TACSIA: A Framework for Passenger Expectations on Trust in Autonomous Cars with Socially Interactive Agents

This is the authors' abstract. We don't add key points for this paper.

With Agentic AI redefining the driving experience, trust has evolved from a static metric into a dynamic outcome of the human-vehicle partnership. While autonomous co-pilots are increasingly integrated into digital cockpits, the specific characteristics required to build and maintain trust during Level 4 "Mind Off" scenarios remain underexplored, particularly the underlying mental models that passengers use to define a trustworthy AI partner. This paper employs an exploratory, bottom-up qualitative study with a representative sample (N = 22) to systematically assess passenger expectations towards Agentic AI Assistants. Utilizing a multi-coder Grounded Theory approach, a comprehensive design framework encompassing user expectations regarding trust-inducing competencies, communication styles and embodiment characteristics is derived. This framework articulates the design space for AI Assistants in Digital Cockpits, providing actionable insights and design heuristics for the development of trust-calibrated partnerships.