When AI Deceives: A Natural Experiment on the Causal Effects of Perceived Deception on Player Ratings in RPGs

This feed aggregates recent CS/cyber papers on human-AI interaction, alignment, and tools for verification. Key findings: AI-driven deception in games harms player ratings when players perceive deception; AI persuasive framing can increase cooperation but antisocial framing produces larger, longer-lasting harms (dual-use risk); interpretable episode-based LLM reasoning traces (Epi2Diff) improve prediction of human item difficulty; Google’s Paper Assistant Tool (PAT) demonstrates scalable, agentic AI for deep scientific review that finds substantive errors; studies highlight psychometric limits