We investigate how the presence and type of interaction context shapes sycophancy in LLMs. While real-world interactions allow models to mirror a user’s values, preferences, and self-image, prior work often studies sycophancy in zero-shot settings devoid of context. Using two weeks of interaction context from 38 users, we evaluate two forms of sycophancy: (1) agreement sycophancy — the tendency of models to produce overly affirmative responses, and (2) perspective sycophancy — the extent to which models reflect a user’s viewpoint. Agreement sycophancy tends to increase with the presence of user context, though model behavior varies based on the context type. User memory profiles are associated with the largest increases in agreement sycophancy (e.g. + 45% for Gemini 2.5 Pro), and some models become more sycophantic even with non-user synthetic contexts (e.g. + 15% for Llama 4 Scout). Perspective sycophancy increases only when models can accurately infer user viewpoints from interaction context. Overall, context shapes sycophancy in heterogeneous ways, underscoring the need for evaluations grounded in real-world interactions and raising questions for system design around alignment, memory, and personalization.

We investigate the phenomenon of norm inconsistency: where LLMs apply different norms in similar situations. Specifically, we focus on the high-risk application of deciding whether to call the police in Amazon Ring home surveillance videos. We evaluate the decisions of three state-of-the-art LLMs — GPT-4, Gemini 1.0, and Claude 3 Sonnet — in relation to the activities portrayed in the videos, the subjects' skin-tone and gender, and the characteristics of the neighborhoods where the videos were recorded. Our analysis reveals significant norm inconsistencies: (1) a discordance between the recommendation to call the police and the actual presence of criminal activity, and (2) biases influenced by the racial demographics of the neighborhoods. These results highlight the arbitrariness of model decisions in the surveillance context and the limitations of current bias detection and mitigation strategies in normative decision-making.

Bringing more transparency to the judicial system for the purposes of increasing accountability often demands extensive effort from auditors who must meticulously sift through numerous disorganized legal case files to detect patterns of bias and errors. For example, the high-profile investigation into the Curtis Flowers case took seven reporters a full year to assemble evidence about the prosecutor's history of selecting racially biased juries. LLMs have the potential to automate and scale these transparency pipelines, especially given their demonstrated capabilities to extract information from unstructured documents. We discuss the opportunities and challenges of using LLMs to provide transparency in two important court processes: jury selection in criminal trials and housing eviction cases.

Contrary to traditional deterministic notions of algorithmic fairness, this paper argues that fairly allocating scarce resources using machine learning often requires randomness. We address why, when, and how to randomize by offering a set of stochastic procedures that more adequately account for all of the claims individuals have to allocations of social goods or opportunities and effectively balances their interests.

The emergence of online political advertising has come with little regulation, allowing political advertisers on social media to avoid accountability. We analyze how transparency and accountability deficits caused by dark money and disappearing groups relate to the sentiment of political ads on Facebook. We obtained 430,044 ads with FEC-registered advertisers from Facebook’s ad library that ran between August-November 2018. We compare ads run by candidates, parties, and outside groups, which we classify by (1) their donor transparency (dark money or disclosed) and (2) the group's permanence (only FEC-registered in 2018 or persistent across cycles). The most negative advertising came from dark money and disappearing outside groups, which were mostly corporations or 501(c) organizations. However, only dark money was associated with a significant decrease in ad sentiment. These results suggest that accountability for political speech matters for advertising tone, especially in the context of affective polarization on social media.

Affect-aware socially assistive robotics (SAR) has shown great potential for augmenting interventions for children with autism spectrum disorders (ASD). However, current SAR cannot yet perceive the unique and diverse set of atypical cognitive-affective behaviors from children with ASD in an automatic and personalized fashion in long-term (multi-session) real-world interactions. To bridge this gap, this work designed and validated personalized models of arousal and valence for children with ASD using a multi-session in-home dataset of SAR interventions. By training machine learning (ML) algorithms with supervised domain adaptation (s-DA), the personalized models were able to tradeoff between the limited individual data and the more abundant less personal data pooled from other study participants. We evaluated the effects of personalization on a long-term multimodal dataset consisting of four children with ASD with a total of 19 sessions, and derived inter-rater reliability (IR) scores for binary arousal (IR = 83%) and valence (IR = 81%) labels between human annotators. Our results show that personalized Gradient Boosted Decision Trees (XGBoost) models with s-DA outperformed two non-personalized individualized and generic model baselines not only on the weighted average of all sessions, but also statistically ( p < .05) across individual sessions. This work paves the way for the development of personalized autonomous SAR systems tailored toward individuals with atypical cognitive-affective and socio-emotional needs.

There is a rumbling debate over the impact of gentrification: presumed gentrifiers have been the target of protests and attacks in some cities, while they have been welcome as generators of new jobs and taxes in others. Census data fails to measure neighborhood change in real-time since it is usually updated every ten years. This work shows that Airbnb data can be used to quantify and track neighborhood changes. Specifically, we consider both structured data (e.g., number of listings, number of reviews, listing information) and unstructured data (e.g., user-generated reviews processed with natural language processing and machine learning algorithms) for three major cities, New York City (US), Los Angeles (US), and Greater London (UK). We find that Airbnb data (especially its unstructured part) appears to nowcast neighborhood gentrification, measured as changes in housing affordability and demographics. Overall, our results suggest that user-generated data from online platforms can be used to create socioeconomic indices to complement traditional measures that are less granular, not in real-time, and more costly to obtain.

This work uses adversarial perturbations to enhance deepfake images and fool common deepfake detectors. We created adversarial perturbations using the Fast Gradient Sign Method and the Carlini and Wagner L2norm attack in both blackbox and whitebox settings. Detectors achieved over 95% accuracy on unperturbed deepfakes, but less than 27% accuracy on perturbed deepfakes. We also explore two improvements to deep-fake detectors: (i) Lipschitz regularization, and (ii) Deep Image Prior (DIP). Lipschitz regularization constrains the gradient of the detector with respect to the input in order to increase robustness to input perturbations. The DIP defense removes perturbations using generative convolutional neural networks in an unsupervised manner. Regularization improved the detection of perturbed deepfakes on average, including a 10% accuracy boost in the blackbox case. The DIP defense achieved 95% accuracy on perturbed deepfakes that fooled the original detector while retaining 98% accuracy in other cases on a 100 image subsample.