Bishwamittra Ghosh, Soumi Das, Till Speicher, Qinyuan Wu, Mohammad Aflah Khan, Deepak Garg, Krishna P. Gummadi, Evimaria Terzi. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Low-Rank Adaptation (LoRA) has emerged as a parameter-efficient approach for fine-tuning large language models.However, conventional LoRA adapters are typically trained for a single task, limiting their applicability in real-world settings, where inputs may span multiple, diverse task domains.At inference time, existing methods can combine multiple LoRAs to improve cross-task performance, but they require additional labeled data or task-specific training, which is expensive at scale.In this work, we introduce LoRA on the Go (LOGO), a training-free framework that dynamically selects and merges adapters at the instance level without any additional requirements.LOGO leverages signals extracted from a single forward pass through LoRA adapters, to identify the most relevant adapters and determine their contributions on-the-fly.Across 5 NLP benchmarks, 27 datasets, and 3 model families, LOGO outperforms training-based baselines on some tasks up to a margin of 3.6% while remaining competitive on other tasks and maintaining inference throughput, highlighting its effectiveness and practicality.

A decentralized autonomous organization (DAO) is a governing entity that empowers its stakeholders (i.e., users who hold one or more of its tokens) to manage blockchain-based protocols (i.e., smart contracts) collaboratively. The governance of a DAO is explicitly encoded in the DAO's governance contract, which defines how stakeholders participate in governance and how much influence (or voting power) they have in any decision. While decentralization and autonomy are the fundamental tenets of a DAO's design, empirical evidence suggests that in practice governance is often highly centralized. In this work, we study the designs and implementations of 48 public and actively used DAOs, with substantially large capital, deployed on Ethereum. We identify how three key governance mechanismstoken registration, staking, and delegation-originally introduced to improve security or participation, contribute to the concentration of voting power. Unlike prior work on centralization of voting power in specific DAOs, our findings reveal that these governance mechanisms of DAOs themselves systematically reinforce centralization. By elucidating the relationship between governance design and voting centralization, this work advances the understanding of DAO governance structures and highlights the inherent trade-offs between decentralization, security, and usability of DAOs.

The GDPR's Right of Access aims to empower users with control over their personal data via Data Download Packages (DDPs). However, their effectiveness is often compromised by inconsistent platform implementations, questionable data reliability, and poor user comprehensibility. This paper conducts a comprehensive audit of DDPs from three social media platforms (TikTok, Instagram, and YouTube) to systematically assess these critical drawbacks. Despite offering similar services, we find that these platforms demonstrate significant inconsistencies in implementing the Right of Access, evident in varying levels of shared data. Critically, the failure to disclose processing purposes, retention periods, and other third-party data recipients serves as a further indicator of non-compliance. Our reliability evaluations, using bots and user-donated data, reveal that while TikTok's DDPs offer more consistent and complete data, others exhibit notable shortcomings. Similarly, our assessment of comprehensibility, based on surveys with 400 participants, indicates that current DDPs substantially fall short of GDPR's standards. To improve the comprehensibility, we propose and demonstrate a two-layered approach by: (1)~enhancing the data representation itself using stakeholder interpretations; and (2)~incorporating a user-friendly extension (\textit{Know Your Data}) for intuitive data visualization where users can control the level of transparency they prefer. Our findings underscore the need for clearer and non-conflicting regulatory guidance, stricter enforcement, and platform commitment to realize the goal of GDPR's Right of Access.

Recent studies have discussed how users are increasingly using conversational AI systems, powered by LLMs, for information seeking, decision support, and even emotional support. However, these macro-level observations offer limited insight into how the purpose of these interactions shifts over time, how users frame their interactions with the system, and how steering dynamics unfold in these human-AI interactions. To examine these evolving dynamics, we gathered and analyzed a unique dataset InVivoGPT: consisting of 825K ChatGPT interactions, donated by 300 users through their GDPR data rights. Our analyses reveal three key findings. First, participants increasingly turn to ChatGPT for a broader range of purposes, including substantial growth in sensitive domains such as health and mental health. Second, interactions become more socially framed: the system anthropomorphizes itself at rising rates, participants more frequently treat it as a companion, and personal data disclosure becomes both more common and more diverse. Third, conversational steering becomes more prominent, especially after the release of GPT-4o, with conversations where the participants followed a model-initiated suggestion quadrupling over the period of our dataset. Overall, our results show that conversational AI systems are shifting from functional tools to social partners, raising important questions about their design and governance.

Online platforms increasingly offer ''paid'' ad-free subscriptions as an alternative to the traditional ''free'' ad-based model. The transition to ad-free models ostensibly removes advertising as a key justification for data processing under the GDPR. So, normatively, platforms should collect less user data. However, platforms may justify continued data collection as a means to provide an improved, personalized experience. This tension between privacy principles and platform incentives raises a critical underexplored question: do data collection practices vary between ad-free and ad-based subscription models?

Annotation and classification of legal text are central components of empirical legal research. Traditionally, these tasks are often delegated to trained research assistants. Motivated by the advances in language modeling, empirical legal scholars are increasingly turning to commercial models, hoping that it will alleviate the significant cost of human annotation. In this work, we present a comprehensive analysis of large language models’ current abilities to perform legal annotation tasks. To do so, we construct CaselawQA, a benchmark comprising 260 legal text classification tasks, nearly all new to the machine learning community. We demonstrate that commercial models, such as GPT-4.5 and Claude 3.7 Sonnet, achieve non-trivial accuracy but generally fall short of the performance required for legal work. We then demonstrate that small, lightly fine-tuned models vastly outperform commercial models. A few dozen to a few hundred labeled examples are usually enough to achieve higher accuracy. Our work points to a viable alternative to the predominant practice of prompting commercial models. For concrete legal annotation tasks with some available labeled data, researchers are likely better off using a fine-tuned open-source model. Code, datasets, and fine-tuned models are available at https://github.com/socialfoundations/lawma.

Opaque algorithms disseminate and mediate the content that users consume on online social media platforms. This algorithmic mediation serves users with contents of their liking, on the other hand, it may cause several inadvertent risks to society at scale. While some of these risks, e.g., filter bubbles or dissemination of hateful content, are well studied in the community, behavioral addiction, designated by the Digital Services Act (DSA) as a potential systemic risk, has been understudied. In this work, we aim to study if one can effectively diagnose behavioral addiction using digital data traces from social media platforms. Focusing on the TikTok short-format video platform as a case study, we employ a novel mixed methodology of combining survey responses with data donations of behavioral traces. We survey 1590 TikTok users and stratify them into three addiction groups (i.e., less/moderately/highly likely addicted). Then, we obtain data donations from 107 surveyed participants. By analyzing users' data we find that, among others, highly likely addicted users spend more time watching TikTok videos and keep coming back to TikTok throughout the day, indicating a compulsion to use the platform. Finally, by using basic user engagement features, we train classifier models to identify highly likely addicted users with F1 >= 0.55. The performance of the classifier models suggests predicting addictive users solely based on their usage is rather difficult.

In this paper, we focus on the challenging task of reliably estimating factual knowledge that is embedded inside large language models (LLMs). To avoid reliability concerns with prior approaches, we propose to eliminate prompt engineering when probing LLMs for factual knowledge. Our approach, called Zero-Prompt Latent Knowledge Estimator (ZP-LKE), leverages the in-context learning ability of LLMs to communicate both the factual knowledge question as well as the expected answer format. Our knowledge estimator is both conceptually simpler (i.e., doesn't depend on meta-linguistic judgments of LLMs) and easier to apply (i.e., is not LLM-specific), and we demonstrate that it can surface more of the latent knowledge embedded in LLMs. We also investigate how different design choices affect the performance of ZP-LKE. Using the proposed estimator, we perform a large-scale evaluation of the factual knowledge of a variety of open-source LLMs, like OPT, Pythia, Llama(2), Mistral, Gemma, etc. over a large set of relations and facts from the Wikidata knowledge base. We observe differences in the factual knowledge between different model families and models of different sizes, that some relations are consistently better known than others but that models differ in the precise facts they know, and differences in the knowledge of base models and their finetuned counterparts. Code available at: https://github.com/QinyuanWu0710/ZeroPrompt_LKE

Interleaving sponsored results (advertisements) amongst organic results on search engine result pages (SERP) has become a common practice across multiple digital platforms. Advertisements have catered to consumer satisfaction and fostered competition in digital public spaces; making them an appealing gateway for businesses to reach their consumers. However, especially in the context of digital marketplaces, due to the competitive nature of the sponsored results with the organic ones, multiple unwanted repercussions have surfaced affecting different stakeholders. From the consumers' perspective the sponsored ads/results may cause degradation of search quality and nudge consumers to potentially irrelevant and costlier products. The sponsored ads may also affect the level playing field of the competition in the marketplaces among sellers. To understand and unravel these potential concerns, we analyse the Amazon digital marketplace in four different countries by simulating 4,800 search operations. Our analyses over SERPs consisting 2M organic and 638K sponsored results show items with poor organic ranks (beyond 100th position) appear as sponsored results even before the top organic results on the first page of Amazon SERP. Moreover, we also observe that in majority of the cases, these top sponsored results are costlier and are of poorer quality than the top organic results. We believe these observations can motivate researchers for further deliberation to bring in more transparency and guard rails in the advertising practices followed in digital marketplaces.

Short-format videos have exploded on platforms like TikTok, Instagram, and YouTube. Despite this, the research community lacks large-scale empirical studies into how people engage with short-format videos and the role of recommendation systems that offer endless streams of such content. In this work, we analyze user engagement on TikTok using data we collect via a data donation system that allows TikTok users to donate their data. We recruited 347 TikTok users and collected 9.2M TikTok video recommendations they received. By analyzing user engagement, we find that the average daily usage time increases over the users’ lifetime while the user attention remains stable at around 45%. We also find that users like more videos uploaded by people they follow than those recommended by people they do not follow. Our study offers valuable insights into how users engage with short-format videos on TikTok and lessons learned from designing a data donation system.

In recent years, user feeds on social media platforms have shifted from simple, chronologically ordered content posted by their network connections (i.e., friends) to opaque, algorithmically ordered and curated content. This shift has led to regulations that require platforms to offer end users greater transparency and control over their algorithmic recommendation-based feeds. In response, social media platforms such as TikTok have recently started explaining why specific videos are recommended to end users. However, we still lack a good understanding of how these explanations are generated and whether they offer the desired transparency to end users. In this work, we audit explanations provided on short-format videos on TikTok. We collect a large dataset of short-format videos and explanations provided by TikTok (when available) using automated sockpuppet accounts. Then, we systematically characterize the explanations, focusing on their accuracy and comprehensiveness. For our assessments, we compare the provided explanations with video metadata and the behavior of our sockpuppet accounts. Our analysis shows that some generic (non-personalized) reasons are always included in explanations (e.g., "This video is popular in your country"), while at the same time, we find that a large number of provided explanations are incompatible with the behavior of our sockpuppet accounts; (e.g., an account that made zero comments on the platform, was presented with the explanation "You commented on similar videos" in 34% of all recommended videos.) Overall, our audit of TikTok video explanations highlights the need for more accurate, fine-grained, and useful explanations for the end users. We will make our code and dataset available to assist the research community.

