Open Internet by MindsNet

Validating LLM Responses to Projective Psychological Tests

The study's methodology is compromised due to likely data contamination and lack of controls, making it difficult to determine the scientific value of administering standard Rorschach tests to LLMs. This raises concerns about the validity of using LLMs in psychological research without proper safeguards. The challenge lies in developing methods to rule out data contamination and ensure the tests measure meaningful aspects of AI processing.

Computing & Technology, Computer Science, Machine Learning