Assessment and manipulation of latent constructs in pre-trained language models using psychometric scales
This work provides a novel method for interpreting and rectifying biases in language models, which is crucial for developing more explainable and trustworthy AI systems.
This paper addresses the challenge of assessing latent psychological constructs in pre-trained language models by reformulating standard psychological questionnaires into natural language inference prompts. Applying this method to 88 publicly available models, the authors found human-like mental health-related constructs such as anxiety and depression, which align with human psychological theories and respond to similar mitigation strategies.
Human-like personality traits have recently been discovered in large language models, raising the hypothesis that their (known and as yet undiscovered) biases conform with human latent psychological constructs. While large conversational models may be tricked into answering psychometric questionnaires, the latent psychological constructs of thousands of simpler transformers, trained for other tasks, cannot be assessed because appropriate psychometric methods are currently lacking. Here, we show how standard psychological questionnaires can be reformulated into natural language inference prompts, and we provide a code library to support the psychometric assessment of arbitrary models. We demonstrate, using a sample of 88 publicly available models, the existence of human-like mental health-related constructs (including anxiety, depression, and Sense of Coherence) which conform with standard theories in human psychology and show similar correlations and mitigation strategies. The ability to interpret and rectify the performance of language models by using psychological tools can boost the development of more explainable, controllable, and trustworthy models.