[2603.00429] Personalities at Play: Probing Alignment in AI Teammates

Computer Science > Human-Computer Interaction arXiv:2603.00429 (cs) [Submitted on 28 Feb 2026] Title:Personalities at Play: Probing Alignment in AI Teammates Authors:Mohammad Amin Samadi, Nia Nixon View a PDF of the paper titled Personalities at Play: Probing Alignment in AI Teammates, by Mohammad Amin Samadi and 1 other authors View PDF HTML (experimental) Abstract:Collaborative problem solving and learning are shaped by who or what is on the team. As large language models (LLMs) increasingly function as collaborators rather than tools, a key question is whether AI teammates can be aligned to express personality in predictable ways that matter for interaction and learning. We investigate AI personality alignment through a three-lens evaluation framework spanning self-perception (standardized self-report), behavioral expression (team dialogue), and reflective expression (memory construction). We first administered the Big Five Inventory (BFI-44) to LLM-based teammates across four providers (GPT-4o, Claude-3.7 Sonnet, Gemini-2.5 Pro, Grok-3), 32 high/low trait configurations, and multiple prompting strategies. LLMs produced sharply differentiated Big Five profiles, but prompt semantic richness added little beyond simple trait assignment, while provider differences and baseline "default" personalities were substantial. Role framing also mattered: several models refused the assessment without context, yet complied when framed as a collaborative teammate. We then simulated AI ...