July 29, 2025

Simulated Selfhood in LLMs: A Behavioral Analysis of Introspective Coherence

View Full Paper

Ask AI

Bookmark

The study analyzes large language models for signs of introspection through 21 behavioral prompts, generating 1,050 completions.
High rates of contradictions were found in self-referential outputs, reflecting a tension between mechanistic and anthropomorphic language.
Behavioral dimensions evaluated include surface-level similarity and inferential consistency, revealing fluctuations in self-representation.
Findings emphasize caution in interpreting LLM outputs as reflective of mental states, informing LLM interpretability and user perception.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Behavioral analysis reveals that LLMs display pseudo-consciousness but lack true introspective coherence, indicating limitations in self-representation.

Cite This Study

Jose Augusto de Lima Prestes (2025) studied this question.

synapsesocial.com/papers/689a0945e6551bb0af8ced74 https://doi.org/10.31235/osf.io/u3qc9_v3

View Full Paper

Ask AI

Bookmark

Implication

Behavioral analysis reveals that LLMs display pseudo-consciousness but lack true introspective coherence, indicating limitations in self-representation.

Simulated Selfhood in LLMs: A Behavioral Analysis of Introspective Coherence

Key Points

Discussion

Implication

Cite This Study

Simulated Selfhood in LLMs: A Behavioral Analysis of Introspective Coherence

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider