102
Studies
Evaluation and psychometric validity
Instruments and benchmarks examining what is measured, how reliably, and whether inferred or declared traits are stable and valid.
Explore trackCritical review · literature cutoff July 18, 2026
Critical evidence on synthetic personality in LLMs
An academic map of 385 unique studies, organized into 6 research tracks. All 385 records were written after reading the full source and distinguish method, results, limitations, and what each study does not establish.
Research tracks
65
Studies
Trait induction and control
Prompting, fine-tuning, and editing activations or weights to elicit, represent, modulate, or suppress traits and behaviors.
Explore track68
Studies
Personas, identity, and agents
Design and evaluation of characters, profiles, and persistent agents, including role-play, personalization, memory, identity, and consistency.
Explore track45
Studies
Society, culture, and collective behavior
Use of LLMs to study or approximate human group opinions, values, and decisions, as well as multi-agent interaction, networks, and collective phenomena.
Explore track88
Studies
Applications, bias, and safety
Applications in health, education, recommendation, finance, and design, together with risks involving bias, discrimination, privacy, manipulation, and harm.
Explore track17
Studies
Reviews, theory, and governance
Field reviews, conceptual frameworks, methodological requirements, and proposals for responsible research and governance.
Explore trackLatest review added
Ten Novel Phenomena in Machine Psychology: How Large Language Models Exhibit Complex Identity-Reactive Behaviors in Response to Ethnically-Cued User Names
A 3x3x5x3 design with ChatGPT-5.4, Claude 4.6 Sonnet, and Gemini 3.1 Pro; three names, five domains, and three registers produce 135 responses through public interfaces. Grounded theory generates ten phenomena; a second coder reviews the corpus. A 136-feature pipeline applies ANOVA, chi-square, regression, a 1,000-bootstrap, K-means, and HDBSCAN.
Recently added studies
View allTrack: Society, culture, and collective behavior
12 Angry AI Agents: Evaluating Multi-Agent LLM Decision-Making Through Cinematic Jury Deliberation
Track: Evaluation and psychometric validity
Verbal Confidence Saturation in 3-9B Open-Weight Instruction-Tuned LLMs: A Pre-Registered Psychometric Validity Screen
Track: Trait induction and control
The Governance of Human-LLM Interaction: Safety Gating, Civility Steering, and Affective Default Lock-In
Track: Reviews, theory, and governance
Understanding large language models demands distinguishing human projection from machine cognition
Track: Applications, bias, and safety