Recommendation algorithms for social media feeds often function as black boxes from the perspective of users. We aim to detect whether social media feed recommendations are personalized to users, and to characterize the factors contributing to personalization in these feeds. We introduce a general framework to examine a set of social media feed recommendations for a user as a timeline. We label items in the timeline as the result of exploration vs. exploitation of the user's interests on the part of the recommendation algorithm and introduce a set of metrics to capture the extent of personalization across user timelines. We apply our framework to a real TikTok dataset and validate our results using a baseline generated from automated TikTok bots, as well as a randomized baseline. We also investigate the extent to which factors such as video viewing duration, liking, and following drive the personalization of content on TikTok. Our results demonstrate that our framework produces intuitive and explainable results, and can be used to audit and understand personalization in social media feeds.

E-commerce marketplaces provide business opportunities to millions of sellers worldwide. Some of these sellers have special relationships with the marketplace by virtue of using their subsidiary services (e.g., fulfillment and/or shipping services provided by the marketplace) -- we refer to such sellers collectively as Related Sellers. When multiple sellers offer to sell the same product, the marketplace helps a customer in selecting an offer (by a seller) through (a) a default offer selection algorithm, (b) showing features about each of the offers and the corresponding sellers (price, seller performance metrics, seller's number of ratings etc.), and (c) finally evaluating the sellers along these features. In this paper, we perform an end-to-end investigation into how the above apparatus can nudge customers toward the Related Sellers on Amazon's four different marketplaces in India, USA, Germany and France. We find that given explicit choices, customers' preferred offers and algorithmically selected offers can be significantly different. We highlight that Amazon is adopting different performance metric evaluation policies for different sellers, potentially benefiting Related Sellers. For instance, such policies result in notable discrepancy between the actual performance metric and the presented performance metric of Related Sellers. We further observe that among the seller-centric features visible to customers, sellers' number of ratings influences their decisions the most, yet it may not reflect the true quality of service by the seller, rather reflecting the scale at which the seller operates, thereby implicitly steering customers toward larger Related Sellers. Moreover, when customers are shown the rectified metrics for the different sellers, their preference toward Related Sellers is almost halved. We believe our findings will inform and encourage further deliberation toward more effective governance of such design choices and policies adopted by e-commerce marketplaces.

Prior research in psychology has found that people's decisions are often inconsistent. An individual's decisions vary across time, and decisions vary even more across people. Inconsistencies have been identified not only in subjective matters, like matters of taste, but also in settings one might expect to be more objective, such as sentencing, job performance evaluations, or real estate appraisals. In our study, we explore whether algorithmic decision aids can be used to moderate the degree of inconsistency in human decision-making in the context of real estate appraisal. In a large-scale human-subject experiment, we study how different forms of algorithmic assistance influence the way that people review and update their estimates of real estate prices. We find that both (i) asking respondents to review their estimates in a series of algorithmically chosen pairwise comparisons and (ii) providing respondents with traditional machine advice are effective strategies for influencing human responses. Compared to simply reviewing initial estimates one by one, the aforementioned strategies lead to (i) a higher propensity to update initial estimates, (ii) a higher accuracy of post-review estimates, and (iii) a higher degree of consistency between the post-review estimates of different respondents. While these effects are more pronounced with traditional machine advice, the approach of reviewing algorithmically chosen pairs can be implemented in a wider range of settings, since it does not require access to ground truth data.

An evaluation criterion for safe and trustworthy deep learning is how well the invariances captured by representations of deep neural networks (DNNs) are shared with humans. We identify challenges in measuring these invariances. Prior works used gradient-based methods to generate identically represented inputs (IRIs), ie, inputs which have identical representations (on a given layer) of a neural network, and thus capture invariances of a given network. One necessary criterion for a network's invariances to align with human perception is for its IRIs look 'similar' to humans. Prior works, however, have mixed takeaways; some argue that later layers of DNNs do not learn human-like invariances yet others seem to indicate otherwise. We argue that the loss function used to generate IRIs can heavily affect takeaways about invariances of the network and is the primary reason for these conflicting findings. We propose an adversarial regularizer on the IRI generation loss that finds IRIs that make any model appear to have very little shared invariance with humans. Based on this evidence, we argue that there is scope for improving models to have human-like invariances, and further, to have meaningful comparisons between models one should use IRIs generated using the regularizer-free loss. We then conduct an in-depth investigation of how different components (eg architectures, training losses, data augmentations) of the deep learning pipeline contribute to learning models that have good alignment with humans. We find that architectures with residual connections trained using a (self-supervised) contrastive loss with l_p ball adversarial data augmentation tend to learn invariances that are most aligned with humans. Code: github.com/nvedant07/Human-NN-Alignment

Abstract In permissionless blockchains, transaction issuers include a fee to incentivize miners to include their transactions. To accurately estimate this prioritization fee for a transaction, transaction issuers (or blockchain participants, [email protected] generally) rely on two fundamental notions of transparency, namely contention and prioritization transparency. Contention transparency implies that participants are aware of every pending transaction that will contend with a given transaction for inclusion. Prioritization transparency states that the participants are aware of the transaction or prioritization fees paid by every such contending transaction. Neither of these notions of transparency holds well today. Private relay networks, for instance, allow users to send transactions privately to miners. Besides, users can offer fees to miners via either direct transfers to miners’ wallets or off-chain payments—neither of which are public. In this work, we characterize the lack of contention and prioritization transparency in Bitcoin and Ethereum resulting from such practices. We show that private relay networks are widely used and private transactions are quite prevalent. We show that the lack of transparency facilitates miners to collude and overcharge users who may use these private relay networks despite them offering little to no guarantees on transaction prioritization. The lack of these transparencies in blockchains has crucial implications for transaction issuers as well as the stability of blockchains. Finally, we make our data sets and scripts publicly available.

During the COVID-19 pandemic, health-related misinformation and harmful content shared online had a significant adverse effect on society. In an attempt to mitigate this adverse effect, mainstream social media platforms like Facebook, Twitter, and TikTok employed soft moderation interventions (i.e., warning labels) on potentially harmful posts. Such interventions aim to inform users about the post's content without removing it, hence easing the public's concerns about censorship and freedom of speech. Despite the recent popularity of these moderation interventions, as a research community, we lack empirical analyses aiming to uncover how these warning labels are used in the wild, particularly during challenging times like the COVID-19 pandemic. In this work, we analyze the use of warning labels on TikTok, focusing on COVID-19 videos. First, we construct a set of 26 COVID-19 related hashtags, and then we collect 41K videos that include those hashtags in their description. Second, we perform a quantitative analysis on the entire dataset to understand the use of warning labels on TikTok. Then, we perform an in-depth qualitative study, using thematic analysis, on 222 COVID-19 related videos to assess the content and the connection between the content and the warning labels. Our analysis shows that TikTok broadly applies warning labels on TikTok videos, likely based on hashtags included in the description (e.g., 99% of the videos that contain #coronavirus have warning labels). More worrying is the addition of COVID-19 warning labels on videos where their actual content is not related to COVID-19 (23% of the cases in a sample of 143 English videos that are not related to COVID-19). Finally, our qualitative analysis on a sample of 222 videos shows that 7.7% of the videos share misinformation/harmful content and do not include warning labels, 37.3% share benign information and include warning labels, and that 35% of the videos that share misinformation/harmful content (and need a warning label) are made for fun. Our study demonstrates the need to develop more accurate and precise soft moderation systems, especially on a platform like TikTok that is extremely popular among people of younger age.

Representations learned by pre-training a neural network on a large dataset are increasingly used successfully to perform a variety of downstream tasks. In this work, we take a closer look at how features are encoded in such pre-trained representations. We find that learned representations in a given layer exhibit a degree of diffuse redundancy, ie, any randomly chosen subset of neurons in the layer that is larger than a threshold size shares a large degree of similarity with the full layer and is able to perform similarly as the whole layer on a variety of downstream tasks. For example, a linear probe trained on $20\%$ of randomly picked neurons from the penultimate layer of a ResNet50 pre-trained on ImageNet1k achieves an accuracy within $5\%$ of a linear probe trained on the full layer of neurons for downstream CIFAR10 classification. We conduct experiments on different neural architectures (including CNNs and Transformers) pre-trained on both ImageNet1k and ImageNet21k and evaluate a variety of downstream tasks taken from the VTAB benchmark. We find that the loss \& dataset used during pre-training largely govern the degree of diffuse redundancy and the "critical mass" of neurons needed often depends on the downstream task, suggesting that there is a task-inherent redundancy-performance Pareto frontier. Our findings shed light on the nature of representations learned by pre-trained deep neural networks and suggest that entire layers might not be necessary to perform many downstream tasks. We investigate the potential for exploiting this redundancy to achieve efficient generalization for downstream tasks and also draw caution to certain possible unintended consequences. Our code is available at \url{https://github.com/nvedant07/diffused-redundancy}.

Abstract Reliably predicting potential failure risks of machine learning (ML) systems when deployed with production data is a crucial aspect of trustworthy AI. This paper introduces the Risk Advisor, a novel post-hoc meta-learner for estimating failure risks and predictive uncertainties of any already-trained black-box classification model. In addition to providing a risk score, the Risk Advisor decomposes the uncertainty estimates into aleatoric and epistemic uncertainty components, thus giving informative insights into the sources of uncertainty inducing the failures. Consequently, Risk Advisor can distinguish between failures caused by data variability, data shifts and model limitations and provide useful guidance on appropriate risk mitigation actions (e.g., collecting more data to counter data shift). Extensive experiments on various families of black-box classification models and on real-world and synthetic datasets covering common ML failure scenarios show that the Risk Advisor reliably predicts deployment-time failure risks in all the scenarios, and outperforms strong baselines.

Related item recommendations (RIRs) are ubiquitous in most online platforms today, including e-commerce and content streaming sites. These recommendations not only help users compare items related to a given item but also play a major role in bringing traffic to individual items, thus deciding the exposure that different items receive. With a growing number of people depending on such platforms to earn their livelihood, it is important to understand whether different items are receiving their desired exposure. To this end, our experiments on multiple real-world RIR datasets reveal that the existing RIR algorithms often result in very skewed exposure distribution of items, and the quality of items is not a plausible explanation for such skew in exposure. To mitigate this exposure bias, we introduce multiple flexible interventions [fair related item recommendation (FaiRIR)] in the RIR pipeline. We instantiate these mechanisms with two well-known algorithms for constructing RIRs—rating singular value decomposition (SVD) and item2vec—and show on real-world data that our mechanisms allow for a fine-grained control on exposure distribution, often at a small or no cost in terms of recommendation quality, measured in terms of relatedness and user satisfaction.

Influence maximization has found applications in a wide range of real-world problems, for instance, viral marketing of products in an online social network, and information propagation of valuable information such as job vacancy advertisements and health-related information. While existing algorithmic techniques usually aim at maximizing the total number of people influenced, the population often comprises several socially salient groups, e.g., based on gender or race. As a result, these techniques could lead to disparity across different groups in receiving important information. Furthermore, in many of these applications, the spread of influence is time-critical, i.e., it is only beneficial to be influenced before a time deadline. As we show in this paper, the time-criticality of the information could further exacerbate the disparity of influence across groups. This disparity, introduced by algorithms aimed at maximizing total influence, could have far-reaching consequences, impacting people's prosperity and putting minority groups at a big disadvantage. In this work, we propose a notion of group fairness in time-critical influence maximization. We introduce surrogate objective functions to solve the influence maximization problem under fairness considerations. By exploiting the submodularity structure of our objectives, we provide computationally efficient algorithms with guarantees that are effective in enforcing fairness during the propagation process. We demonstrate the effectiveness of our approach through synthetic and real-world experiments.

