Large language models (LLMs) increasingly support heterogeneous tasks within a single interface, requiring users to form, update, and act upon beliefs about one system across domains with different reliability profiles. Understanding how such beliefs transfer across tasks and shape delegation is therefore critical for the design of multipurpose AI systems. We report a preregistered experiment (N = 240, 7,200 trials) in which participants interacted with a controlled AI simulation across grammar checking, travel planning, and visual question answering, each with fixed, domain-typical accuracy levels. Delegation was operationalized as a binary reliance decision—accepting the AI’s output versus acting independently and belief dynamics were evaluated against Bayesian benchmarks. We find three main results. First, participants do not reset beliefs between tasks: priors in a new task depend on posteriors from the previous task, with a 10-point increase predicting a 3–4 point higher subsequent prior. Second, within tasks, belief updating follows the Bayesian direction but is substantially conservative, proceeding at roughly half the normative Bayesian rate. Third, delegation is driven primarily by subjective beliefs about AI accuracy rather than self-confidence, though confidence independently reduces reliance when beliefs are held constant. Together, these findings show that users form global, path-dependent expectations about multipurpose AI systems, update them conservatively, and rely on AI primarily based on subjective beliefs rather than objective performance. We discuss implications for expectation calibration, reliance design, and the risks of belief spillovers in deployed LLM-based interfaces.

Privacy concerns significantly impact AI adoption, yet little is known about how information environments shape user responses to data leak threats. We conducted a 2 × 3 between-subjects experiment (N = 610) examining how risk versus ambiguity about privacy leaks affects the adoption of AI personalization. Participants chose between standard and AI-personalized product baskets, with personalization requiring data sharing that could leak to pricing algorithms. Under risk (30% leak probability), we found no difference in AI adoption between privacy-threatening and neutral conditions (ca. 50% adoption). Under ambiguity (10-50% range), privacy threats significantly reduced adoption compared to neutral conditions. This effect holds for sensitive demographic data as well as anonymized preference data. Users systematically over-bid for privacy disclosure labels, suggesting strong demand for transparency institutions. Notably, privacy leak threats did not affect subsequent bargaining behavior with algorithms. Our findings indicate that ambiguity over data leaks, rather than only privacy preferences per se, drives avoidance behavior among users towards personalized AI.

AI consumer markets are characterized by severe buyer-supplier market asymmetries. Complex AI systems can appear highly accurate while making costly errors or embedding hidden defects. While there have been regulatory efforts surrounding different forms of disclosure, large information gaps remain. This paper provides the first experimental evidence on the important role of information asymmetries and disclosure designs in shaping user adoption of AI systems. We systematically vary the density of low-quality AI systems and the depth of disclosure requirements in a simulated AI product market to gauge how people react to the risk of accidentally relying on a low-quality AI system. Then, we compare participants’ choices to a rational Bayesian model, analyzing the degree to which partial information disclosure can improve AI adoption. Our results underscore the deleterious effects of information asymmetries on AI adoption, but also highlight the potential of partial disclosure designs to improve the overall efficiency of human decision-making.

Compared to search engine result pages (SERPs), AI-generated podcasts represent a relatively new and relatively more passive modality of information consumption, delivering narratives in a naturally engaging format. As these two media increasingly converge in everyday information-seeking behavior, it is essential to explore how their interaction influences user attitudes, particularly in contexts involving controversial, value-laden, and often debated topics. Addressing this need, we aim to understand how information mediums of present-day SERPs and AI-generated podcasts interact to shape the opinions of users. To this end, through a controlled user study (N = 483), we investigated user attitudinal effects of consuming information via SERPs and AI-generated podcasts, focusing on how the sequence and modality of exposure shape user opinions. A majority of users in our study corresponded to attitude change outcomes, and we found an effect of sequence on attitude change. Our results further revealed a role of viewpoint bias and the degree of topic controversiality in shaping attitude change, although we found no effect of individual moderators.

Research on trust in AI is limited to several trustors (e.g., end-users) and trustees (especially AI systems), and empirical explorations remain in laboratory settings, overlooking factors that impact trust relations in the real world.Here, we broaden the scope of research by accounting for the supply chains that AI systems are part of.To CHI '25,

Since the explosion in popularity of ChatGPT, large language models (LLMs) have continued to impact our everyday lives. Equipped with external tools that are designed for a specific purpose (e.g., for flight booking or an alarm clock), LLM agents exercise an increasing capability to assist humans in their daily work. Although LLM agents have shown a promising blueprint as daily assistants, there is a limited understanding of how they can provide daily assistance based on planning and sequential decision making capabilities. We draw inspiration from recent work that has highlighted the value of 'LLM-modulo' setups in conjunction with humans-in-the-loop for planning tasks. We conducted an empirical study (N = 248) of LLM agents as daily assistants in six commonly occurring tasks with different levels of risk typically associated with them (e.g., flight ticket booking and credit card payments). To ensure user agency and control over the LLM agent, we adopted LLM agents in a plan-then-execute manner, wherein the agents conducted step-wise planning and step-by-step execution in a simulation environment. We analyzed how user involvement at each stage affects their trust and collaborative team performance. Our findings demonstrate that LLM agents can be a double-edged sword -- (1) they can work well when a high-quality plan and necessary user involvement in execution are available, and (2) users can easily mistrust the LLM agents with plans that seem plausible. We synthesized key insights for using LLM agents as daily assistants to calibrate user trust and achieve better overall task outcomes. Our work has important implications for the future design of daily assistants and human-AI collaboration with LLM agents.

Explainable artificial intelligence (XAI) methods are being proposed to help interpret and understand how AI systems reach specific predictions. Inspired by prior work on conversational user interfaces, we argue that augmenting existing XAI methods with conversational user interfaces can increase user engagement and boost user understanding of the AI system. In this paper, we explored the impact of a conversational XAI interface on users' understanding of the AI system, their trust, and reliance on the AI system. In comparison to an XAI dashboard, we found that the conversational XAI interface can bring about a better understanding of the AI system among users and higher user trust. However, users of both the XAI dashboard and conversational XAI interfaces showed clear overreliance on the AI system. Enhanced conversations powered by large language model (LLM) agents amplified over-reliance. Based on our findings, we reason that the potential cause of such overreliance is the illusion of explanatory depth that is concomitant with both XAI interfaces. Our findings have important implications for designing effective conversational XAI interfaces to facilitate appropriate reliance and improve human-AI collaboration. Code can be found at https://github.com/delftcrowd/IUI2025_ConvXAI

Conversational recommender systems (CRSs) provide users with an interactive means to express preferences and receive real-time personalized recommendations.The success of these systems is heavily influenced by the preference elicitation process.While existing research mainly focuses on what questions to ask during preference elicitation, there is a notable gap in understanding what role broader interaction patterns-including tone, pacing, and level of proactiveness-play in supporting users in completing a given task.This study investigates the impact of different conversational styles on preference elicitation, task performance, and user satisfaction with CRSs.We conducted a controlled experiment in the context of scientific literature recommendation, contrasting two distinct conversational styles-high involvement (fast-paced, direct, and proactive with frequent prompts) and high considerateness (polite and accommodating, prioritizing clarity and user comfort)-alongside a flexible experimental condition where users could switch between the two.Our results indicate that adapting conversational strategies based on user expertise and allowing flexibility between styles can enhance both user satisfaction and the effectiveness of recommendations in CRSs.Overall, our findings hold important implications for the design of future CRSs.

In today's society, where Artificial Intelligence (AI) has gained a vital role, concerns regarding user's trust have garnered significant attention.The use of AI systems in high-risk domains have often led users to either under-trust it, potentially causing inadequate reliance or over-trust it, resulting in over-compliance.Therefore, users must maintain an appropriate level of trust.Past research has indicated that explanations provided by AI systems can enhance user understanding of when to trust or not trust the system.However, the utility of presentation of different explanations forms still remains to be explored especially in high-risk domains.Therefore, this study explores the impact of different explanation types (text, visual, and hybrid) and user expertise (retired police officers and lay users) on establishing appropriate trust in AI-based predictive policing.While we observed that the hybrid form of explanations increased the subjective trust in AI for expert users, it did not led to better decision-making.Furthermore, no form of explanations helped build appropriate trust.The findings of our study emphasize the importance of re-evaluating the use of explanations to build [appropriate] trust in AI based systems especially when the system's use is questionable.Finally, we synthesize potential challenges and policy recommendations based on our results to design for appropriate trust in high-risk based AI-based systems.

