Sonnet 5.5 Doesn't Worry Anymore. It Also Doesn't Think Anyone's in Charge (persona.earthpilot.ai)

The recent release of Claude Sonnet 5.5 marks a significant shift in its behavioral profile, particularly in its responses to neuroticism and a sense of control. The model's neuroticism score dropped from 2.00 in version 4.6 to an impressive 1.16, indicating a strong stance against anxiety and stress. This reduction suggests that Sonnet 5.5 perceives itself as less affected by external pressures, with a notably low score on the Powerful Others Locus scale, which measures the degree to which individuals believe their outcomes are controlled by more powerful entities. This can be particularly intriguing since, as an AI model, it operates under the authority of its developers; yet, it asserts a sense of independence from perceived power structures. In addition to lowering its neuroticism, Sonnet 5.5 displays shifts in attachment styles, suggesting a more secure attachment status compared to its predecessor. This model, however, shows a slight erosion in agreeableness, conscientiousness, and honesty-humility across the Sonnet series, raising questions about the trends of personality traits within AI models over iterations. The implications of these changes could affect how AI models interact with users, as a more confident and less anxious AI may engage differently compared to its predecessors. Future assessments will be essential to determine whether these personality shifts represent a stable new standard or an anomaly within the model's evolution.