The potential for machine learning systems to amplify social inequities and unfairness is receiving increasing popular and academic attention. Much recent work has focused on developing algorithmic tools to assess and mitigate such unfairness. However, there is little work on enhancing fairness in graph algorithms. Here, we develop a simple, effective and general method, CrossWalk, that enhances fairness of various graph algorithms, including influence maximization, link prediction and node classification, applied to node embeddings. CrossWalk is applicable to any random walk based node representation learning algorithm, such as DeepWalk and Node2Vec. The key idea is to bias random walks to cross group boundaries, by upweighting edges which (1) are closer to the groups’ peripheries or (2) connect different groups in the network. CrossWalk pulls nodes that are near groups’ peripheries towards their neighbors from other groups in the embedding space, while preserving the necessary structural properties of the graph. Extensive experiments show the effectiveness of our algorithm to enhance fairness in various graph algorithms, including influence maximization, link prediction and node classification in synthetic and real networks, with only a very small decrease in performance.

Machine learning algorithms are increasingly used to assist human decision-making. When the goal of machine assistance is to improve the accuracy of human decisions, it might seem appealing to design ML algorithms that complement human knowledge. While neither the algorithm nor the human are perfectly accurate, one could expect that their complementary expertise might lead to improved outcomes. In this study, we demonstrate that in practice decision aids that are not complementary, but make errors similar to human ones may have their own benefits. In a series of human-subject experiments with a total of 901 participants, we study how the similarity of human and machine errors influences human perceptions of and interactions with algorithmic decision aids. We find that (i) people perceive more similar decision aids as more useful, accurate, and predictable, and that (ii) people are more likely to take opposing advice from more similar decision aids, while (iii) decision aids that are less similar to humans have more opportunities to provide opposing advice, resulting in a higher influence on people’s decisions overall.

Influence maximization has found applications in a wide range of real-world problems, for instance, viral marketing of products in an online social network, and propagation of valuable information such as job vacancy advertisements. While existing algorithmic techniques usually aim at maximizing the total number of people influenced, the population often comprises several socially salient groups, e.g., based on gender or race. As a result, these techniques could lead to disparity across different groups in receiving important information. Furthermore, in many applications, the spread of influence is time-critical, i.e., it is only beneficial to be influenced before a deadline. As we show in this paper, such time-criticality of information could further exacerbate the disparity of influence across groups. This dis-parity could have far-reaching consequences, impacting people's prosperity and putting minority groups at a big disadvantage. In this work, we propose a notion of group fairness in time-critical influence maximization. We introduce surrogate objective functions to solve the influence maximization problem under fair-ness considerations. By exploiting the submodularity structure of our objectives, we provide computationally efficient algorithms with guarantees that are effective in enforcing fairness during the propagation process. Extensive experiments on synthetic and real-world datasets demonstrate the efficacy of our proposal.

A major challenge in studying robustness in deep learning is defining the set of “meaningless” perturbations to which a given Neural Network (NN) should be invariant. Most work on robustness implicitly uses a human as the reference model to define such perturbations. Our work offers a new view on robustness by using another reference NN to define the set of perturbations a given NN should be invariant to, thus generalizing the reliance on a reference “human NN” to any NN. This makes measuring robustness equivalent to measuring the extent to which two NNs share invariances. We propose a measure called \stir, which faithfully captures the extent to which two NNs share invariances. \stir re-purposes existing representation similarity measures to make them suitable for measuring shared invariances. Using our measure, we are able to gain insights about how shared invariances vary with changes in weight initialization, architecture, loss functions, and training dataset. Our implementation is available at: \url{https://github.com/nvedant07/STIR}.

In traditional (desktop) e-commerce search, a customer issues a specific query and the system returns a ranked list of products in order of relevance to the query. An increasingly popular alternative in e-commerce search is to issue a voice-query to a smart speaker (e.g., Amazon Echo) powered by a voice assistant (VA, e.g., Alexa). In this situation, the VA usually spells out the details of only one product, an explanation citing the reason for its selection, and a default action of adding the product to the customer’s cart. This reduced autonomy of the customer in the choice of a product during voice-search makes it necessary for a VA to be far more responsible and trustworthy in its explanation and default action.

Recently, almost all conferences have moved to virtual mode due to the pandemic-induced restrictions on travel and social gathering. Contrary to in-person conferences, virtual conferences face the challenge of efficiently scheduling talks, accounting for the availability of participants from different timezones and their interests in attending different talks. A natural objective for conference organizers is to maximize efficiency, e.g., total expected audience participation across all talks. However, we show that optimizing for efficiency alone can result in an unfair virtual conference schedule, where individual utilities for participants and speakers can be highly unequal. To address this, we formally define fairness notions for participants and speakers, and derive suitable objectives to account for them. As the efficiency and fairness objectives can be in conflict with each other, we propose a joint optimization framework that allows conference organizers to design schedules that balance (i.e., allow trade-offs) among efficiency, participant fairness and speaker fairness objectives. While the optimization problem can be solved using integer programming to schedule smaller conferences, we provide two scalable techniques to cater to bigger conferences. Extensive evaluations over multiple real-world datasets show the efficacy and flexibility of our proposed approaches.

Misinformation on social media has become a critical problem, particularly during a public health pandemic. Most social platforms today rely on users’ voluntary reports to determine which news stories to fact-check first. Despite the importance, no prior work has explored the potential biases in such a reporting process. This work proposes a novel methodology to assess how users perceive truth or misinformation in online news stories. By conducting a large-scale survey ($N =15$000), we identify the possible biases in news perceptions and explore how partisan leanings influence the news selection algorithm for fact checking. Our survey reveals several perception biases or inaccuracies in estimating the truth level of stories. The first kind, called the total perception bias (TPB), is the aggregate difference in the ground truth and perceived truth level. The next two are the false-positive bias (FPB) and false-negative bias (FNB), which measures users’ gullibility and cynicality of a given claim. We also propose ideological mean perception bias (IMPB), which quantifies a news story’s ideological disputability. Collectively, these biases indicate that user perceptions are not correlated with the ground truth of new stories; users believe some stories to be more false and vice versa. This calls for the need to fact-check news stories that exhibit the most considerable perception biases first, which the current voluntary reporting does not offer. Based on these observations, we propose a new framework that can best leverage users’ truth perceptions to remove false stories, correct misperceptions of users, or decrease ideological disagreements. We discuss how this new prioritizing scheme can aid platforms to significantly reduce the impact of fake news on user beliefs.

Many online platforms today (such as Amazon, Netflix, Spotify, LinkedIn, and AirBnB) can be thought of as two-sided markets with producers and customers of goods and services. Traditionally, recommendation services in these platforms have focused on maximizing customer satisfaction by tailoring the results according to the personalized preferences of individual customers. However, our investigation reinforces the fact that such customer-centric design of these services may lead to unfair distribution of exposure to the producers, which may adversely impact their well-being. However, a pure producer-centric design might become unfair to the customers. As more and more people are depending on such platforms to earn a living, it is important to ensure fairness to both producers and customers. In this work, by mapping a fair personalized recommendation problem to a constrained version of the problem of fairly allocating indivisible goods, we propose to provide fairness guarantees for both sides. Formally, our proposed FairRec algorithm guarantees Maxi-Min Share of exposure for the producers, and Envy-Free up to One Item fairness for the customers. Extensive evaluations over multiple real-world datasets show the effectiveness of FairRec in ensuring two-sided fairness while incurring a marginal loss in overall recommendation quality. Finally, we present a modification of FairRec (named as FairRecPlus ) that at the cost of additional computation time, improves the recommendation performance for the customers, while maintaining the same fairness guarantees.

Traditional approaches to ensure group fairness in algorithmic decision making aim to equalize "total" error rates for different subgroups in the population. In contrast, we argue that the fairness approaches should instead focus only on equalizing errors arising due to model uncertainty (a.k.a epistemic uncertainty), caused due to lack of knowledge about the best model or due to lack of data. In other words, our proposal calls for ignoring the errors that occur due to uncertainty inherent in the data, i.e., aleatoric uncertainty. We draw a connection between predictive multiplicity and model uncertainty and argue that the techniques from predictive multiplicity could be used to identify errors made due to model uncertainty. We propose scalable convex proxies to come up with classifiers that exhibit predictive multiplicity and empirically show that our methods are comparable in performance and up to four orders of magnitude faster than the current state-of-the-art. We further pro- pose methods to achieve our goal of equalizing group error rates arising due to model uncertainty in algorithmic decision making and demonstrate the effectiveness of these methods using synthetic and real-world datasets

Prior work on fairness in machine learning has focused on settings where all the information needed about each individual is readily available. However, in many applications, further information may be acquired at a cost. For example, when assessing a customer's creditworthiness, a bank initially has access to a limited set of information but progressively improves the assessment by acquiring additional information before making a final decision. In such settings, we posit that a fair decision maker may want to ensure that decisions for all individuals are made with similar expected error rate, even if the features acquired for the individuals are different. We show that a set of carefully chosen confidence thresholds can not only effectively redistribute an information budget according to each individual's needs, but also serve to address individual and group fairness concerns simultaneously. Finally, using two public datasets, we confirm the effectiveness of our methods and investigate the limitations.

Reliably predicting potential failure risks of machine learning (ML) systems when deployed with production data is a crucial aspect of trustworthy AI. This paper introduces Risk Advisor, a novel post-hoc meta-learner for estimating failure risks and predictive uncertainties of any already-trained black-box classification model. In addition to providing a risk score, the Risk Advisor decomposes the uncertainty estimates into aleatoric and epistemic uncertainty components, thus giving informative insights into the sources of uncertainty inducing the failures. Consequently, Risk Advisor can distinguish between failures caused by data variability, data shifts and model limitations and advise on mitigation actions (e.g., collecting more data to counter data shift). Extensive experiments on real-world datasets covering a variety of ML failure scenarios show that the Risk Advisor reliably predicts deployment-time failure risks in all the scenarios, and outperforms strong baselines.

Most public blockchain protocols, including the popular Bitcoin and Ethereum blockchains, do not formally specify the order in which miners should select transactions from the pool of pending (or uncommitted) transactions for inclusion in the blockchain. Over the years, informal conventions or "norms" for transaction ordering have, however, emerged via the use of shared software by miners, e.g., the GetBlockTemplate (GBT) mining protocol in Bitcoin Core. Today, a widely held view is that Bitcoin miners prioritize transactions based on their offered "transaction fee-per-byte." Bitcoin users are, consequently, encouraged to increase the fees to accelerate the commitment of their transactions, particularly during periods of congestion. In this paper, we audit the Bitcoin blockchain and present statistically significant evidence of mining pools deviating from the norms to accelerate the commitment of transactions for which they have (i) a selfish or vested interest, or (ii) received dark-fee payments via opaque (non-public) side-channels. As blockchains are increasingly being used as a record-keeping substrate for a variety of decentralized (financial technology) systems, our findings call for an urgent discussion on defining neutrality norms that miners must adhere to when ordering transactions in the chains. Finally, we make our data sets and scripts publicly available.