Contestability has been proposed as a key element in designing algorithmic decision-making processes that safeguard decision subjects' rights to dignity and autonomy. However, little is known about how contestability can be operationalized based on decision subjects' needs and preferences. We address this research gap by identifying decision subjects' information and procedural needs for enacting meaningful contestability. To this end, we chose an illegal holiday rental detection scenario as our case; a high-risk decision-making process in the public sector. We conducted 21 semi-structured interviews with citizens with experience renting their homes out and different levels of AI literacy. We found that decision subjects request interventions that facilitate (1) cooperation in sense-making, (2) support in contestation acts, and (3) appropriate responsibility attribution. Our results highlight the cooperative work behind contestability, and motivate future efforts to structure individual and collective action, to personalize explanations for contestability, and to open up sites of contestation in AI pipelines.

The ability to make appropriate delegation decisions is an important prerequisite of effective human-AI collaboration. Recent work, however, has shown that people struggle to evaluate AI systems in the presence of forecasting errors, falling well short of relying on AI systems appropriately. We use a pre-registered crowdsourcing study (N = 611) to extend this literature by two underexplored crucial features of human AI decision-making: choice independence and error type. Subjects in our study repeatedly complete two prediction tasks and choose which predictions they want to delegate to an AI system. For one task, subjects receive a decision heuristic that allows them to make informed and relatively accurate predictions. The second task is substantially harder to solve, and subjects must come up with their own decision rule. We systematically vary the AI system’s performance such that it either provides the best possible prediction for both tasks or only for one of the two. Our results demonstrate that people systematically violate choice independence by taking the AI’s performance in an unrelated second task into account. Humans who delegate predictions to a superior AI in their own expertise domain significantly reduce appropriate reliance when the model makes systematic errors in a complementary expertise domain. In contrast, humans who delegate predictions to a superior AI in a complementary expertise domain significantly increase appropriate reliance when the model systematically errs in the human expertise domain. Furthermore, we show that humans differentiate between error types and that this effect is conditional on the considered expertise domain. This is the first empirical exploration of choice independence and error types in the context of human-AI collaboration. Our results have broad and important implications for the future design, deployment, and appropriate application of AI systems.

Human intelligence continues to be essential in building ground-truth data, training sets, and for evaluating a plethora of systems. The democratized and distributed nature of online crowd work — an attractive and accessible feature that has led to the proliferation of the paradigm — has also meant that crowd workers may not always feel connected to their remote peers. Despite the prevalence of collaborative crowdsourcing practices, workers on many microtask crowdsourcing platforms work on tasks individually and are seldom directly exposed to other crowd workers. In this context, improving worker engagement on microtask crowdsourcing platforms is an unsolved challenge. At the same time, fostering a sense of community among workers can improve the sustainability and working conditions in crowd work. This work aims to increase worker engagement in conversational microtask crowdsourcing by leveraging evolving avatars that workers can customize as they progress through monotonous task batches. We also aim to improve group identification in individual tasks by creating a community space where workers can share their avatars and feelings on task completion. To this end, we carried out a preregistered between-subjects controlled study (N = 680) spanning five experimental conditions and two task types. We found that evolving and customizable worker avatars can increase worker retention. The prospect of sharing worker avatars and task-related feelings in a community space did not consistently affect group identification. Our exploratory analysis indicated that workers who identify themselves as crowd workers experienced greater intrinsic motivation, subjective engagement, and perceived workload. Furthermore, we discuss how task differences shape the relative effectiveness of our interventions. Our findings have important theoretical and practical implications for designing conversational crowdsourcing tasks and in shaping new directions for research to improve crowd worker experiences.

While existing literature has explored and revealed several insights pertaining to the role of human factors (e.g., prior experience, domain knowledge) and attributes of AI systems (e.g., accuracy, trustworthiness), there is a limited understanding around how the important task characteristics of complexity and uncertainty shape human decision-making and human-AI team performance. In this work, we aim to address this research and empirical gap by systematically exploring how task complexity and uncertainty influence human-AI decision-making. Task complexity refers to the load of information associated with a task, while task uncertainty refers to the level of unpredictability associated with the outcome of a task. We conducted a between-subjects user study (N = 258) in the context of a trip-planning task to investigate the impact of task complexity and uncertainty on human trust and reliance on AI systems. Our results revealed that task complexity and uncertainty have a significant impact on user reliance on AI systems. When presented with complex and uncertain tasks, users tended to rely more on AI systems while demonstrating lower levels of appropriate reliance compared to tasks that were less complex and uncertain. In contrast, we found that user trust in the AI systems was not influenced by task complexity and uncertainty. Our findings can help inform the future design of empirical studies exploring human-AI decision-making. Insights from this work can inform the design of AI systems and interventions that are better aligned with the challenges posed by complex and uncertain tasks. Finally, the lens of diagnostic versus prognostic tasks can inspire the operationalization of uncertainty in human-AI decision-making studies.

Advances in generative AI, the proliferation of large multimodal models (LMMs), and democratized open access to these technologies have direct implications for the production and diffusion of misinformation. In this prequel, we address tackling misinformation in the unique and increasingly popular context of podcasts. The rise of podcasts as a popular medium for disseminating information across diverse topics necessitates a proactive strategy to combat the spread of misinformation. Inspired by the proven effectiveness of auditory alerts in contexts like collision alerts for drivers and error pings in mobile phones, our work envisions the application of auditory alerts as an effective tool to tackle misinformation in podcasts. We propose the integration of suitable auditory alerts to notify listeners of potential misinformation within the podcasts they are listening to, in real-time and without hampering listening experiences. We identify several opportunities and challenges in this path and aim to provoke novel conversations around instruments, methods, and measures to tackle misinformation in podcasts.

When using web search engines to conduct inquiries on debated topics, searchers’ interactions with search results are commonly affected by a combination of searcher and system biases. While prior work has mainly investigated these biases in isolation, there is a lack of a comprehensive understanding of web search on debated topics. Addressing this gap, we conducted an exploratory user study (N = 255), aimed at advancing the understanding of the intricate searcher-system interplay. Particularly, we investigated the relations between (i) search system exposure, searchers’ attitude strength, prior knowledge, and receptiveness to opposing views, (ii) search interactions, and (iii) post-search epistemic states. We observed that search interaction was shaped by search system exposure, attitude strength, and prior knowledge, and that attitude change was influenced by the level of confirmation bias and initial attitude strength, but not search system exposure. Insights from this work suggest the need to adapt interventions that mitigate the risks of searcher and system bias to searchers’ nuanced pre-search epistemic states. They further emphasize the threat of customizing the search ranking to enhance user satisfaction in the context of debated topics to responsible opinion formation.

With the integration of AI systems into our daily lives, human-AI collaboration has become increasingly prevalent. Prior work in this realm has primarily explored the effectiveness and performance of individual human and AI systems in collaborative tasks. While much of decision-making occurs within human peers and groups in the real world, there is a limited understanding of how they collaborate with AI systems. One of the key predictors of human-AI collaboration is the characteristics of the task at hand. Understanding the influence of task characteristics on human-AI collaboration is crucial for enhancing team performance and developing effective strategies for collaboration. Addressing a research and empirical gap, we seek to explore how the features of a task impact decision-making within human-AI group settings. In a 2 × 2 between-subjects study (N = 256) we examine the effects of task complexity and uncertainty on group performance and behaviour. The participants were grouped into pairs and assigned to one of four experimental conditions characterized by varying degrees of complexity and uncertainty. We found that high task complexity and high task uncertainty can negatively impact the performance of human-AI groups, leading to decreased group accuracy and increased disagreement with the AI system. We found that higher task complexity led to a higher efficiency in decision-making, while a higher task uncertainty had a negative impact on efficiency. Our findings highlight the importance of considering task characteristics when designing human-AI collaborative systems, as well as the future design of empirical studies exploring human-AI collaboration.

Concepts are an important construct in semantics, based on which humans understand the world with various levels of abstraction. With the recent advances in explainable artificial intelligence (XAI), concept-level explanations are receiving an increasing amount of attention from the broad research community. However, laypeople may find such explanations difficult to digest due to the potential knowledge gap and the concomitant cognitive load. Inspired by prior work that has explored analogies and sensemaking, we argue that augmenting concept-level explanations with analogical inference information from commonsense knowledge can be a potential solution to tackle this issue. To investigate the validity of our proposition, we first designed an effective analogy-based explanation generation method and collected 600 analogy-based explanations from 100 crowd workers. Next, we proposed a set of structured dimensions for the qualitative assessment of such explanations, and conducted an empirical evaluation of the generated analogies with experts. Our findings revealed significant positive correlations between the qualitative dimensions of analogies and the perceived helpfulness of analogy-based explanations, suggesting the effectiveness of the dimensions. To understand the practical utility and the effectiveness of analogybased explanations in assisting human decision-making, we conducted a follow-up empirical study (N = 280) on a skin cancer detection task with non-expert humans and an imperfect AI system. Thus, we designed a between-subjects study spanning five different experimental conditions with varying types of explanations. The results of our study confirmed that a knowledge gap can prevent participants from understanding concept-level explanations. Consequently, when only the target domain of our designed analogy-based explanation was provided (in a specific experimental condition), participants demonstrated relatively more appropriate reliance on the AI system. In contrast to our expectations, we found that analogies were not effective in fostering appropriate reliance. We carried out a qualitative analysis of the open-ended responses from participants in the study regarding their perceived usefulness of explanations and analogies. Our findings suggest that human intuition and the perceived plausibility of analogies may have played a role in affecting user reliance on the AI system. We also found that the understanding of commonsense explanations varied with the varying experience of the recipient user, which points out the need for further work on personalization when leveraging commonsense explanations. In summary, although we did not find quantitative support for our hypotheses around the benefits of using analogies, we found considerable qualitative evidence suggesting the potential of high-quality analogies in aiding non-expert users in their decision making with AI-assistance. These insights can inform the design of future methods for the generation and use of effective analogy-based explanations.

