Algorithmic Reflectivity in Large Language Models: Methodological Study of Linguistic Structure in Mentalization
The study tested if a Large Language Model could generate mentalization profiles matching clinical standards for Mentalization-Based Treatment (MBT). The model produced linguistically coherent profiles, scoring well on some axes, but failed to integrate internal states with external contexts dynamically. This shows the model can mimic the language of reflection without possessing genuine feeling or intentional experience.