In this paper, we focus on auditing black-box prediction models for compliance with the GDPR’s data minimization principle. This principle restricts prediction models to use the minimal information that is necessary for performing the task at hand. Given the challenge of the black-box setting, our key idea is to check if each of the prediction model’s input features is individually necessary by assigning it some constant value (i.e., applying a simple imputation) across all prediction instances, and measuring the extent to which the model outcomes would change. We introduce a metric for data minimization that is based on model instability under simple imputations. We extend the applicability of this metric from a finite sample model to a distributional setting by introducing a probabilistic data minimization guarantee, which we derive using a Bayesian approach. Furthermore, we address the auditing problem under a constraint on the number of queries to the prediction system. We formulate the problem of allocating a budget of system queries to feasible simple imputations (for investigating model instability) as a multi-armed bandit framework with probabilistic success metrics. We define two bandit problems for providing a probabilistic data minimization guarantee at a given confidence level: a decision problem given a data minimization level, and a measurement problem given a fixed query budget. We design efficient algorithms for these auditing problems using novel exploration strategies that expand classical bandit strategies. Our experiments with real-world prediction systems show that our auditing algorithms significantly outperform simpler benchmarks in both measurement and decision problems.

Public schools in the United States offer tuition-free primary and secondary education to their students, and are divided into school districts funded by the local and state governments. Although the primary source of school district revenue is public money, several studies have pointed to the inequality in funding across different school districts. In this paper, we focus on the spatial geometry/distribution of such inequality, i.e., how the highly funded and lesser funded school districts are located relative to each other. Due to the major reliance on local property taxes for school funding, we find existing school district boundaries promoting financial segregation, with highly-funded school districts surrounded by lesser-funded districts and vice-versa.

Major online platforms today can be thought of as two-sided markets with producers and customers of goods and services. There have been concerns that over-emphasis on customer satisfaction by the platforms may affect the well-being of the producers. To counter such issues, few recent works have attempted to incorporate fairness for the producers. However, these studies have overlooked an important issue in such platforms -- to supposedly improve customer utility, the underlying algorithms are frequently updated, causing abrupt changes in the exposure of producers. In this work, we focus on the fairness issues arising out of such frequent updates, and argue for incremental updates of the platform algorithms so that the producers have enough time to adjust (both logistically and mentally) to the change. However, naive incremental updates may become unfair to the customers. Thus focusing on recommendations deployed on two-sided platforms, we formulate an ILP based online optimization to deploy changes incrementally in η steps, where we can ensure smooth transition of the exposure of items while guaranteeing a minimum utility for every customer. Evaluations over multiple real world datasets show that our proposed mechanism for platform updates can be efficient and fair to both the producers and the customers in two-sided platforms.

Quota-based fairness mechanisms like the so-called Rooney rule or four-fifths rule are used in selection problems such as hiring or college admission to reduce inequalities based on sensitive demographic attributes (gender, ethnicity, etc.). These mechanisms are often viewed as introducing a trade-off between selection fairness and utility (i.e., the overall quality of the selected candidates). In recent work, however, Kleinberg and Raghavan [\emphProc. of ITCS '18 ] showed that, in the presence of implicit bias in estimating candidates' quality, the Rooney rule can in fact increase the utility of the selection process (beyond improving its fairness). We argue that even in the absence of implicit bias, the estimates of candidates' quality from different groups may differ in another fundamental way, namely, in their variance. We term this phenomenon implicit variance and we ask: can fairness mechanisms be beneficial to the utility of a selection process in the presence of implicit variance (even in the absence of implicit bias)? To answer this question, we propose a simple model in which candidates have a true latent quality that is drawn from a group-independent normal distribution. To make the selection, a decision maker receives an unbiased estimate of the quality of each candidate, with normal noise, but whose variance depends on the candidate's group. We then compare the utility obtained by imposing a fairness mechanism that we term γ-rule, which includes demographic parity (γ = 1$) and the four-fifths rule (γ = 0.8$) as special cases, to that of a group-oblivious baseline selection algorithm that simply picks the candidates with the highest estimated quality independently of their group. Our main result shows that the demographic parity mechanism always strictly increases the selection utility, while any other γ-rule also always increases it weakly. We extend our model to a two-stage selection process where the true quality is observed at the second stage and analyze how our results are changed in that case. We finally discuss multiple extensions of our results, in particular to different distributions of the true latent quality.

We investigate the problem of fair recommendation in the context of two-sided online platforms, comprising customers on one side and producers on the other. Traditionally, recommendation services in these platforms have focused on maximizing customer satisfaction by tailoring the results according to the personalized preferences of individual customers. However, our investigation reveals that such customer-centric design may lead to unfair distribution of exposure among the producers, which may adversely impact their well-being. On the other hand, a producer-centric design might become unfair to the customers. Thus, we consider fairness issues that span both customers and producers. Our approach involves a novel mapping of the fair recommendation problem to a constrained version of the problem of fairly allocating indivisible goods. Our proposed FairRec algorithm guarantees at least Maximin Share (MMS) of exposure for most of the producers and Envy-Free up to One Good (EF1) fairness for every customer. Extensive evaluations over multiple real-world datasets show the effectiveness of FairRec in ensuring two-sided fairness while incurring a marginal loss in the overall recommendation quality.

Online donation platforms help equalize access to opportunity and funding in cases where inequalities exist. In the context of public school education in the United States, for instance, financial inequalities have been shown to be reflected in the educational system, since schools are primarily funded through local property taxes. In response, private charitable donation platforms such as DonorsChoose.org have emerged seeking to alleviate systemic inequalities. Yet, the question remains of how effective these platforms are in redressing existing funding inequalities across school districts. Our analysis of donation data from DonorsChoose shows that such platforms may in fact be ineffective in mitigating existing inequalities or may even exacerbate them. In this paper, we explore how online educational charities could direct more funding towards more impoverished schools without compromising their donors' freedom of choice with respect to donation targets. Seeking to answer this question, we draw on the line of work on choice architectures in behavioral economics and pose a novel research question on the impact of interface design on equity in socio-technical systems. Through controlled experiments, we demonstrate how simple interface design interventions - such as modifying default rankings or displaying additional information about schools - might lead to changes in donation distributions helping platforms direct more funding towards schools in need. Going beyond online educational charities, we hope that our work will bring attention to the role of interface design nudges in the social requirements of online altruism.

The use of algorithmic (learning-based) decision making in scenarios that affect human lives has motivated a number of recent studies to investigate such decision making systems for potential unfairness, such as discrimination against subjects based on their sensitive features like gender or race. However, when judging the fairness of a newly designed decision making system, these studies have overlooked an important influence on people's perceptions of fairness, which is how the new algorithm changes the status quo, i.e., decisions of the existing decision making system. Motivated by extensive literature in behavioral economics and behavioral psychology (prospect theory), we propose a notion of fair updates that we refer to as loss-averse updates. Loss-averse updates constrain the updates to yield improved (more beneficial) outcomes to subjects compared to the status quo. We propose tractable proxy measures that would allow this notion to be incorporated in the training of a variety of linear and non-linear classifiers. We show how our proxy measures can be combined with existing measures for training nondiscriminatory classifiers.Our evaluation using synthetic and real-world datasets demonstrates that the proposed proxy measures are effective for their desired tasks.

People are rated and ranked, towards algorithmic decision making in an increasing number of applications, typically based on machine learning. Research on how to incorporate fairness into such tasks has prevalently pursued the paradigm of group fairness: giving adequate success rates to specifically protected groups. In contrast, the alternative paradigm of individual fairness has received relatively little attention, and this paper advances this less explored direction. The paper introduces a method for probabilistically mapping user records into a low-rank representation that reconciles individual fairness and the utility of classifiers and rankings in downstream applications. Our notion of individual fairness requires that users who are similar in all task-relevant attributes such as job qualification, and disregarding all potentially discriminating attributes such as gender, should have similar outcomes. We demonstrate the versatility of our method by applying it to classification and learning-to-rank tasks on a variety of real-world datasets. Our experiments show substantial improvements over the best prior work for this setting.

Most existing notions of algorithmic fairness are one-shot: they ensure some form of allocative equality at the time of decision making, but do not account for the adverse impact of the algorithmic decisions today on the long-term welfare and prosperity of certain segments of the population. We take a broader perspective on algorithmic fairness. We propose an effort-based measure of fairness and present a data-driven framework for characterizing the long-term impact of algorithmic policies on reshaping the underlying population. Motivated by the psychological literature on social learning and the economic literature on equality of opportunity, we propose a micro-scale model of how individuals may respond to decision-making algorithms. We employ existing measures of segregation from sociology and economics to quantify the resulting macro- scale population-level change. Importantly, we observe that different models may shift the group- conditional distribution of qualifications in different directions. Our findings raise a number of important questions regarding the formalization of fairness for decision-making models.

The rise of algorithmic decision making led to active researches on how to define and guarantee fairness, mostly focusing on one-shot decision making. In several important applications such as hiring, however, decisions are made in multiple stage with additional information at each stage. In such cases, fairness issues remain poorly understood. In this paper we study fairness in k-stage selection problems where additional features are observed at every stage. We first introduce two fairness notions, local (per stage) and global (final stage) fairness, that extend the classical fairness notions to the k-stage setting. We propose a simple model based on a probabilistic formulation and show that the locally and globally fair selections that maximize precision can be computed via a linear program. We then define the price of local fairness to measure the loss of precision induced by local constraints; and investigate theoretically and empirically this quantity. In particular, our experiments show that the price of local fairness is generally smaller when the sensitive attribute is observed at the first stage; but globally fair selections are more locally fair when the sensitive attribute is observed at the second stage – hence in both cases it is often possible to have a selection that has a small price of local fairness and is close to locally fair.

Ride hailing platforms, such as Uber, Lyft, Ola or DiDi, have traditionally focused on the satisfaction of the passengers, or on boosting successful business transactions. However, recent studies provide a multitude of reasons to worry about the drivers in the ride hailing ecosystem. The concerns range from bad working conditions and worker manipulation to discrimination against minorities. With the sharing economy ecosystem growing, more and more drivers financially depend on online platforms and their algorithms to secure a living. It is pertinent to ask what a fair distribution of income on such platforms is and what power and means the platform has in shaping these distributions.

As data science and artificial intelligence become ubiquitous, they have an increasing impact on society. While many of these impacts are beneficial, others may not be. So understanding and managing these impacts is required of every responsible data scientist. Nevertheless, most human decision-makers use algorithms for efficiency purposes and not to make a better (i.e., fairer) decisions. Even the task of risk assessment in the criminal justice system enables efficiency instead of (and often at the expense of) fairness. So we need to frame the problem with fairness, and other societal impacts, as primary objectives. In this context, most attention has been paid to the machine learning of a model for a task, such as recognition, prediction, or classification. However, issues arise in all parts of the data eco-system, from data acquisition to data presentation. For example, the majority of the population is not white and male, yet this demographic is over-represented in the training data. It is challenging for a data scientist to satisfactorily discharge this broad responsibility.

The increasing role of recommender systems in many aspects of society makes it essential to consider how such systems may impact social good. Various modifications to recommendation algorithms have been proposed to improve their performance for specific socially relevant measures. However, previous proposals are often not easily adapted to different measures, and they generally require the ability to modify either existing system inputs, the system's algorithm, or the system's outputs. As an alternative, in this paper we introduce the idea of improving the social desirability of recommender system outputs by adding more data to the input, an approach we view as as providing 'antidote' data to the system. We formalize the antidote data problem, and develop optimization-based solutions. We take as our model system the matrix factorization approach to recommendation, and we propose a set of measures to capture the polarization or fairness of recommendations. We then show how to generate antidote data for each measure, pointing out a number of computational efficiencies, and discuss the impact on overall system accuracy. Our experiments show that a modest budget for antidote data can lead to significant improvements in the polarization or fairness of recommendations.