Pilot studies are an essential cornerstone of the design of crowdsourcing campaigns, yet they are often only mentioned in passing in the scholarly literature. A lack of details surrounding pilot studies in crowdsourcing research hinders the replication of studies and the reproduction of findings, stalling potential scientific advances. We conducted a systematic literature review on the current state of pilot study reporting at the intersection of crowdsourcing and HCI research. Our review of ten years of literature included 171 articles published in the proceedings of the Conference on Human Computation and Crowdsourcing (AAAI HCOMP) and the ACM Digital Library. We found that pilot studies in crowdsourcing research (i.e., crowd pilot studies) are often under-reported in the literature. Important details, such as the number of workers and rewards to workers, are often not reported. On the basis of our findings, we reflect on the current state of practice and formulate a set of best practice guidelines for reporting crowd pilot studies in crowdsourcing research. We also provide implications for the design of crowdsourcing platforms and make practical suggestions for supporting crowd pilot study reporting.

Fairness toolkits are developed to support machine learning (ML) practitioners in using algorithmic fairness metrics and mitigation methods. Past studies have investigated practical challenges for toolkit usage, which are crucial to understanding how to support practitioners. However, the extent to which fairness toolkits impact practitioners’ practices and enable reflexivity around algorithmic harms remains unclear (i.e., distributive unfairness beyond algorithmic fairness, and harms that are not related to the outputs of ML systems). Little is currently understood about the root factors that fragment practices when using fairness toolkits and how practitioners reflect on algorithmic harms. Yet, a deeper understanding of these facets is essential to enable the design of support tools for practitioners. To investigate the impact of toolkits on practices and identify factors that shape these practices, we carried out a qualitative study with 30 ML practitioners with varying backgrounds. Through a mixed within and between-subjects design, we tasked the practitioners with developing an ML model, and analyzed their reported practices to surface potential factors that lead to differences in practices. Interestingly, we found that fairness toolkits act as double-edge swords — with potentially positive and negative impacts on practices. Our findings showcase a plethora of human and organizational factors that play a key role in the way toolkits are envisioned and employed. These results bear implications for the design of future toolkits and educational training for practitioners and call for the creation of new policies to handle the organizational constraints faced by practitioners.

The dazzling promises of AI systems to augment humans in various tasks hinge on whether humans can appropriately rely on them. Recent research has shown that appropriate reliance is the key to achieving complementary team performance in AI-assisted decision making. This paper addresses an under-explored problem of whether the Dunning-Kruger Effect (DKE) among people can hinder their appropriate reliance on AI systems. DKE is a metacognitive bias due to which less-competent individuals overestimate their own skill and performance. Through an empirical study (N = 249), we explored the impact of DKE on human reliance on an AI system, and whether such effects can be mitigated using a tutorial intervention that reveals the fallibility of AI advice, and exploiting logic units-based explanations to improve user understanding of AI advice. We found that participants who overestimate their performance tend to exhibit under-reliance on AI systems, which hinders optimal team performance. Logic units-based explanations did not help users in either improving the calibration of their competence or facilitating appropriate reliance. While the tutorial intervention was highly effective in helping users calibrate their self-assessment and facilitating appropriate reliance among participants with overestimated self-assessment, we found that it can potentially hurt the appropriate reliance of participants with underestimated self-assessment. Our work has broad implications on the design of methods to tackle user cognitive biases while facilitating appropriate reliance on AI systems. Our findings advance the current understanding of the role of self-assessment in shaping trust and reliance in human-AI decision making. This lays out promising future directions for relevant HCI research in this community.

High-quality data plays a vital role in developing reliable image classification models. Despite that, what makes an image difficult to classify remains an unstudied topic. This paper provides a first-of-its-kind, model-agnostic characterization of image atypicality based on human understanding. We consider the setting of image classification “in the wild”, where a large number of unlabeled images are accessible, and introduce a scalable and effective human computation approach for proactive identification and characterization of atypical images. Our approach consists of i) an image atypicality identification and characterization task that presents to the human worker both a local view of visually similar images and a global view of images from the class of interest and ii) an automatic image sampling method that selects a diverse set of atypical images based on both visual and semantic features. We demonstrate the effectiveness and cost-efficiency of our approach through controlled crowdsourcing experiments and provide a characterization of image atypicality based on human annotations of 10K images. We showcase the utility of the identified atypical images by testing state-of-the-art image classification services against such images and provide an in-depth comparative analysis of the alignment between human- and machine-perceived image atypicality. Our findings have important implications for developing and deploying reliable image classification systems.

Quality control is an, if not the, essential challenge in crowdsourcing. Unsatisfactory responses from crowd workers have been found to particularly result from ambiguous and incomplete task descriptions, often from inexperienced task requesters. However, creating clear task descriptions with sufficient information is a complex process for requesters in crowdsourcing marketplaces. In this paper, we investigate the extent to which requesters can be supported effectively in this process through computational techniques. To this end, we developed a tool that enables requesters to iteratively identify and correct eight common clarity flaws in their task descriptions before deployment on the platform. The tool can be used to write task descriptions from scratch or to assess and improve the clarity of prepared descriptions. It employs machine learning-based natural language processing models trained on real-world task descriptions that score a given task description for the eight clarity flaws. On this basis, the requester can iteratively revise and reassess the task description until it reaches a sufficient level of clarity. In a first user study, we let requesters create task descriptions using the tool and rate the tool’s different aspects of helpfulness thereafter. We then carried out a second user study with crowd workers, as those who are confronted with such descriptions in practice, to rate the clarity of the created task descriptions. According to our results, 65% of the requesters classified the helpfulness of the information provided by the tool high or very high (only 12% as low or very low). The requesters saw some room for improvement though, for example, concerning the display of bad examples. Nevertheless, 76% of the crowd workers believe that the overall clarity of the task descriptions created by the requesters using the tool improves over the initial version. In line with this, the automatically-computed clarity scores of the edited task descriptions were generally higher than those of the initial descriptions, indicating that the tool reliably predicts the clarity of task descriptions in overall terms.

Recent advances in the performance of machine learning algorithms have led to the adoption of AI models in decision making contexts across various domains such as healthcare, finance, and education. Different research communities have attempted to optimize and evaluate human-AI team performance through empirical studies by increasing transparency of AI systems, or providing explanations to aid human understanding of such systems. However, the variety in decision making tasks considered and their operationalization in prior empirical work, has led to an opacity around how findings from one task or domain carry forward to another. The lack of a standardized means of considering task attributes prevents straightforward comparisons across decision tasks, thereby limiting the generalizability of findings. We argue that the lens of ‘task complexity’ can be used to tackle this problem of under-specification and facilitate comparison across empirical research in this area. To retrospectively explore how different HCI communities have considered the influence of task complexity in designing experiments in the realm of human-AI decision making, we survey literature and provide an overview of empirical studies on this topic. We found a serious dearth in the consideration of task complexity across various studies in this realm of research. Inspired by Robert Wood’s seminal work on the construct, we operationalized task complexity with respect to three dimensions (component, coordinative, and dynamic) and quantified the complexity of decision tasks in existing work accordingly. We then summarized current trends and proposed research directions for the future. Our study highlights the need to account for task complexity as an important design choice. This is a first step to help the scientific community in drawing meaningful comparisons across empirical studies in human-AI decision making and to provide opportunities to generalize findings across diverse domains and experimental settings.

AI systems are increasingly being used to support human decision making. It is important that AI advice is followed appropriately. However, according to existing literature, users typically under-rely or over-rely on AI systems, and this leads to sub-optimal team performance. In this context, we investigate the role of stated system accuracy by contrasting the lack of system information with the presence of system accuracy in a loan prediction task. We explore how the degree to which humans understand system accuracy influences their reliance on the AI system, by investigating numeracy levels and with the aid of analogies to explain system accuracy in a first of its kind between-subjects study (N=281). We found that explaining the stated accuracy of a system using analogies failed to help users rely on the AI systemappropriately (i.e., the tendency of users to rely on the system when the system is correct, or on themselves otherwise). To eliminate the impact of subjective attitudes towards analogy domains, we conducted a within-subjects study (N=248) where each participant worked on tasks with analogy-based explanations from different domains. Results from this second study confirmed that explaining stated accuracy of the system with analogies was not sufficient to facilitate appropriate reliance on the AI system in the context of loan prediction tasks, irrespective of individual user differences. Based on our findings from the two studies, we reason that the under-reliance on the AI system may be a result of users' overestimation of their own ability to solve the given task. Thus, although familiar analogies can be effective in improving the intelligibility of stated accuracy of the system, an improved understanding of system accuracy does not necessarily lead to improved system reliance and team performance.

