Ch-AI-Tanya: A Research Repository for Exploring Psychological Phenomena of Large Language Models

Ch-AI-Tanya is a research repository focused on the psychology-like phenomena of large language models (LLMs), aiming to collect, organize, and analyze the psychology-like behavioral characteristics observed in LLMs. From an interdisciplinary perspective, this project explores analogies between human psychological phenomena (such as cognitive biases and memory patterns) and those exhibited by AI systems in interactions, providing references for researchers, developers, and ethicists.

Ch-AI-Tanya is corely positioned as a research repository rather than an application tool, systematically collecting and classifying LLM-related psychological phenomena. Its content structure includes: phenomenon definitions, observation cases, mechanism explanations, relevant papers, and discussion controversies, facilitating the systematic presentation of interdisciplinary knowledge.

The project covers four categories of psychology-related phenomena:

1. Cognitive bias category: confirmation bias, anchoring effect, framing effect, and other reasoning biases;
2. Memory and forgetting category: context attention allocation, early information forgetting curve, confabulated memory;
3. Social psychology category: conformity in role-playing, authority obedience, group thinking simulation;
4. Emotion and motivation category: emotion understanding simulation, multi-goal motivation conflict (involving affective computing and AI alignment).

Multi-level research methods are adopted:

1. Phenomenon observation: Identify repeated behavioral patterns through large-scale human-AI interaction logs, requiring systematic data collection and annotation;
2. Controlled experiments: Design specific prompt scenarios to test hypotheses (e.g., comparing the impact of quantitative biases in outputs under different frameworks);
3. Theoretical construction: Attempt to explain phenomena using existing psychological theories or new conceptual frameworks (challenging due to the fundamental differences between AI mechanisms and the human brain).

Positive values: Help predict and understand LLM behaviors, design effective interaction strategies, identify risk patterns, and provide conceptual tools for AI alignment; Controversies: Over-anthropomorphism may misinterpret AI capabilities, obscure statistical nature, be criticized as a category error, and easily mislead the public; The project needs to maintain methodological rigor and distinguish between descriptive analogies and essential assertions.

Ch-AI-Tanya represents the frontier of interdisciplinary AI research, exploring LLM behavioral characteristics. Regardless of whether the analogies are valid, they deepen the understanding of AI. For AI practitioners: Help understand the limitations of model capabilities; For psychology researchers: Provide an opportunity to observe mental phenomena on a new substrate; The repository provides a shared reference framework for interdisciplinary dialogue.