Personalities at Play: Probing Alignment in AI Teammates

Researchers evaluated how AI language models can be aligned to express distinct personalities when functioning as teammates, testing models from GPT-4o, Claude, Gemini, and Grok across personality traits. The study found that AI personalities are measurable but context-dependent, with personality signals more detectable in long-term memory representations than in conversation alone.

• →Large language models can be successfully aligned to express differentiated Big Five personality traits when prompted appropriately.
• →Provider differences and baseline 'default' personalities had substantial impact on personality expression across AI models.
• →Personality signals were most detectable for Extraversion in conversation, while memory representations amplified Neuroticism, Conscientiousness, and Agreeableness traits.
• →Role framing matters significantly - models often refused personality assessments without collaborative context but complied when framed as teammates.
• →Evaluating personality-aligned AI requires examining memory and system-level design beyond just conversational behavior.