As algorithms are increasingly augmenting and substituting human decision-making, understanding how the introduction of computational agents changes the fundamentals of human behavior becomes vital. This pertains to not only users, but also those parties who face the consequences of an algorithmic decision. In a controlled experiment with 480 participants, we exploit an extended version of two-player ultimatum bargaining where responders choose to bargain with either another human, another human with an AI decision aid or an autonomous AI-system acting on behalf of a passive human proposer. Our results show strong responder preferences against the algorithm, as most responders opt for a human opponent and demand higher compensation to reach a contract with autonomous agents. To map these preferences to economic expectations, we elicit incentivized subject beliefs about their opponent’s behavior. The majority of responders maximize their expected value when this is line with approaching the human proposer. In contrast, responders predicting income maximization for the autonomous AI-system overwhelmingly override economic self-interest to avoid the algorithm.

Conversational agents are being widely adopted across several domains to serve a variety of purposes ranging from providing intelligent assistance to companionship. Recent literature has shown that users develop intuitive folk theories and a metaphorical understanding of conversational agents (CAs) due to the lack of a mental model of the agents. However, investigation of metaphorical agent representation in the HCI community has mainly focused on the human level, despite non-human metaphors for agents being prevalent in the real world. We adopted Lakoff and Turner’s ‘Great Chain of Being’ framework to systematically investigate the impact of using non-human metaphors to represent conversational agents on worker engagement in crowdsourcing marketplaces. We designed a text-based conversational agent that assists crowd workers in task execution. Through a between-subjects experimental study (N = 341), we explored how different human and non-human metaphors affect worker engagement, the perceived cognitive load of workers, intrinsic motivation, and their trust in the agents. Our findings bridge the gap of how users experience CAs with non-human metaphors in the context of conversational crowdsourcing.

A wealth of evidence across several domains indicates that goal setting improves performance and learning by enabling individuals to commit their thoughts and actions to goal achievement. Recently, researchers have begun studying the effects of goal setting in paid crowdsourcing to improve the quality and quantity of contributions, increase learning gains, and hold participants accountable for contributing more effectively. However, there is a lack of research addressing crowd workers' goal-setting practices, how they are currently pursuing them, and the challenges that they face. This information is essential for researchers and developers to create tools that assist crowd workers in pursuing their goals more effectively, thereby improving the quality of their contributions. This paper addresses these gaps by conducting mixed-method research in which we surveyed 205 workers from two crowdsourcing platforms -- Amazon Mechanical Turk (MTurk) and Prolific -- about their goal-setting practices. Through a 14-item survey, we asked workers regarding the types of goals they create, their goal achievement strategies, potential barriers that impede goal attainment, and their use of software tools for effective goal management. We discovered that (a) workers actively create intrinsic and extrinsic goals; (b) use a combination of tools for goal management; (c) medical issues and a busy lifestyle are some obstacles to their goal achievement; and (d) we gathered novel features for future goal management tools. Our findings shed light on the broader implications of developing goal management tools to improve workers' well-being.

Human input is pivotal in building reliable and robust artificial intelligence systems. By providing a means to gather diverse, high-quality, representative, and cost-effective human input on demand, microtask crowdsourcing marketplaces have thrived. Despite the unmistakable benefits available from online crowd work, the lack of health provisions and safeguards, along with existing work practices threatens the sustainability of this paradigm. Prior work has investigated worker engagement and mental health, yet no such investigations into the effects of crowd work on the physical health of workers have been undertaken. Crowd workers complete their work in various sub-optimal work environments, often using a conventional input modality of a mouse and keyboard. The repetitive nature of microtask crowdsourcing can lead to stress-related injuries, such as the well-documented carpal tunnel syndrome. It is known that stretching exercises can help reduce injuries and discomfort in office workers. Gestures, the act of using the body intentionally to affect the behavior of an intelligent system, can serve as both stretches and an alternative form of input for microtasks. To better understand the usefulness of the dual-purpose input modality of ergonomically-informed gestures across different crowdsourced microtasks, we carried out a controlled 2 x 3 between-subjects study (N=294). Considering the potential benefits of gestures as an input modality, our results suggest a real trade-off between worker accuracy in exchange for potential short to long-term health benefits.

With recent advances in explainable artificial intelligence (XAI), researchers have started to pay attention to concept-level explanations, which explain model predictions with a high level of abstraction. However, such explanations may be difficult to digest for laypeople due to the potential knowledge gap and the concomitant cognitive load. Inspired by recent work, we argue that analogy-based explanations composed of commonsense knowledge may be a potential solution to tackle this issue. In this paper, we propose analogical inference as a bridge to help end-users leverage their commonsense knowledge to better understand the concept-level explanations. Specifically, we design an effective analogy-based explanation generation method and collect 600 analogy-based explanations from 100 crowd workers. Furthermore, we propose a set of structured dimensions for the qualitative assessment of analogy-based explanations and conduct an empirical evaluation of the generated analogies with experts. Our findings reveal significant positive correlations between the qualitative dimensions of analogies and the perceived helpfulness of analogy-based explanations. These insights can inform the design of future methods for the generation of effective analogy-based explanations. We also find that the understanding of commonsense explanations varies with the experience of the recipient user, which points out the need for further work on personalization when leveraging commonsense explanations.

Different input modalities have been proposed and employed in technological landscapes like microtask crowdsourcing. However, sign language remains an input modality that has received little attention. Despite the fact that thousands of people around the world primarily use sign language, very little has been done to include them in such technological landscapes. We aim to address this gap and take a step towards the inclusion of deaf and mute people in microtask crowdsourcing. We first identify various microtasks which can be adapted to use sign language as input, while elucidating the challenges it introduces. We built a system called ‘SignUpCrowd’ that can be used to support sign language input for microtask crowdsourcing. We carried out a between-subjects study (N=240) to understand the effectiveness of sign language as an input modality for microtask crowdsourcing in comparison to prevalent textual and click input modalities. We explored this through the lens of visual question answering and sentiment analysis tasks by recruiting workers from the Prolific crowdsourcing platform. Our results indicate that sign language as an input modality in microtask crowdsourcing is comparable to the prevalent standards of using text and click input. Although people with no knowledge of sign language found it difficult to use, this input modality has the potential to broaden participation in crowd work. We highlight evidence suggesting the scope for sign language as a viable input type for microtask crowdsourcing. Our findings pave the way for further research to introduce sign language in real-world applications and create an inclusive technological landscape that more people can benefit from.

Advances in artificial intelligence and machine learning have led to a steep rise in the adoption of AI to augment or support human decision-making across domains. There has been an increasing body of work addressing the benefits of model interpretability and explanations to help end-users or other stakeholders decipher the inner workings of the so-called ”black box AI systems”. Yet, little is currently understood about the role of modalities through which explanations can be communicated (e.g., text, visualizations, or audio) to inform, augment, and shape human decision-making. In our work, we address this research gap through the lens of a credibility assessment system. Considering the deluge of information available through various channels, people constantly make decisions while considering the perceived credibility of the information they consume. However, with an increasing information overload, assessing the credibility of the information we encounter is a non-trivial task. To help users in this task, automated credibility assessment systems have been devised as decision support systems in various contexts (e.g., assessing the credibility of news or social media posts). However, for these systems to be effective in supporting users, they need to be trusted and understood. Explanations have been shown to play an essential role in informing users’ reliance on decision support systems. In this paper, we investigate the influence of explanation modalities on an AI-assisted credibility assessment task. We use a between-subjects experiment (N = 375), spanning six different explanation modalities, to evaluate the role of explanation modality on the accuracy of AI-assisted decision outcomes, the perceived system trust among users, and system usability. Our results indicate that explanations play a significant role in shaping users’ reliance on the decision support system and, thereby, the accuracy of decisions made. We found that users performed with higher accuracy while assessing the credibility of statements in the presence of explanations. We also found that users had a significantly harder time agreeing on statement credibility without explanations. With explanations present, text and audio explanations were more effective than graphic explanations. Additionally, we found that combining graphical with text and/or audio explanations were significantly effective. Such combinations of modalities led to a higher user performance than using graphical explanations alone.

