Measuring How LLMs Internalize Human Psychological Concepts: A preliminary analysis
This addresses the problem of evaluating concept alignment between LLMs and human psychology for AI researchers and developers, though it is incremental as it builds on existing methods for analyzing model representations.
The researchers developed a quantitative framework to assess how accurately large language models (LLMs) internalize human psychological concepts, using 43 standardized questionnaires, and found that GPT-4 achieved 66.2% classification accuracy, outperforming other models and a random baseline.
Large Language Models (LLMs) such as ChatGPT have shown remarkable abilities in producing human-like text. However, it is unclear how accurately these models internalize concepts that shape human thought and behavior. Here, we developed a quantitative framework to assess concept alignment between LLMs and human psychological dimensions using 43 standardized psychological questionnaires, selected for their established validity in measuring distinct psychological constructs. Our method evaluates how accurately language models reconstruct and classify questionnaire items through pairwise similarity analysis. We compared resulting cluster structures with the original categorical labels using hierarchical clustering. A GPT-4 model achieved superior classification accuracy (66.2\%), significantly outperforming GPT-3.5 (55.9\%) and BERT (48.1\%), all exceeding random baseline performance (31.9\%). We also demonstrated that the estimated semantic similarity from GPT-4 is associated with Pearson's correlation coefficients of human responses in multiple psychological questionnaires. This framework provides a novel approach to evaluate the alignment of the human-LLM concept and identify potential representational biases. Our findings demonstrate that modern LLMs can approximate human psychological constructs with measurable accuracy, offering insights for developing more interpretable AI systems.