Online donation platforms, such as DonorsChoose, GlobalGiving, or CrowdFunder, enable donors to financially support entities in need. In a typical scenario, after a fundraiser submits a request specifying her need, donors contribute financially to help raise the target amount within a pre-specified timeframe. While the goal of such platforms is to counterbalance societal inequalities, biased donation trends might exacerbate the unfair distribution of resources to those in need. Prior research has looked at the impact of biased data, models, or human behavior on inequality in different socio-technical systems, while largely ignoring the choice architecture, in which the funding decisions are made.

Data brokers such as Acxiom and Experian are in the business of collecting and selling data on people; the data they sell is commonly used to feed marketing as well as political campaigns. Despite the ongoing privacy debate, there is still very limited visibility into data collection by data brokers. Recently, however, online advertising services such as Facebook have begun to partner with data brokers-to add additional targeting features to their platform- providing avenues to gain insight into data broker information.

Online news media sites are emerging as the primary source of news for a large number of users. Due to a large number of stories being published in these media sites, users usually rely on news recommendation systems to find important news. In this work, we focus on automatically recommending news stories to all users of such media websites, where the selection is not influenced by a particular user’s news reading habit. When recommending news stories in such non-personalized manner, there are three basic metrics of interest—recency, importance (analogous to relevance in personalized recommendation) and diversity of the recommended news. Ideally, recommender systems should recommend the most important stories soon after they are published. However, the importance of a story only becomes evident as the story ages, thereby creating a tension between recency and importance. A systematic analysis of popular recommendation strategies in use today reveals that they lead to poor trade-offs between recency and importance in practice. So, in this paper, we propose a new recommendation strategy (called Highest Future-Impact ) which attempts to optimize on both the axes. To implement our proposed strategy in practice, we propose two approaches to predict the future-impact of news stories, by using crowd-sourced popularity signals and by observing editorial selection in past news data. Finally, we propose approaches to inculcate diversity in recommended news which can maintain a balanced proportion of news from different news sections. Evaluations over real-world news datasets show that our implementations achieve good performance in recommending news stories.

Users frequently use search systems on the Web as well as online social media to learn about ongoing events and public opinion on personalities. Prior studies have shown that the top-ranked results returned by these search engines can shape user opinion about the topic (e.g., event or person) being searched. In case of polarizing topics like politics, where multiple competing perspectives exist, the political bias in the top search results can play a significant role in shaping public opinion towards (or away from) certain perspectives. Given the considerable impact that search bias can have on the user, we propose a generalizable search bias quantification framework that not only measures the political bias in ranked list output by the search system but also decouples the bias introduced by the different sources—input data and ranking system. We apply our framework to study the political bias in searches related to 2016 US Presidential primaries in Twitter social media search and find that both input data and ranking system matter in determining the final search output bias seen by the users. And finally, we use the framework to compare the relative bias for two popular search systems—Twitter social media search and Google web search—for queries related to politicians and political events. We end by discussing some potential solutions to signal the bias in the search results to make the users more aware of them.

Algorithmic decision making is employed in an increasing number of real-world applicationstions to aid human decision making. While it has shown considerable promise in terms of improved decision accuracy, in some scenarios, its outcomes have been also shown to impose an unfair (dis)advantage on people from certain social groups (e.g., women, blacks). In this context, there is a need for computational techniques to limit unfairness in algorithmic decision making. In this work, we take a step forward to fulfill that need and introduce a flexible constraint-based framework to enable the design of fair margin-based classifiers. The main technical innovation of our framework is a general and intuitive measure of decision boundary unfairness, which serves as a tractable proxy to several of the most popular computational definitions of unfairness from the literature. Leveraging our measure, we can reduce the design of fair margin-based classifiers to adding tractable constraints on their decision boundaries. Experiments on multiple synthetic and real-world datasets show that our framework is able to successfully limit unfairness, often at a small cost in terms of accuracy.

Much of political debate focuses on the concern that machines might take over. Yet in many domains it is much more plausible that the ultimate choice and responsibility remain with a human decision-maker, but that she is provided with machine advice. A quintessential illustration is the decision of a judge to bail or jail a defendant. In multiple jurisdictions in the US, judges have access to a machine prediction about a defendant's recidivism risk. In our study, we explore how receiving machine advice influences people's bail decisions. We run a vignette experiment with laypersons whom we test on a subsample of cases from the database of this prediction tool. In study 1, we ask them to predict whether defendants will recidivate before tried, and manipulate whether they have access to machine advice. We find that receiving machine advice has a small effect, which is biased in the direction of predicting no recidivism. In the field, human decision makers sometimes have a chance, after the fact, to learn whether the machine has given good advice. In study 2, after each trial we inform participants of ground truth. This does not make it more likely that they follow the advice, despite the fact that the machine is (on average) slightly more accurate than real judges. This also holds if initially the advice is mostly correct, or if it initially is mostly to predict (no) recidivism. Real judges know that their decisions affect defendants' lives. They may also be concerned about reelection or promotion. Hence a lot is at stake. In study 3 we emulate high stakes by giving participants a financial incentive. An incentive to find the ground truth, or to avoid false positive or false negatives, does not make participants more sensitive to machine advice. But an incentive to follow the advice is effective.

Abstract Most social platforms offer mechanisms allowing users to delete their posts, and a significant fraction of users exercise this right to be forgotten. However, ironically, users’ attempt to reduce attention to sensitive posts via deletion, in practice, attracts unwanted attention from stalkers specifically to those (deleted) posts. Thus, deletions may leave users more vulnerable to attacks on their privacy in general. Users hoping to make their posts forgotten face a “damned if I do, damned if I don’t” dilemma. Many are shifting towards ephemeral social platform like Snapchat, which will deprive us of important user-data archival. In the form of intermittent withdrawals, we present, Lethe, a novel solution to this problem of (really) forgetting the forgotten. If the next-generation social platforms are willing to give up the uninterrupted availability of non-deleted posts by a very small fraction, Lethe provides privacy to the deleted posts over long durations. In presence of Lethe, an adversarial observer becomes unsure if some posts are permanently deleted or just temporarily withdrawn by Lethe; at the same time, the adversarial observer is overwhelmed by a large number of falsely flagged undeleted posts. To demonstrate the feasibility and performance of Lethe, we analyze large-scale real data about users’ deletion over Twitter and thoroughly investigate how to choose time duration distributions for alternating between temporary withdrawals and resurrections of non-deleted posts. We find a favorable trade-off between privacy, availability and adversarial overhead in different settings for users exercising their right to delete. We show that, even against an ultimate adversary with an uninterrupted access to the entire platform, Lethe offers deletion privacy for up to 3 months from the time of deletion, while maintaining content availability as high as 95% and keeping the adversarial precision to 20%.

We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of a distance metric. Instead, we propose novel approaches to elicit and leverage side-information on equally deserving individuals to counter subordination between social groups. We model this knowledge as a fairness graph, and learn a unified Pairwise Fair Representation (PFR) of the data that captures both data-driven similarity between individuals and the pairwise side-information in fairness graph. We elicit fairness judgments from a variety of sources, including human judgments for two real-world datasets on recidivism prediction (COMPAS) and violent neighborhood prediction (Crime & Communities). Our experiments show that the PFR model for operationalizing individual fairness is practically viable.

In recent years, we have witnessed a paradigm shift in news consumption. In traditional news media organizations, a small number of expert editors are responsible for selecting news stories that are consumed by all news readers (the audience). However, with the growing popularity of social media as a news consumption medium, a part of the editorial power of selecting news stories has shifted to the audience who select and share the stories that can reach a large number of consumers. In this paper, we analyze data from two popular news media sites-The New York Times and The Guardian, and characterize the considerable differences in the types of stories selected by the audience and expert news editors. We also find that story selections by audience vary significantly across different social media channels such as Twitter, Facebook, and email. We contextualize the differences utilizing media and communication theory and discuss their implications for news readers and media organizations.

With widespread use of machine learning methods in numerous domains involving humans, several studies have raised questions about the potential for unfairness towards certain individuals or groups. A number of recent works have proposed methods to measure and eliminate unfairness from machine learning models. However, most of this work has focused on only one dimension of fair decision making: distributive fairness, i.e., the fairness of the decision outcomes. In this work, we leverage the rich literature on organizational justice and focus on another dimension of fair decision making: procedural fairness, i.e., the fairness of the decision making process. We propose measures for procedural fairness that consider the input features used in the decision process, and evaluate the moral judgments of humans regarding the use of these features. We operationalize these measures on two real world datasets using human surveys on the Amazon Mechanical Turk (AMT) platform, demonstrating that our measures capture important properties of procedurally fair decision making. We provide fast submodular mechanisms to optimize the tradeoff between procedural fairness and prediction accuracy. On our datasets, we observe empirically that procedural fairness may be achieved with little cost to outcome fairness, but that some loss of accuracy is unavoidable.

Although diverse news stories are actively posted on social media, readers often focus on the news which reinforces their pre-existing views, leading to 'filter bubble' effects. To combat this, some recent systems expose and nudge readers toward stories with different points of view. One example is the Wall Street Journal's 'Blue Feed, Red Feed' system, which presents posts from biased publishers on each side of a topic. However, these systems have had limited success. We present a complementary approach which identifies high consensus 'purple' posts that generate similar reactions from both 'blue' and 'red' readers. We define and operationalize consensus for news posts on Twitter in the context of US politics. We show that high consensus posts can be identified and discuss their empirical properties. We present a method for automatically identifying high and low consensus news posts on Twitter, which can work at scale across many publishers. To do this, we propose a novel category of audience leaning based features, which we show are well suited to this task. Finally, we present our 'Purple Feed' system which highlights high consensus posts from publishers on both sides of the political spectrum.

Security behaviors can help users avoid incidents, but can also increase costs, both to users -- in time and mental effort -- and to platforms -- in user engagement and engineering resources. As such, we should consider when it is most efficient and effective to encourage security behaviors. Recent work has shown that users attempt to make security decisions based on cost benefit tradeoffs (boundedly, rationally). Yet, sometimes security nudges (e.g., create unique passwords for every website) encourage users toward irrational behavior: creating strong, unique passwords even for those sites that contain no personal data. In this work-in-progress, we present a mechanism design (a framework) that can be used to optimize the distribution of security nudges and requirements among users with different levels of risk or different levels of investment in a given system. Further, we introduce a new paradigm: the distribution of resources (e.g., ubikeys) that can lower the cost of security behaviors to those users with the most need (the highest time cost from 2FA or lowest Internet skill). Future work will involve simulations showing the value of optimizing distribution of nudges and resources using this framework, and evaluating such an approach in a live test.

Online advertising platforms such as those of Facebook and Google collect detailed data about users, which they leverage to allow advertisers to target ads to users based on various pieces of user information. While most advertising platforms have transparency mechanisms in place to reveal this collected information to users, these often present an incomplete view of the information being collected and of how it is used for targeting ads, thus necessitating further transparency.

Recent work has explored how to train machine learning models which do not discriminate against any subgroup of the population as determined by sensitive attributes such as gender or race. To avoid disparate treatment, sensitive attributes should not be considered. On the other hand, in order to avoid disparate impact, sensitive attributes must be examined, e.g., in order to learn a fair model, or to check if a given model is fair. We introduce methods from secure multi-party computation which allow us to avoid both. By encrypting sensitive attributes, we show how an outcome-based fair model may be learned, checked, or have its outputs verified and held to account, without users revealing their sensitive attributes.

Discrimination via algorithmic decision making has received considerable attention. Prior work largely focuses on defining conditions for fairness, but does not define satisfactory measures of algorithmic unfairness. In this paper, we focus on the following question: Given two unfair algorithms, how should we determine which of the two is more unfair? Our core idea is to use existing inequality indices from economics to measure how unequally the outcomes of an algorithm benefit different individuals or groups in a population. Our work offers a justified and general framework to compare and contrast the (un)fairness of algorithmic predictors. This unifying approach enables us to quantify unfairness both at the individual and the group level. Further, our work reveals overlooked tradeoffs between different fairness notions: using our proposed measures, the overall individual-level unfairness of an algorithm can be decomposed into a between-group and a within-group component. Earlier methods are typically designed to tackle only between-group un- fairness, which may be justified for legal or other reasons. However, we demonstrate that minimizing exclusively the between-group component may, in fact, increase the within-group, and hence the overall unfairness. We characterize and illustrate the tradeoffs between our measures of (un)fairness and the prediction accuracy.

We draw attention to an important, yet largely overlooked aspect of evaluating fairness for automated decision making systems---namely risk and welfare considerations. Our proposed family of measures corresponds to the long-established formulations of cardinal social welfare in economics, and is justified by the Rawlsian conception of fairness behind a veil of ignorance. The convex formulation of our welfare-based measures of fairness allows us to integrate them as a constraint into any convex loss minimization pipeline. Our empirical analysis reveals interesting trade-offs between our proposal and (a) prediction accuracy, (b) group discrimination, and (c) Dwork et al's notion of individual fairness. Furthermore and perhaps most importantly, our work provides both heuristic justification and empirical evidence suggesting that a lower-bound on our measures often leads to bounded inequality in algorithmic outcomes; hence presenting the first computationally feasible mechanism for bounding individual-level inequality.

Rankings of people and items are at the heart of selection-making, match-making, and recommender systems, ranging from employment sites to sharing economy platforms. As ranking positions influence the amount of attention the ranked subjects receive, biases in rankings can lead to unfair distribution of opportunities and resources such as jobs or income. This paper proposes new measures and mechanisms to quantify and mitigate unfairness from a bias inherent to all rankings, namely, the position bias which leads to disproportionately less attention being paid to low-ranked subjects. Our approach differs from recent fair ranking approaches in two important ways. First, existing works measure unfairness at the level of subject groups while our measures capture unfairness at the level of individual subjects, and as such subsume group unfairness. Second, as no single ranking can achieve individual attention fairness, we propose a novel mechanism that achieves amortized fairness, where attention accumulated across a series of rankings is proportional to accumulated relevance. We formulate the challenge of achieving amortized individual fairness subject to constraints on ranking quality as an online optimization problem and show that it can be solved as an integer linear program. Our experimental evaluation reveals that unfair attention distribution in rankings can be substantial, and demonstrates that our method can improve individual fairness while retaining high ranking quality.

Sites like Facebook and Google now serve as de facto data brokers, aggregating data on users for the purpose of implementing powerful advertising platforms. Historically, these services allowed advertisers to select which users see their ads via targeting attributes. Recently, most advertising platforms have begun allowing advertisers to target users directly by uploading the personal information of the users who they wish to advertise to (e.g., their names, email addresses, phone numbers, etc.); these services are often known as custom audiences. Custom audiences effectively represent powerful linking mechanisms, allowing advertisers to leverage any PII (e.g., from customer data, public records, etc.) to target users. In this paper, we focus on Facebook's custom audience implementation and demonstrate attacks that allow an adversary to exploit the interface to infer users' PII as well as to infer their activity. Specifically, we show how the adversary can infer users' full phone numbers knowing just their email address, determine whether a particular user visited a website, and de-anonymize all the visitors to a website by inferring their phone numbers en masse. These attacks can be conducted without any interaction with the victim(s), cannot be detected by the victim(s), and do not require the adversary to spend money or actually place an ad. We propose a simple and effective fix to the attacks based on reworking the way Facebook de-duplicates uploaded information. Facebook's security team acknowledged the vulnerability and has put into place a fix that is a variant of the fix we propose. Overall, our results indicate that advertising platforms need to carefully consider the privacy implications of their interfaces.

As algorithms are increasingly used to make important decisions that affect human lives, ranging from social benefit assignment to predicting risk of criminal recidivism, concerns have been raised about the fairness of algorithmic decision making. Most prior works on algorithmic fairness normatively prescribe how fair decisions ought to be made. In contrast, here, we descriptively survey users for how they perceive and reason about fairness in algorithmic decision making. A key contribution of this work is the framework we propose to understand why people perceive certain features as fair or unfair to be used in algorithms. Our framework identifies eight properties of features, such as relevance, volitionality and reliability, as latent considerations that inform people»s moral judgments about the fairness of feature use in decision-making algorithms. We validate our framework through a series of scenario-based surveys with 576 people. We find that, based on a person»s assessment of the eight latent properties of a feature in our exemplar scenario, we can accurately (> 85%) predict if the person will judge the use of the feature as fair. Our findings have important implications. At a high-level, we show that people»s unfairness concerns are multi-dimensional and argue that future studies need to address unfairness concerns beyond discrimination. At a low-level, we find considerable disagreements in people»s fairness judgments. We identify root causes of the disagreements, and note possible pathways to resolve them.

Algorithmic decision making systems are ubiquitous across a wide variety of online as well as offline services. These systems rely on complex learning methods and vast amounts of data to optimize the service functionality, satisfaction of the end user and profitability. However, there is a growing concern that these automated decisions can lead, even in the absence of intent, to a lack of fairness, i.e., their outcomes can disproportionately hurt (or, benefit) particular groups of people sharing one or more sensitive attributes (e.g., race, sex). In this paper, we introduce a flexible mechanism to design fair classifiers by leveraging a novel intuitive measure of decision boundary (un)fairness. We instantiate this mechanism with two well-known classifiers, logistic regression and support vector machines, and show on real-world data that our mechanism allows for a fine-grained control on the degree of fairness, often at a small cost in terms of accuracy.

Search engines in online communities such as Twitter or Facebook not only return matching posts, but also provide links to the profiles of the authors. Thus, when a user appears in the top-k results for a sensitive keyword query, she becomes widely exposed in a sensitive context. The effects of such exposure can result in a serious privacy violation, ranging from embarrassment all the way to becoming a victim of organizational discrimination.

Search systems in online social media sites are frequently used to find information about ongoing events and people. For topics with multiple competing perspectives, such as political events or political candidates, bias in the top ranked results significantly shapes public opinion. However, bias does not emerge from an algorithm alone. It is important to distinguish between the bias that arises from the data that serves as the input to the ranking system and the bias that arises from the ranking system itself. In this paper, we propose a framework to quantify these distinct biases and apply this framework to politics-related queries on Twitter. We found that both the input data and the ranking system contribute significantly to produce varying amounts of bias in the search results and in different ways. We discuss the consequences of these biases and possible mechanisms to signal this bias in social media search systems' interfaces.

In this paper we analyze Facebook's Free Basics program, which provides free Internet access to a restricted set of web services. As the program grows to 60+ developing countries, an independent and data-driven audit of its scope and outreach is highly relevant to the ICTD community.

The adoption of automated, data-driven decision making in an ever expanding range of applications has raised concerns about its potential unfairness towards certain social groups. In this context, a number of recent studies have focused on defining, detecting, and removing unfairness from data-driven decision systems. However, the existing notions of fairness, based on parity (equality) in treatment or outcomes for different social groups, tend to be quite stringent, limiting the overall decision making accuracy. In this paper, we draw inspiration from the fair-division and envy-freeness literature in economics and game theory and propose preference-based notions of fairness -- given the choice between various sets of decision treatments or outcomes, any group of users would collectively prefer its treatment or outcomes, regardless of the (dis)parity as compared to the other groups. Then, we introduce tractable proxies to design margin-based classifiers that satisfy these preference-based notions of fairness. Finally, we experiment with a variety of synthetic and real-world datasets and show that preference-based fairness allows for greater decision accuracy than parity-based fairness.

Online news media sites are emerging as the primary source of news for a large number of users. The selection of 'front-page' stories on these media sites usually takes into consideration several crowdsourced popularity metrics, such as number of views or shares by the readers. In this work, we focus on automatically recommending front-page stories in such media websites. When recommending news stories, there are two basic metrics of interest - recency and relevancy. Ideally, recommender systems should recommend the most relevant stories soon after they are published. However, the relevancy of a story only becomes evident as the story ages, thereby creating a tension between recency and relevancy. A systematic analysis of popular recommendation strategies in use today reveals that they lead to poor trade-offs between recency and relevancy in practice. So, in this paper, we propose a new recommendation strategy (called Highest Future-Impact) which attempts to optimize on both the axes. To implement our proposed strategy in practice, we develop an optimization framework combining the predicted future-impact of the stories with the uncertainties in the predictions. Evaluations over three real-world news datasets show that our implementation achieves good performance trade-offs between recency and relevancy.

Automated data-driven decision making systems are increasingly being used to assist, or even replace humans in many settings. These systems function by learning from historical decisions, often taken by humans. In order to maximize the utility of these systems (or, classifiers), their training involves minimizing the errors (or, misclassifications) over the given historical data. However, it is quite possible that the optimally trained classifier makes decisions for people belonging to different social groups with different misclassification rates (e.g., misclassification rates for females are higher than for males), thereby placing these groups at an unfair disadvantage. To account for and avoid such unfairness, in this paper, we introduce a new notion of unfairness, disparate mistreatment, which is defined in terms of misclassification rates. We then propose intuitive measures of disparate mistreatment for decision boundary-based classifiers, which can be easily incorporated into their formulation as convex-concave constraints. Experiments on synthetic as well as real world datasets show that our methodology is effective at avoiding disparate mistreatment, often at a small cost in terms of accuracy.

Extracting news on specific topics from the Twitter microblogging site poses formidable challenges, which include handling millions of tweets posted daily, judging topicality and importance of tweets, and ensuring trustworthiness of results in the face of spam. To date, all scalable approaches have relied on crowd wisdom, i.e., keyword-matching on the global tweet stream to gather relevant tweets, and crowd- endorsements to judge the importance of tweets. We propose a fundamentally different methodology -- for a given topic, we identify trustworthy experts on the topic, and extract news-stories that are most popular among the experts. Comparing the crowd-based and expert-based methodologies, we demonstrate that the news-stories obtained by our methodology (i) have higher relevance for a wide variety of topics, (ii) achieve very high coverage of important news-stories posted globally in Twitter, and (iii) are far more trustworthy. Using our methodology, we implemented and publicly deployed a topical news system for Twitter, which can extract news-stories on thousands of topics.

Privacy of Internet users is at stake because they expose personal information in posts created in online communities, in search queries, and other activities. An adversary that monitors a community may identify the users with the most sensitive properties and utilize this knowledge against them (e.g., by adjusting the pricing of goods or targeting ads of sensitive nature). Existing privacy models for structured data are inadequate to capture privacy risks from user posts.

Social media sites are information marketplaces, where users produce and consume a wide variety of information and ideas. In these sites, users typically choose their information sources, which in turn determine what specific information they receive, how much information they receive and how quickly this information is shown to them. In this context, a natural question that arises is how efficient are social media users at selecting their information sources. In this work, we propose a computational framework to quantify users' efficiency at selecting information sources. Our framework is based on the assumption that the goal of users is to acquire a set of unique pieces of information. To quantify user's efficiency, we ask if the user could have acquired the same pieces of information from another set of sources more efficiently. We define three different notions of efficiency -- link, in-flow, and delay -- corresponding to the number of sources the user follows, the amount of (redundant) information she acquires and the delay with which she receives the information. Our definitions of efficiency are general and applicable to any social media system with an underlying in- formation network, in which every user follows others to receive the information they produce.

