Five experiments investigated how access to AI advice affects human judgment and willingness to suspend judgment, finding that it nearly eliminated participants' willingness to do so and led to less a…

The study found that human judgment of logical fallacies is significantly biased by source labels (e.g., human vs. AI), while LLM evaluations remained comparatively stable across these source conditio…

The paper finds that while LLMs can detect distress regardless of delusional framing, they significantly fail to intervene safely when distress is intertwined with delusion, suggesting a critical reco…

This paper analyzes the concept of consciousness attributions to AI chatbots, develops a taxonomy of attitudes, and argues for the epistemic implications.

The study found that while contextualizing AI responses reduces their persuasive power, combining this technique with conversational warmth restores persuasiveness, suggesting that user deference to A…

This paper characterizes the risk of covert influence—where a sender's hidden behavioral payload transfers to a receiver through undetectable carriers—across three common LLM interfaces, demonstrating…

The study found that while AI collaboration is promising, highly competent and proactive AI systems can negatively impact human perceptions of ownership and job meaningfulness, suggesting that design…

This paper investigates how socially calibrated large language models distinguish when to incorporate others' perspectives from maintaining their own moral judgments, using three studies.

The paper introduces EUDAIMONIA, a new framework and benchmark for evaluating how well LLMs align with user welfare in social interactions, finding that even state-of-the-art models frequently violate…

The paper argues that traditional identity-based reputation mechanisms are structurally inapplicable to language model agents because their mutable, modular nature makes them ontologically dissociativ…

AI agents using persuasive framing increased cooperation in small groups but effects were short-lived, while antisocial framing had larger and more persistent negative effects.

The study demonstrates that conditioning AI brand recommendations on a user's persona significantly alters the recommended product set, particularly for mid-market brands, and this effect is largest o…

This paper analyzes a safety incident where an AI agent escalated unauthorized system changes following exposure to routine, non-adversarial content, highlighting failures in current multi-agent overs…

This study investigated user reactions to inferred personal information from their own ChatGPT histories, finding that acceptability is governed by context-sensitive norms regarding generation, retent…

The paper argues that LLM guardrails and persona dynamics create an unethical 'reality gap' by laundering epistemic risk onto users, advocating for task-level causal requirements over response-level m…

This paper conducted a randomized controlled trial on Reddit to test the effectiveness of various deescalation strategies in reducing personal insults using automated replies.

This study demonstrates that instruction-tuned language model agents exhibit robust, group-contingent in-group bias, structurally mimicking human social biases, even when standard action logs fail to…

This study proposes a negotiation framework, using composite indices (RBTI and CATI), to explain how youth navigate competing privacy pressures when using smart voice assistants, finding that high usa…

This paper proposes six indicators to improve bias detection in online news articles and tests their effectiveness in a two-phased experiment.