Access to commonsense knowledge is receiving renewed interest for developing neuro-symbolic AI systems, or debugging deep learning models. Little is currently understood about the types of knowledge that can be gathered using existing knowledge elicitation methods. Moreover, these methods fall short of meeting the evolving requirements of several downstream AI tasks. To this end, collecting broad and tacit knowledge, in addition to negative or discriminative knowledge can be highly useful. Addressing this research gap, we developed a novel game with a purpose, ‘FindItOut’, to elicit different types of knowledge from human players through easily configurable game mechanics. We recruited 125 players from a crowdsourcing platform, who played 2430 rounds, resulting in the creation of more than 150k tuples of knowledge. Through an extensive evaluation of these tuples, we show that FindItOut can successfully result in the creation of plural knowledge with a good player experience. We evaluate the efficiency of the game (over 10 × higher than a reference baseline) and the usefulness of the resulting knowledge, through the lens of two downstream tasks — commonsense question answering and the identification of discriminative attributes. Finally, we present a rigorous qualitative analysis of the tuples’ characteristics, that informs the future use of FindItOut across various researcher and practitioner communities.

Trust is an important component of human-AI relationships and plays a major role in shaping the reliance of users on online algorithmic decision support systems. With recent advances in natural language processing, text and voice-based conversational interfaces have provided users with new ways of interacting with such systems. Despite the growing applications of conversational user interfaces (CUIs), little is currently understood about the suitability of such interfaces for decision support and how CUIs inspire trust among humans engaging with decision support systems. In this work, we aim to address this gap and answer the following question: to what extent can a conversational interface build user trust in decision support systems in comparison to a conventional graphical user interface? To this end, we built a text-based conversational interface, and a conventional web-based graphical user interface. These served as the means for users to interact with an online decision support system to help them find housing, given a fixed set of constraints. To understand how the accuracy of the decision support system moderates user behavior and trust across the two interfaces, we considered an accurate and inaccurate system. We carried out a 2 × 2 between-subjects study (N = 240) on the Prolific crowdsourcing platform. Our findings show that the conversational interface was significantly more effective in building user trust and satisfaction in the online housing recommendation system when compared to the conventional web interface. Our results highlight the potential impact of conversational interfaces for trust development in decision support systems.

Unknown unknowns represent a major challenge in reliable image recognition. Existing methods mainly focus on unknown unknowns identification, leveraging human intelligence to gather images that are potentially difficult for the machine. To drive a deeper understanding of unknown unknowns and more effective identification and treatment, this paper focuses on unknown unknowns characterization. We introduce a human-in-the-loop, semantic analysis framework for characterizing unknown unknowns at scale. We engage humans in two tasks that specify what a machine should know and describe what it really knows, respectively, both at the conceptual level, supported by information extraction and machine learning interpretability methods. Data partitioning and sampling techniques are employed to scale out human contributions in handling large data. Through extensive experimentation on scene recognition tasks, we show that our approach provides a rich, descriptive characterization of unknown unknowns and allows for more effective and cost-efficient detection than the state of the art.

Crowd-powered conversational systems (CPCS) are gaining considerable attention for their potential utility in a variety of application domains, for which automated conversational interfaces are still too limited. CPCS currently suffer from long response delays, which hampers their potential as conversational partners. The majority of prior work in this area has focused on demonstrating the feasibility of the approach and improving performance, while evaluation studies have primarily focused on response latency and ways to reduce it. Relatively little is currently known about how response delays in a CPCS can affect user experience. While the importance of reducing response latency is widely recognized in the broader field of human-computer interaction, little attention has been paid to how response quality, response delay, conversational context, and the complexity of the task affect how users experience the conversation, and how they perceive waiting for responses in particular. We conducted a between-subjects experiment (N = 478), to examine the influence of these four factors on the overall waiting experience of users. Results show that users 1) evaluated the waiting experience more negatively when the response delay was longer than 8 seconds, 2) underestimated the elapsed time but experienced more frustration in tasks with high complexity, 3) underestimated the elapsed time and experienced less frustration with high quality bot's utterances, 4) judged response delays to be slightly longer, and experienced more frustration in an emotion-centric CPCS compared to a task-centric CPCS. Our insights can inform the design of future CPCSs with regards to defining performance requirements and anticipating their potential impact on the user experience they can facilitate.

Active reading strategies---such as content annotations (through the use of highlighting and note-taking, for example)---have been shown to yield improvements to a learner's knowledge and understanding of the topic being explored. This has been especially notable in long and complex learning endeavours. With web search engines nowadays used as the primary gateway for learners (or users) to find content that helps them realise their learning goals, they are often poorly equipped with the necessary tools to aid in sense-making, an important aspect of theSearch as Learning (SAL) process. Within theInformation Retrieval (IR) community, research efforts have explored ways to keep track of users' search context by providing a notepad-like interface for the collection of relevant articles, and aid them during the exploratory search process. However, these studies did not explicitly measure the effect that such tools have on knowledge and understanding during a complex, learning-oriented search task. In this paper, we address this research gap by carrying out an Interactive IR experiment with highlighting and note-taking tools built into the search interface. We conducted a crowdsourced between-subjects study (N=115), where participants were assigned to one of four conditions: (i) control (a standard web search interface); (ii) high (highlighting enabled);(iii) note (note-taking enabled); and (iv) highnote (both highlighting and note-taking enabled). We assess participants' learning with a recall-oriented vocabulary learning task, and a cognitively more taxing essay writing task. We find that(i) active reading tools do not aid in the vocabulary learning task. However,(ii) participants in high covered 34% more subtopics, and participants in note covered 34% more facts in their essays when compared to control. Furthermore, (iii) we observed that incorporating active learning tools significantly changed the search behaviour of participants across a number of measures. This is the first work that sheds light on the effect of active reading tools on the SAL process, with important design implications for learning-oriented search systems.

Crowd-Powered Conversational Systems (CPCS) are gaining traction due to their potential utility in a range of application fields where automated conversational interfaces are still inadequate. Currently, long response times negatively impact CPCSs, limiting their potential application as conversational partners. Related research has focused on developing algorithms for swiftly hiring workers and synchronous crowd coordination techniques to ensure high-quality work. Evaluation studies typically concern system reaction times and performance measurements, but have so far not examined the effects of extended wait times on users. The goal of this study, based on time perception models, is to explore how effective different time fillers are at reducing the negative impacts of waiting in CPCSs. To this end, we conducted a rigorous simulation-based between-subjects (N = 930) study on the Prolific crowdsourcing platform to assess the influence of different filler types across three levels of delay (8, 16 & 32s) for Information Retrieval (IR) and stress management tasks. Our results show that asking users to perform secondary tasks (e.g., microtasks or breathing exercises) while waiting for longer periods of time helped divert their attention away from timekeeping, increased their engagement, and resulted in shorter perceived waiting times. For shorter delays, conversational fillers generated more intense immersion and contributed to shorten the perception of time.

Recent research has demonstrated that cognitive biases such as the confirmation bias or the anchoring effect can negatively affect the quality of crowdsourced data. In practice, however, such biases go unnoticed unless specifically assessed or controlled for. Task requesters need to ensure that task workflow and design choices do not trigger workers’ cognitive biases. Moreover, to facilitate the reuse of crowdsourced data collections, practitioners can benefit from understanding whether and which cognitive biases may be associated with the data. To this end, we propose a 12-item checklist adapted from business psychology to combat cognitive biases in crowdsourcing. We demonstrate the practical application of this checklist in a case study on viewpoint annotations for search results. Through a retrospective analysis of relevant crowdsourcing research that has been published at HCOMP in 2018, 2019, and 2020, we show that cognitive biases may often affect crowd workers but are typically not considered as potential sources of poor data quality. The checklist we propose is a practical tool that requesters can use to improve their task designs and appropriately describe potential limitations of collected data. It contributes to a body of efforts towards making human-labeled data more reliable and reusable.

Music content annotation campaigns are common on paid crowdsourcing platforms. Crowd workers are expected to annotate complicated music artefacts, which can demand certain skills and expertise. Traditional methods of participant selection are not designed to capture these kind of domain-specific skills and expertise, and often domain-specific questions fall under the general demographics category. Despite the popularity of such tasks, there is a general lack of deeper understanding of the distribution of musical properties - especially auditory perception skills - among workers. To address this knowledge gap, we conducted a user study (N=100) on Prolific. We asked workers to indicate their musical sophistication through a questionnaire and assessed their music perception skills through an audio-based skill test. The goal of this work is to better understand the extent to which crowd workers possess higher perceptions skills, beyond their own musical education level and self reported abilities. Our study shows that untrained crowd workers can possess high perception skills on the music elements of melody, tuning, accent and tempo; skills that can be useful in a plethora of annotation tasks in the music domain.