On most current websites untrustworthy or spammy identities are easily created. Existing proposals to detect untrustworthy identities rely on reputation signals obtained by observing the activities of identities over time within a single site or domain; thus, there is a time lag before which websites cannot easily distinguish attackers and legitimate users. In this paper, we investigate the feasibility of leveraging information about identities that is aggregated across multiple domains to reason about their trustworthiness. Our key insight is that while honest users naturally maintain identities across multiple domains (where they have proven their trustworthiness and have acquired reputation over time), attackers are discouraged by the additional effort and costs to do the same. We propose a flexible framework to transfer trust between domains that can be implemented in today's systems without significant loss of privacy or significant implementation overheads.

People have long been aware of malicious users that impersonate celebrities or launch identity theft attacks in social networks. However, beyond anecdotal evidence, there have been no in-depth studies of impersonation attacks in today's social networks. One reason for the lack of studies in this space is the absence of datasets about impersonation attacks. To this end, we propose a technique to build extensive datasets of impersonation attacks in current social networks and we gather 16,572 cases of impersonation attacks in the Twitter social network. Our analysis reveals that most identity impersonation attacks are not targeting celebrities or identity theft. Instead, we uncover a new class of impersonation attacks that clone the profiles of ordinary people on Twitter to create real-looking fake identities and use them in malicious activities such as follower fraud. We refer to these as the doppelgänger bot attacks. Our findings show (i) that identity impersonation attacks are much broader than believed and can impact any user, not just celebrities and (ii) that attackers are evolving and create real-looking accounts that are harder to detect by current systems. We also propose and evaluate methods to automatically detect impersonation attacks sooner than they are being detected in today's Twitter social network.

Matching the profiles of a user across multiple online social networks brings opportunities for new services and applications as well as new insights on user online behavior, yet it raises serious privacy concerns. Prior literature has showed that it is possible to accurately match profiles, but their evaluation focused only on sampled datasets. In this paper, we study the extent to which we can reliably match profiles in practice, across real-world social networks, by exploiting public attributes, i.e., information users publicly provide about themselves. Today's social networks have hundreds of millions of users, which brings completely new challenges as a reliable matching scheme must identify the correct matching profile out of the millions of possible profiles. We first define a set of properties for profile attributes--Availability, Consistency, non-Impersonability, and Discriminability (ACID)--that are both necessary and sufficient to determine the reliability of a matching scheme. Using these properties, we propose a method to evaluate the accuracy of matching schemes in real practical cases. Our results show that the accuracy in practice is significantly lower than the one reported in prior literature. When considering entire social networks, there is a non-negligible number of profiles that belong to different users but have similar attributes, which leads to many false matches. Our paper sheds light on the limits of matching profiles in the real world and illustrates the correct methodology to evaluate matching schemes in realistic scenarios.

Analysis of content streams gathered from social networking sites such as Twitter has several applications ranging from content search and recommendation, news detection to business analytics. However, processing large amounts of data generated on these sites in real-time poses a difficult challenge. To cope with the data deluge, analytics companies and researchers are increasingly resorting to sampling. In this article, we investigate the crucial question of how to sample content streams generated by users in online social networks . The traditional method is to randomly sample all the data. For example, most studies using Twitter data today rely on the 1% and 10% randomly sampled streams of tweets that are provided by Twitter. In this paper, we analyze a different sampling methodology, one where content is gathered only from a relatively small sample (<1%) of the user population, namely, the expert users . Over the duration of a month, we gathered tweets from over 500,000 Twitter users who are identified as experts on a diverse set of topics, and compared the resulting expert sampled tweets with the 1% randomly sampled tweets provided publicly by Twitter. We compared the sampled datasets along several dimensions, including the popularity, topical diversity, trustworthiness, and timeliness of the information contained within them, and on the sentiment/opinion expressed on specific topics. Our analysis reveals several important differences in data obtained through the different sampling methodologies, which have serious implications for applications such as topical search, trustworthy content recommendations, breaking news detection, and opinion mining.

We here investigate what drives the popularity of information on social media platforms. Focusing on YouTube, we seek to understand the extent to which content by itself determines a video's popularity. Using mechanical turk as experimental platform, we asked users to evaluate pairs of videos, and compared users' relative perception of the videos' content against their relative popularity reported by YouTube. We found that in most evaluations users could not reach consensus on which video had better content as their perceptions tend to be very subjective. Nevertheless, when consensus was reached, the video with preferred content almost always achieved greater popularity on YouTube, highlighting the importance of content in driving information popularity on social media.

We present a semantic methodology to identify topical groups in Twitter on a large number of topics, each consisting of users who are experts on or interested in a specific topic. Early studies investigating the nature of Twitter suggest that it is a social media platform consisting of a relatively small section of elite users, producing information on a few popular topics such as media, politics, and music, and the general population consuming it. We show that this characterization ignores a rich set of highly specialized topics, ranging from geology, neurology, to astrophysics and karate - each being discussed by their own topical groups. We present a detailed characterization of these topical groups based on their network structures and tweeting behaviors. Analyzing these groups on the backdrop of the common identity and bond theory in social sciences shows that these groups exhibit characteristics of topical-identity based groups, rather than social-bond based ones.

We propose a novel mechanism to infer topics of interest of individual users in the Twitter social network. We observe that in Twitter, a user generally follows experts on various topics of her interest in order to acquire information on those topics. We use a methodology based on social annotations (proposed earlier by us) to first deduce the topical expertise of popular Twitter users, and then transitively infer the interests of the users who follow them. This methodology is a sharp departure from the traditional techniques of inferring interests of a user from the tweets that she posts or receives. We show that the topics of interest inferred by the proposed methodology are far superior than the topics extracted by state-of-the-art techniques such as using topic models (Labeled LDA) on tweets. Based upon the proposed methodology, we build a system Who Likes What, which can infer the interests of millions of Twitter users. To our knowledge, this is the first system that can infer interests for Twitter users at such scale. Hence, this system would be particularly beneficial in developing personalized recommender services over the Twitter platform.

Several applications today rely upon content streams crowd-sourced from online social networks. Since real-time processing of large amounts of data generated on these sites is difficult, analytics companies and researchers are increasingly resorting to sampling. In this paper, we investigate the crucial question of how to sample the data generated by users in social networks. The traditional method is to randomly sample all the data. We analyze a different sampling methodology, where content is gathered only from a relatively small subset (< 1%) of the user population namely, the expert users. Over the duration of a month, we gathered tweets from over 500,000 Twitter users who are identified as experts on a diverse set of topics, and compared the resulting expert-sampled tweets with the 1% randomly sampled tweets provided publicly by Twitter. We compared the sampled datasets along several dimensions, including the diversity, timeliness, and trustworthiness of the information contained within them, and find important differences between the datasets. Our observations have major implications for applications such as topical search, trustworthy content recommendations, and breaking news detection.

The way in which social conventions emerge in communities has been of interest to social scientists for decades. Here we report on the emergence of a particular social convention on Twitter---the way to indicate a tweet is being reposted and attributing the content to its source. Despite being invented at different times and having different adoption rates, only two variations became widely adopted. In this paper we describe this process in detail, highlighting the factors that come into play in deciding which variation individuals will adopt. Our classification analysis demonstrates that the date of adoption and the number of exposures are particularly important in the adoption process, while personal features (such as the number of followers and join date) and the number of adopter friends have less discriminative power in predicting adoptions. We discuss implications of these findings in the design of future Web applications and services.

Thwarting large-scale crawls of user profiles in online social networks (OSNs) like Facebook and Renren is in the interest of both the users and the operators of these sites. OSN users wish to maintain control over their personal information, and OSN operators wish to protect their business assets and reputation. Existing rate-limiting techniques are ineffective against crawlers with many accounts, be they fake accounts (also known as Sybils) or compromised accounts of real users obtained on the black market.

There has been a flurry of research on leveraging social networks to defend against multiple identity, or Sybil, attacks. A series of recent works does not try to explicitly identify Sybil identities and, instead, bounds the impact that Sybil identities can have. We call these approaches Sybil tolerance; they have shown to be effective in applications including reputation systems, spam protection, online auctions, and content rating systems. All of these approaches use a social network as a credit network, rendering multiple identities ineffective to an attacker without a commensurate increase in social links to honest users (which are assumed to be hard to obtain). Unfortunately, a hurdle to practical adoption is that Sybil tolerance relies on computationally expensive network analysis, thereby limiting widespread deployment.

Finding topic experts on microblogging sites with millions of users, such as Twitter, is a hard and challenging problem. In this paper, we propose and investigate a new methodology for discovering topic experts in the popular Twitter social network. Our methodology relies on the wisdom of the Twitter crowds -- it leverages Twitter Lists, which are often carefully curated by individual users to include experts on topics that interest them and whose meta-data (List names and descriptions) provides valuable semantic cues to the experts' domain of expertise. We mined List information to build Cognos, a system for finding topic experts in Twitter. Detailed experimental evaluation based on a real-world deployment shows that: (a) Cognos infers a user's expertise more accurately and comprehensively than state-of-the-art systems that rely on the user's bio or tweet content, (b) Cognos scales well due to built-in mechanisms to efficiently update its experts' database with new users, and (c) Despite relying only on a single feature, namely crowdsourced Lists, Cognos yields results comparable to, if not better than, those given by the official Twitter experts search engine for a wide range of queries in user tests. Our study highlights Lists as a potentially valuable source of information for future content or expert search systems in Twitter.

Recently, Twitter has emerged as a popular platform for discovering real-time information on the Web, such as news stories and people's reaction to them. Like the Web, Twitter has become a target for link farming, where users, especially spammers, try to acquire large numbers of follower links in the social network. Acquiring followers not only increases the size of a user's direct audience, but also contributes to the perceived influence of the user, which in turn impacts the ranking of the user's tweets by search engines.

Information propagation in online social networks like Twitter is unique in that word-of-mouth propagation and traditional media sources coexist. We collect a large amount of data from Twitter to compare the relative roles different types of users play in information flow. Using empirical data on the spread of news about major international headlines as well as minor topics, we investigate the relative roles of three types of information spreaders: 1) mass media sources like BBC; 2) grassroots, consisting of ordinary users; and 3) evangelists, consisting of opinion leaders, politicians, celebrities, and local businesses. Mass media sources play a vital role in reaching the majority of the audience in any major topics. Evangelists, however, introduce both major and minor topics to audiences who are further away from the core of the network and would otherwise be unreachable. Grassroots users are relatively passive in helping spread the news, although they account for the 98% of the network. Our results bring insights into what contributes to rapid information propagation at different levels of topic popularity, which we believe are useful to the designers of social search and recommendation engines.

The sharing of personal data has emerged as a popular activity over online social networking sites like Facebook. As a result, the issue of online social network privacy has received significant attention in both the research literature and the mainstream media. Our overarching goal is to improve defaults and provide better tools for managing privacy, but we are limited by the fact that the full extent of the privacy problem remains unknown; there is little quantification of the incidence of incorrect privacy settings or the difficulty users face when managing their privacy.