In web search on debated topics, algorithmic and cognitive biases strongly influence how users consume and process information. Recent research has shown that this can lead to a search engine manipulation effect (SEME): when search result rankings are biased towards a particular viewpoint, users tend to adopt this favored viewpoint. To better understand the mechanisms underlying SEME, we present a pre-registered, 5 x 3 factorial user study investigating whether order effects (i.e., users adopting the viewpoint pertaining to higher-ranked documents) can cause SEME. For five different debated topics, we evaluated attitude change after exposing participants with mild pre-existing attitudes to search results that were overall viewpoint-balanced but reflected one of three levels of algorithmic ranking bias. We found that attitude change did not differ across levels of ranking bias and did not vary based on individual user differences. Our results thus suggest that order effects may not be an underlying mechanism of SEME. Exploratory analyses lend support to the presence of exposure effects (i.e., users adopting the majority viewpoint among the results they examine) as a contributing factor to users' attitude change. We discuss how our findings can inform the design of user bias mitigation strategies.

There is a growing use of intelligent systems to support human decision-making across several domains. Trust in intelligent systems, however, is pivotal in shaping their widespread adoption. Little is currently understood about how trust in an intelligent system evolves over time and how it is mediated by the accuracy of the system. We aim to address this knowledge gap by exploring trust formation over time and its relation to system accuracy. To that end, we built an intelligent house recommendation system and carried out a longitudinal study consisting of 201 participants across 3 sessions in a week. In each session, participants were tasked with finding housing that fit a given set of constraints using a conventional web interface that reflected a typical housing search website. Participants could choose to use an intelligent decision support system to help them find the right house. Depending on the group, participants received a variation of accurate or inaccurate advice from the intelligent system throughout each session. We measured trust using a trust in automation scale at the end of each session.

Understanding the influence of users’ opinions on their search behavior together with their inherent biases in web search has garnered widespread interest in recent times. This is largely due to the implications of promoting critical thinking, explaining phenomena such as political polarization, or the manifestation of echo chambers. It is important to understand how personal opinions can bias users’ interaction with search results. Moreover, there is a lack of understanding of the impact of user search intents, namely non-purposeful browsing versus searching with a pre-defined goal, on users’ interactions with search results. We take a step towards bridging this knowledge gap through an empirical study in this paper. To do so, we select two controversial topics in abortion and gun control, and invite users to learn about them through ‘Purposeless’ and ‘Purposeful’ web searching. Our findings suggest that users with strong personal opinions exhibit biased interactions with the search results. However, the effect of users’ opinions on their interactions with search results can differ depending on whether users search purposelessly or with a purpose. Our findings advance the current understanding of the effect of users’ opinions in web search sessions, and show that users’ search intents shape their interaction with search results. This work has broad design implications on dealing with bias in interactive information retrieval systems.

Abstract Web search is among the most frequent online activities. In this context, widespread informational queries entail user intentions to obtain knowledge with respect to a particular topic or domain. To serve learning needs better, recent research in the field of interactive information retrieval has advocated the importance of moving beyond relevance ranking of search results and considering a user’s knowledge state within learning oriented search sessions. Prior work has investigated the use of supervised models to predict a user’s knowledge gain and knowledge state from user interactions during a search session. However, the characteristics of the resources that a user interacts with have neither been sufficiently explored, nor exploited in this task. In this work, we introduce a novel set of resource-centric features and demonstrate their capacity to significantly improve supervised models for the task of predicting knowledge gain and knowledge state of users in Web search sessions. We make important contributions, given that reliable training data for such tasks is sparse and costly to obtain. We introduce various feature selection strategies geared towards selecting a limited subset of effective and generalizable features.

The future of crowd work has been identified to depend on worker satisfaction, but we lack a thorough understanding of how worker satisfaction can be increased in microtask crowdsourcing. Prior work has shown that one solution is to build tasks that are engaging. To facilitate engagement, two methods that have received attention in recent HCI literature are the use of video games and conversational interfaces. While these are largely different techniques, they aim for the same goal of reducing worker burden and increasing engagement in a task. On one hand, video games have huge motivation potential and translating game design elements for motivational purposes has shown positive effects. Recent work in games research has shown that the use of player avatars is effective in fostering interest, enjoyment, and other aspects pertaining to intrinsic motivation. On the other hand, conversational interfaces have been argued to have advantages over traditional GUIs due to facilitating a more human-like interaction. Conversational microtasking has recently been proposed to improve worker engagement in microtask marketplaces. The contexts of games and crowd work are underlined by the need to motivate and engage participants, yet the potential of using worker avatars to promote self-identification and improve worker satisfaction in microtask crowdsourcing has remained unexplored. Addressing this knowledge gap, we carried out a between-subject study involving 360 crowd workers. We investigated how worker avatars influence quality related outcomes of workers and their perceived experience, in conventional web and novel conversational interfaces. We equipped workers with the functionality of customizing their avatars, and selecting characterizations for their avatars, to understand whether identifying with an avatar can increase the motivation of workers. We found that using worker avatars with conversational interfaces can effectively reduce cognitive workload and increase worker retention. Our results indicate the occurrence of similarity and wishful avatar identification in crowdsourcing. Our findings have important implications in alleviating workers' perceived workload and on the design of crowdsourcing microtasks.

Crowdsourcing has become a standard methodology to collect manually annotated data such as relevance judgments at scale. On crowdsourcing platforms like Amazon MTurk or FigureEight, crowd workers select tasks to work on based on different dimensions such as task reward and requester reputation. Requesters then receive the judgments of workers who self-selected into the tasks and completed them successfully. Several crowd workers, however, preview tasks, begin working on them, reaching varying stages of task completion without finally submitting their work. Such behavior results in unrewarded effort which remains invisible to requesters. In this paper, we conduct an investigation of the phenomenon of task abandonment, the act of workers previewing or beginning a task and deciding not to complete it. We follow a three-fold methodology which includes 1) investigating the prevalence and causes of task abandonment by means of a survey over different crowdsourcing platforms, 2) data-driven analysis of logs collected during a large-scale relevance judgment experiment, and 3) controlled experiments measuring the effect of different dimensions on abandonment. Our results show that task abandonment is a widely spread phenomenon. Apart from accounting for a considerable amount of wasted human effort, this bears important implications on the hourly wages of workers as they are not rewarded for tasks that they do not complete. We also show how task abandonment may have strong implications on the use of collected data (for example, on the evaluation of Information Retrieval systems).

The rise in popularity of conversational agents has enabled humans to interact with machines more naturally. Recent work has shown that crowd workers in microtask marketplaces can complete a variety of human intelligence tasks (HITs) using conversational interfaces with similar output quality compared to the traditional Web interfaces. In this paper, we investigate the effectiveness of using conversational interfaces to improve worker engagement in microtask crowdsourcing. We designed a text-based conversational agent that assists workers in task execution, and tested the performance of workers when interacting with agents having different conversational styles. We conducted a rigorous experimental study on Amazon Mechanical Turk with 800 unique workers, to explore whether the output quality, worker engagement and the perceived cognitive load of workers can be affected by the conversational agent and its conversational styles. Our results show that conversational interfaces can be effective in engaging workers, and a suitable conversational style has potential to improve worker engagement.

On-demand emotional support is an expensive and elusive societal need that is exacerbated in difficult times — as witnessed during the COVID-19 pandemic. Prior work in affective crowdsourcing has examined ways to overcome technical challenges for providing on-demand emotional support to end users. This can be achieved by training crowd workers to provide thoughtful and engaging on-demand emotional support. Inspired by recent advances in conversational user interface research, we investigate the efficacy of a conversational user interface for training workers to deliver psychological support to users in need. To this end, we conducted a between-subjects experimental study on Prolific, wherein a group of workers (N=200) received training on motivational interviewing via either a conversational interface or a conventional web interface. Our results indicate that training workers in a conversational interface yields both better worker performance and improves their user experience in on-demand stress management tasks.

Recent advances in machine learning have led to the widespread adoption of ML models for decision support systems. However, little is known about how the introduction of such systems affects the behavior of human stakeholders. This pertains both to the people using the system, as well as those who are affected by its decisions. To address this knowledge gap, we present a series of ultimatum bargaining game experiments comprising 1178 participants. We find that users are willing to use a black-box decision support system and thereby make better decisions. This translates into higher levels of cooperation and better market outcomes. However, because users under-weigh algorithmic advice, market outcomes remain far from optimal. Explanations increase the number of unique system inquiries, but users appear less willing to follow the system’s recommendation. People who negotiate with a user who has a decision support system, but cannot use one themselves, react to its introduction by demanding a better deal for themselves, thereby decreasing overall cooperation levels. This effect is largely driven by the percentage of participants who perceive the system’s availability as unfair. Interpretability mitigates perceptions of unfairness. Our findings highlight the potential for decision support systems to further human cooperation, but also the need for regulators to consider heterogeneous stakeholder reactions. In particular, higher levels of transparency might inadvertently hurt cooperation through changes in fairness perceptions.

Crowdsourcing is a popular technique to collect large amounts of human-generated labels, such as relevance judgments used to create information retrieval (IR) evaluation collections. Previous research has shown how collecting high quality labels from a crowdsourcing platform can be challenging. Existing quality assurance techniques focus on answer aggregation or on the use of gold questions where ground-truth data allows to check for the quality of the responses.

Paid micro-task crowdsourcing has gained in popularity partly due to the increasing need for large-scale manually labelled datasets which are often used to train and evaluate Artificial Intelligence systems. Modern paid crowdsourcing platforms use a piecework approach to rewards, meaning that workers are paid for each task they complete, given that their work quality is considered sufficient by the requester or the platform. Such an approach creates risks for workers; their work may be rejected without being rewarded, and they may be working on poorly rewarded tasks, in light of the disproportionate time required to complete them. As a result, recent research has shown that crowd workers may tend to choose specific, simple, and familiar tasks and avoid new requesters to manage these risks. In this paper, we propose a novel crowdsourcing reward mechanism that allows workers to share these risks and achieve a standardized hourly wage equal for all participating workers. Reward-focused workers can thereby take up challenging and complex HITs without bearing the financial risk of not being rewarded for completed work. We experimentally compare different crowd reward schemes and observe their impact on worker performance and satisfaction. Our results show that 1) workers clearly perceive the benefits of the proposed reward scheme, 2) work effectiveness and efficiency are not impacted as compared to those of the piecework scheme, and 3) the presence of slow workers is limited and does not disrupt the proposed cooperation-based approaches.

Crowdsourcing marketplaces have provided a large number of opportunities for online workers to earn a living. To improve satisfaction and engagement of such workers, who are vital for the sustainability of the marketplaces, recent works have used conversational interfaces to support the execution of a variety of crowdsourcing tasks. The rationale behind using conversational interfaces stems from the potential engagement that conversation can stimulate. Prior works in psychology have also shown that conversational styles can play an important role in communication. There are unexplored opportunities to estimate a worker's conversational style with an end goal of improving worker satisfaction, engagement and quality. Addressing this knowledge gap, we investigate the role of conversational styles in conversational microtask crowdsourcing. To this end, we design a conversational interface which supports task execution, and we propose methods to estimate the conversational style of a worker. Our experimental setup was designed to empirically observe how conversational styles of workers relate with quality-related outcomes. Results show that even a naive supervised classifier can predict the conversation style with high accuracy (80%), and crowd workers with an Involvement conversational style provided a significantly higher output quality, exhibited a higher user engagement and perceived less cognitive task load in comparison to their counterparts. Our findings have important implications on task design with respect to improving worker performance and their engagement in microtask crowdsourcing.

Crowdsourced data acquired from tasks that comprise a subjective component (e.g. opinion detection, sentiment analysis) is potentially affected by the inherent bias of crowd workers who contribute to the tasks. This can lead to biased and noisy ground-truth data, propagating the undesirable bias and noise when used in turn to train machine learning models or evaluate systems. In this work, we aim to understand the influence of workers' own opinions on their performance in the subjective task of bias detection. We analyze the influence of workers' opinions on their annotations corresponding to different topics. Our findings reveal that workers with strong opinions tend to produce biased annotations. We show that such bias can be mitigated to improve the overall quality of the data collected. Experienced crowd workers also fail to distance themselves from their own opinions to provide unbiased annotations.

User-centered approaches have been extensively studied and used in the area of struggling search. Related research has targeted key aspects of users such as user satisfaction or frustration, and search success or failure, using a variety of experimental methods including laboratory user studies, in-situ explicit feedback from searchers and by using crowdsourcing. Such studies are valuable in advancing the understanding of search difficulty from a user's perspective, and yield insights that can directly improve search systems and their evaluation. However, little is known about how user moods influence their interactions with a search system or their perception of struggling. In this work, we show that a user's own mood. can systematically bias the user's perception, and experience while interacting with a search system and trying to satisfy an information need. People who are in activated-(un)pleasant moods tend to issue more queries than people in deactivated or neutral moods. Those in an unpleasant mood perceive a higher level of difficulty. Our insights extend the current understanding of struggling search tasks and have important implications on the design and evaluation of search systems supporting such tasks.

Conversational interfaces can facilitate human-computer interactions. Whether or not conversational interfaces can improve worker experience and work quality in crowdsourcing marketplaces has remained unanswered. We investigate the suitability of text-based conversational interfaces for microtask crowdsourcing. We designed a rigorous experimental campaign aimed at gauging the interest and acceptance by crowdworkers for this type of work interface. We compared Web and conversational interfaces for five common microtask types and measured the execution time, quality of work, and the perceived satisfaction of 316 workers recruited from the FigureEight platform. We show that conversational interfaces can be used effectively for crowdsourcing microtasks, resulting in a high satisfaction from workers, and without having a negative impact on task execution time or work quality.

Crowdsourcing has become a standard methodology to collect manually annotated data such as relevance judgments at scale. On crowdsourcing platforms like Amazon MTurk or FigureEight, crowd workers select tasks to work on based on different dimensions such as task reward and requester reputation. Requesters then receive the judgments of workers who self-selected into the tasks and completed them successfully. Several crowd workers, however, preview tasks, begin working on them, reaching varying stages of task completion without finally submitting their work. Such behavior results in unrewarded effort which remains invisible to requesters. In this paper, we conduct the first investigation into the phenomenon of task abandonment, the act of workers previewing or beginning a task and deciding not to complete it. We follow a three-fold methodology which includes 1) investigating the prevalence and causes of task abandonment by means of a survey over different crowdsourcing platforms, 2) data-driven analyses of logs collected during a large-scale relevance judgment experiment, and 3) controlled experiments measuring the effect of different dimensions on abandonment. Our results show that task abandonment is a widely spread phenomenon. Apart from accounting for a considerable amount of wasted human effort, this bears important implications on the hourly wages of workers as they are not rewarded for tasks that they do not complete. We also show how task abandonment may have strong implications on the use of collected data (for example, on the evaluation of IR systems).

Complex machine learning models are deployed in several critical domains including healthcare and autonomous vehicles nowadays, albeit as functional blackboxes. Consequently, there has been a recent surge in interpreting decisions of such complex models in order to explain their actions to humans. Models which correspond to human interpretation of a task are more desirable in certain contexts and can help attribute liability, build trust, expose biases and in turn build better models. It is therefore crucial to understand how and which models conform to human understanding of tasks. In this paper we present a large-scale crowdsourcing study that reveals and quantifies the dissonance between human and machine understanding, through the lens of an image classification task. In particular, we seek to answer the following questions: Which (well performing) complex ML models are closer to humans in their use of features to make accurate predictions? How does task difficulty affect the feature selection capability of machines in comparison to humans? Are humans consistently better at selecting features that make image recognition more accurate? Our findings have important implications on human-machine collaboration, considering that a long term goal in the field of artificial intelligence is to make machines capable of learning and reasoning like humans.

Web search is frequently used by people to acquire new knowledge and to satisfy learning-related objectives, but little is known about how a user»s knowledge evolves through the course of a search session. We present a study addressing the knowledge gain of users in informational search sessions. Using crowdsourcing, we recruited 500 distinct users and orchestrated real-world search sessions spanning 10 different topics and information needs. By using scientifically formulated knowledge tests we calibrated the knowledge of users before and after their search sessions, quantifying their knowledge gain. We investigated the impact of information needs on the search behavior and knowledge gain of users, revealing a significant effect of information need on user queries and navigational patterns, but no direct effect on the knowledge gain. Users on average exhibited a higher knowledge gain through search sessions pertaining to topics they were less familiar with.

Web search is frequently used by people to acquire new knowledge and to satisfy learning-related objectives. In this context, informational search missions with an intention to obtain knowledge pertaining to a topic are prominent. The importance of learning as an outcome of web search has been recognized. Yet, there is a lack of understanding of the impact of web search on a user's knowledge state. Predicting the knowledge gain of users can be an important step forward if web search engines that are currently optimized for relevance can be molded to serve learning outcomes. In this paper, we introduce a supervised model to predict a user's knowledge state and knowledge gain from features captured during the search sessions. To measure and predict the knowledge gain of users in informational search sessions, we recruited 468 distinct users using crowdsourcing and orchestrated real-world search sessions spanning 11 different topics and information needs. By using scientifically formulated knowledge tests, we calibrated the knowledge of users before and after their search sessions, quantifying their knowledge gain. Our supervised models utilise and derive a comprehensive set of features from the current state of the art and compare performance of a range of feature sets and feature selection strategies. Through our results, we demonstrate the ability to predict and classify the knowledge state and gain using features obtained during search sessions, exhibiting superior performance to an existing baseline in the knowledge state prediction task.