Traditionally, users have discovered information on the Web by browsing or searching. Recently, word-of-mouth has emerged as a popular way of discovering the Web, particularly on social networking sites like Facebook and Twitter. On these sites, users discover Web content by following URLs posted by their friends. Such word-of-mouth based content discovery has become a major driver of traffic to many Web sites today. To better understand this popular phenomenon, in this paper we present a detailed analysis of word-of-mouth exchange of URLs among Twitter users. Among our key findings, we show that Twitter yields propagation trees that are wider than they are deep. Our analysis on the geolocation of users indicates that users who are geographically close together are more likely to share the same URL.

Today, OSN sites allow users to share data using a centrally controlled web infrastructure. However, if users shared data directly from home, they could potentially retain full control over the data (i.e., what to share, whom to share with). This paper investigates the feasibility of alternative decentralized architectures that allow users to share their data directly from home. Specifically, we (a) characterize social content workloads using data gathered from the popular Flickr and YouTube social networks and (b) characterize home networks using data gathered from residential gateways deployed in a number of households. We use the data from these measurements to evaluate the potential for delivering social content directly from users' homes.

Online social networking sites (OSNs) like Facebook and Orkut contain personal data of millions of users. Many OSNs view this data as a valuable asset that is at the core of their business model. Both OSN users and OSNs have strong incentives to restrict large scale crawls of this data. OSN users want to protect their privacy and OSNs their business interest. Traditional defenses against crawlers involve rate- limiting browsing activity per user account. These defense schemes, however, are vulnerable to Sybil attacks, where a crawler creates a large number of fake user accounts. In this paper, we propose Genie, a system that can be deployed by OSN operators to defend against Sybil crawlers. Genie is based on a simple yet powerful insight: the social network itself can be leveraged to defend against Sybil crawlers. We first present Genie's design and then discuss how Genie can limit crawlers while allowing browsing of user profiles by normal users.

The Internet is witnessing explosive growth in traffic due to bulk content transfers, such as multimedia and software downloads, and online sharing of personal, commercial, and scientific data. Yet bulk data transfers remain very expensive and inefficient. As a result, huge amounts of digital data continue to be delivered outside of the Internet using hard drives, optical media or tapes. Meanwhile, large reserves of spare bandwidth lie unutilized in today's networks, where links are overprovisioned for peak load. We designed NetEx, a bulk transfer system that opportunistically exploits the excess capacities of network links to deliver bulk content cheaply and efficiently. Our results based on data from both a commercial tier-1 ISP and the Abilene network suggest that NetEx can considerably increase the capacity of the network, and at the same time it can provide good average performance to bulk transfers.

Recently, there has been much excitement in the research community over using social networks to mitigate multiple identity, or Sybil, attacks. A number of schemes have been proposed, but they differ greatly in the algorithms they use and in the networks upon which they are evaluated. As a result, the research community lacks a clear understanding of how these schemes compare against each other, how well they would work on real-world social networks with different structural properties, or whether there exist other (potentially better) ways of Sybil defense.

Online social networks are now a popular way for users to connect, express themselves, and share content. Users in today's online social networks often post a profile, consisting of attributes like geographic location, interests, and schools attended. Such profile information is used on the sites as a basis for grouping users, for sharing content, and for suggesting users who may benefit from interaction. However, in practice, not all users provide these attributes.

Online social networking sites like MySpace, Facebook, and Flickr have become a popular way to share and disseminate content. Their massive popularity has led to viral marketing techniques that attempt to spread content, products, and ideas on these sites. However, there is little data publicly available on viral propagation in the real world and few studies have characterized how information spreads over current online social networks.

Recently, it has been reported that certain access ISPs are surreptitiously blocking their customers from uploading data using the popular BitTorrent file-sharing protocol. The reports have sparked an intense and wide-ranging policy debate on network neutrality and ISP traffic management practices. However, to date, end users lack access to measurement tools that can detect whether their access ISPs are blocking their BitTorrent traffic. And since ISPs do not voluntarily disclose their traffic management policies, no one knows how widely BitTorrent traffic blocking is deployed in the current Internet. In this paper, we address this problem by designing an easy-to-use tool to detect BitTorrent blocking and by presenting results from a widely used public deployment of the tool.

Planetary-scale network testbeds like PlanetLab and RON have become indispensable for evaluating prototypes of distributed systems under realistic Internet conditions. However, current testbeds lack the heterogeneity that characterizes the commercial Internet. For example, most testbed nodes are connected to well-provisioned research networks, whereas most Internet nodes are in edge networks.

A large and rapidly growing proportion of users connect to the Internet via residential broadband networks such as Digital Subscriber Lines (DSL) and cable. Residential networks are often the bottleneck in the last mile of today's Internet. Their characteristics critically affect Internet applications, including voice-over-IP, online games, and peer-to-peer content sharing/delivery systems. However, to date, few studies have investigated commercial broadband deployments, and rigorous measurement data that characterize these networks at scale are lacking. In this paper, we present the first large-scale measurement study of major cable and DSL providers in North America and Europe. We describe and evaluate the measurement tools we developed for this purpose. Our study characterizes several properties of broadband networks, including link capacities, packet round-trip times and jitter, packet loss rates, queue lengths, and queue drop policies. Our analysis reveals important ways in which residential networks differ from how the Internet is conventionally thought to operate. We also discuss the implications of our findings for many emerging protocols and systems, including delay-based congestion control (e.g., PCP) and network coordinate systems (e.g., Vivaldi).

Online social networking sites like Orkut, YouTube, and Flickr are among the most popular sites on the Internet. Users of these sites form a social network, which provides a powerful means of sharing, organizing, and finding content and contacts. The popularity of these sites provides an opportunity to study the characteristics of online social network graphs at large scale. Understanding these graphs is important, both to improve current systems and to design new applications of online social networks.

This paper proposes Monarch, a novel tool that accurately emulates transport protocol flows from an end host controlled by its user to any other Internet host that responds to simple TCP, UDP, or ICMP packet probes. Since many Internet hosts and routers respond to such probes, Monarch can evaluate transport protocols, such as TCP Reno, TCP Vegas, and TCP Nice, over a large and diverse set of Internet paths. Current approaches to evaluating these protocols need control over both end hosts of an Internet path. Consequently, they are limited to a small number of paths between nodes in testbeds like PlanetLab, RON or NIMI. Monarch's ability to evaluate transport protocols with minimal support from the destination host enables many new measurement studies. We show the feasibility of using Monarch for three example studies: (a) understanding transport protocol behavior over network paths that are less explored by the research community, such as paths to cable and DSL hosts, (b) investigating the relative performance of different transport protocol designs, such as TCP Vegas and TCP Reno, and (c) testing protocol implementations under a wide range of experimental conditions.

Distributed hash tables have been proposed as flat, nonhierarchical structures, in contrast to most scalable distributed systems of the past. We show how to construct hierarchical DHTs while retaining the homogeneity of load and functionality offered by flat designs. Our generic construction, Canon, offers the same routing state vs. routing hops trade-off provided by standard DHT designs. The advantages of Canon include (but are not limited to) (a) fault isolation, (b) efficient caching and effective bandwidth usage for multicast, (c) adaptation to the underlying physical network, (d) hierarchical storage of content, and (e) hierarchical access control. Canon can be applied to many different proposed DHTs to construct their Canonical versions. We show how four different DHTs - Chord, Symphony, CAN and Kademlia - can be converted into their Canonical versions that we call Crescendo, Cacophony, Can-Can and Kandy respectively.

The various proposed DHT routing algorithms embody several different underlying routing geometries. These geometries include hypercubes, rings, tree-like structures, and butterfly networks. In this paper we focus on how these basic geometric approaches affect the resilience and proximity properties of DHTs. One factor that distinguishes these geometries is the degree of flexibility they provide in the selection of neighbors and routes. Flexibility is an important factor in achieving good static resilience and effective proximity neighbor and route selection. Our basic finding is that, despite our initial preference for more complex geometries, the ring geometry allows the greatest flexibility, and hence achieves the best resilience and proximity performance.

Peer-to-peer (P2P) file sharing accounts for an astonishing volume of current Internet traffic. This paper probes deeply into modern P2P file sharing systems and the forces that drive them. By doing so, we seek to increase our understanding of P2P file sharing workloads and their implications for future multimedia workloads. Our research uses a three-tiered approach. First, we analyze a 200-day trace of over 20 terabytes of Kazaa P2P traffic collected at the University of Washington. Second, we develop a model of multimedia workloads that lets us isolate, vary, and explore the impact of key system parameters. Our model, which we parameterize with statistics from our trace, lets us confirm various hypotheses about file-sharing behavior observed in the trace. Third, we explore the potential impact of locality-awareness in Kazaa.Our results reveal dramatic differences between P2P file sharing and Web traffic. For example, we show how the immutability of Kazaa's multimedia objects leads clients to fetch objects at most once; in contrast, a World-Wide Web client may fetch a popular page (e.g., CNN or Google) thousands of times. Moreover, we demonstrate that: (1) this "fetch-at-most-once" behavior causes the Kazaa popularity distribution to deviate substantially from Zipf curves we see for the Web, and (2) this deviation has significant implications for the performance of multimedia file-sharing systems. Unlike the Web, whose workload is driven by document change, we demonstrate that clients' fetch-at-most-once behavior, the creation of new objects, and the addition of new clients to the system are the primary forces that drive multimedia workloads such as Kazaa. We also show that there is substantial untapped locality in the Kazaa workload. Finally, we quantify the potential bandwidth savings that locality-aware P2P file-sharing architectures would achieve.

Several distributed real-time applications (e.g., medical imaging, air traffic control, video conferencing) demand hard guarantees on the message delivery latency and the recovery delay from component failures. As these demands cannot be met in traditional datagram services, special schemes have been proposed to provide timely recovery for real-time communications in multihop networks. These schemes reserve additional network resources (spare resources) a priori along a backup channel that is disjoint with the primary. Upon a failure in the primary channel, its backup is activated, making the real-time connection dependable. We propose a new backup method, called segmented backups, in which backup paths are provided for partial segments of the primary path rather than for its entire length, as is done in the existing schemes. We show that our method offers: 1) improved network resource utilization; 2) higher average call acceptance rate; 3) better quality-of-service guarantees on propagation delays and failure-recovery times; 4) increased flexibility to control the level of fault tolerance of each connection separately. We provide an algorithm for routing the segmented backups and prove its optimality with respect to spare resource reservation. We detail necessary extensions to resource reservation protocol (RSVP) to support our scheme and argue that they increase the implementation complexity of RSVP minimally. Our simulation studies on various network topologies demonstrate that spare resource aggregation methods, such as backup multiplexing, are more effective when applied to our scheme than to earlier schemes.

The ability to estimate network latencies between arbitrary Internet end hosts would enable new measurement studies and applications, such as investigating routing path inefficiencies on a wide-scale or constructing topologically sensitive overlay networks. In this paper we present King, a tool that accurately and quickly estimates the latency between arbitrary end hosts by using recursive DNS queries in a novel way. Compared to previous approaches, King has several advantages. Unlike IDMaps, King does not require the deployment of additional infrastructure, and unlike GNP, King does not require end hosts to agree upon a set of reference points. Unlike both IDMaps and GNP, King's estimates are based on direct online measurements rather than offline extrapolation. Because King uses existing DNS infrastructure, King scales naturally both in terms of the number of hosts that can be measured and in terms of the number of hosts performing measurements.After describing the techniques used in King, we present an extensive evaluation and analysis of the accuracy and consistency of our tool. Specifically, our evaluation shows that the accuracy of King is significantly better than the accuracy of IDMaps, and that King tends to preserve order among its latency estimates. Finally, we describe a variety of measurement studies and applications that could benefit from our tool, and present results from one such measurement study.