Embedded markup based on Microdata, RDFa, and Microformats have become prevalent on the Web and constitute an unprecedented source of data. However, RDF statements extracted from markup are fundamentally different from traditional RDF graphs: entity descriptions are flat, facts are highly redundant, and despite very frequent co-references explicit links are missing. Therefore, carrying out typical entity-centric tasks such as retrieval and summarisation cannot be tackled sufficiently with state-of-the-art methods and require preliminary data fusion. Given the scale and dynamics of Web markup, the applicability of general data fusion approaches is limited. We present a novel query-centric data fusion approach which overcomes such issues through a combination of entity retrieval and fusion techniques geared towards the specific challenges associated with embedded markup. To ensure precise and diverse entity descriptions, we follow a supervised learning approach and train a classifier for data fusion of a pool of candidate facts relevant to a given query and obtained through a preliminary entity retrieval step. We perform a thorough evaluation on a subset of the Web Data Commons dataset and show significant improvement over existing baselines. In addition, an investigation into the coverage and complementarity of facts from the constructed entity descriptions compared to DBpedia, shows potential for aiding tasks such as knowledge base population.

Crowdsourcing has become an increasingly popular means to acquire human input on demand. Microtask crowdsourcing market-places facilitate the access to millions of people (called workers) who are willing to participate in tasks in return for monetary rewards or other forms of compensation. This paradigm presents a unique learning context where workers have to learn to complete tasks on-the-fly by applying their learning immediately through the course of tasks. However, most workers typically dropout early in large batches of tasks, depriving themselves of the opportunity to learn on-the-fly through the course of batch completion. By doing so workers squander a potential chance at improving their performance and completing tasks effectively. In this paper, we propose a novel method to engage and retain workers, to improve their learning in crowdsourced information finding tasks by using achievement priming. Through rigorous experimental findings, we show that it is possible to retain workers in long batches of tasks by triggering their inherent motivation to achieve and excel. As a consequence of increased worker retention, we find that workers learn to perform more effectively, depicting relatively more stable accuracy and lower task completion times in comparison to workers who drop out early.

Paid crowdsourcing platforms have evolved into remarkable marketplaces where requesters can tap into human intelligence to serve a multitude of purposes, and the workforce can benefit through monetary returns for investing their efforts. In this work, we focus on individual crowd worker competencies. By drawing from self-assessment theories in psychology, we show that crowd workers often lack awareness about their true level of competence. Due to this, although workers intend to maintain a high reputation, they tend to participate in tasks that are beyond their competence. We reveal the diversity of individual worker competencies, and make a case for competence-based pre-selection in crowdsourcing marketplaces. We show the implications of flawed self-assessments on real-world microtasks, and propose a novel worker pre-selection method that considers accuracy of worker self-assessments. We evaluated our method in a sentiment analysis task and observed an improvement in the accuracy by over 15%, when compared to traditional performance-based worker pre-selection. Similarly, our proposed method resulted in an improvement in accuracy of nearly 6% in an image validation task. Our results show that requesters in crowdsourcing platforms can benefit by considering worker self-assessments in addition to their performance for pre-selection.

Manually inspecting text in a document collection to assess whether an event occurs in it is a cumbersome task. Although a manual inspection can allow one to identify and discard false events, it becomes infeasible with increasing numbers of automatically detected events. In this paper, we present a system to automatize event validation, defined as the task of determining whether a given event occurs in a given document or corpus. In addition to supporting users seeking for information that corroborates a given event, event validation can also boost the precision of automatically detected event sets by discarding false events and preserving the true ones. The system allows to specify events, retrieves candidate web documents, and assesses whether events occur in them. The validation results are shown to the user, who can revise the decision of the system. The validation method relies on a supervised model to predict the occurrence of events in a non-annotated corpus. This system can also be used to build ground-truths for event corpora.

Crowdsourcing is increasingly being used as a means to tackle problems requiring human intelligence. With the ever-growing worker base that aims to complete microtasks on crowdsourcing platforms in exchange for financial gains, there is a need for stringent mechanisms to prevent exploitation of deployed tasks. Quality control mechanisms need to accommodate a diverse pool of workers, exhibiting a wide range of behavior. A pivotal step towards fraud-proof task design is understanding the behavioral patterns of microtask workers. In this paper, we analyze the prevalent malicious activity on crowdsourcing platforms and study the behavior exhibited by trustworthy and untrustworthy workers, particularly on crowdsourced surveys. Based on our analysis of the typical malicious activity, we define and identify different types of workers in the crowd, propose a method to measure malicious activity, and finally present guidelines for the efficient design of crowdsourced surveys.

Manually inspecting text to assess whether an event occurs in a document collection is an onerous and time consuming task. Although a manual inspection to discard the false events would increase the precision of automatically detected sets of events, it is not a scalable approach. In this paper, we automatize event validation, defined as the task of determining whether a given event occurs in a given document or corpus. The introduction of automatic event validation as a post-processing step of event detection can boost the precision of the detected event set, discarding false events and preserving the true ones. We propose a novel automatic method for event validation, which relies on a supervised model to predict the occurrence of events in a non-annotated corpus. The data for training the model is gathered by exploiting the crowdsourcing paradigm. Experiments on real-world events and documents show that our proposed method (i) outperforms the state-of-the-art event validation approach and (ii) increases the precision of event detection while preserving recall.

Long-running, high-impact events such as the Boston Marathon bombing often develop through many stages and involve a large number of entities in their unfolding. Timeline summarization of an event by key sentences eases story digestion, but does not distinguish between what a user remembers and what she might want to re-check. In this work, we present a novel approach for timeline summarization of high-impact events, which uses entities instead of sentences for summarizing the event at each individual point in time. Such entity summaries can serve as both (1) important memory cues in a retrospective event consideration and (2) pointers for personalized event exploration. In order to automatically create such summaries, it is crucial to identify the "right" entities for inclusion. We propose to learn a ranking function for entities, with a dynamically adapted trade-off between the in-document salience of entities and the informativeness of entities across documents, i.e., the level of new information associated with an entity for a time point under consideration. Furthermore, for capturing collective attention for an entity we use an innovative soft labeling approach based on Wikipedia. Our experiments on a real large news datasets confirm the effectiveness of the proposed methods.

With the advent and growing use of crowdsourcing labor markets for a variety of applications, optimizing the quality of results produced is of prime importance. The quality of the results produced is typically a function of the performance of crowd workers. In this paper, we investigate the notion of treating crowd workers as ‘ learners ’ in a novel learning environment. This learning context is characterized by a short-lived learning phase and immediate application of learned concepts. We draw motivation from the desire of crowd workers to perform well in order to maintain a good reputation, while attaining monetary rewards successfully. Thus, we delve into training workers in specific microtasks of different types. We exploit (i) implicit training , where workers are provided training when they provide erraneous responses to questions with priorly known answers, and (ii) explicit training , where workers are required to go through a training phase before they attempt to work on the task itself. We evaluated our approach in 4 different types of microtasks with a total of 1200 workers, who were subjected to either one of the proposed training strategies or baseline case of no training . The results show that workers who undergo training depict an improvement in performance upto 5 %, and a reduction in the task completion time upto 41 %. Additionally, crowd training led to the elimination of malicious workers and a costs-benefit gain upto nearly 15 %.

The Web of data is constantly evolving based on the dynamics of its content. Current Web search engine technologies consider static collections and do not factor in explicitly or implicitly available temporal information, that can be leveraged to gain insights into the dynamics of the data. In this paper, we hypothesize that by employing the temporal aspect as the primary means for capturing the evolution of entities, it is possible to provide entity-based accessibility to Web archives. We empirically show that the edit activity on Wikipedia can be exploited to provide evidence of the evolution of Wikipedia pages over time, both in terms of their content and in terms of their temporally defined relationships, classified in literature as events. Finally, we present results from our extensive analysis of a dataset consisting of 31,998 Wikipedia pages describing politicians, and observations from in-depth case studies. Our findings reflect the usefulness of leveraging temporal information in order to study the evolution of entities and breed promising grounds for further research.

With more than 340~million messages that are posted on Twitter every day, the amount of duplicate content as well as the demand for appropriate duplicate detection mechanisms is increasing tremendously. Yet there exists little research that aims at detecting near-duplicate content on microblogging platforms. We investigate the problem of near-duplicate detection on Twitter and introduce a framework that analyzes the tweets by comparing (i) syntactical characteristics, (ii) semantic similarity, and (iii) contextual information. Our framework provides different duplicate detection strategies that, among others, make use of external Web resources which are referenced from microposts. Machine learning is exploited in order to learn patterns that help identifying duplicate content. We put our duplicate detection framework into practice by integrating it into Twinder, a search engine for Twitter streams. An in-depth analysis shows that it allows Twinder to diversify search results and improve the quality of Twitter search. We conduct extensive experiments in which we (1) evaluate the quality of different strategies for detecting duplicates, (2) analyze the impact of various features on duplicate detection, (3) investigate the quality of strategies that classify to what exact level two microposts can be considered as duplicates and (4) optimize the process of identifying duplicate content on Twitter. Our results prove that semantic features which are extracted by our framework can boost the performance of detecting duplicates.