Designing and Understanding Risk Communication in Human-AI Interaction: Perceptual Effects and Implications
Year: 2026
Type: dissertation
Abstract: AI systems increasingly inform and autonomously make decisions that carry real-world consequence, from medical guidance to autonomous physical assistance.These systems are non-deterministic and introduce uncertainty that users must infer and navigate, often by relying on cues embedded in the interaction itself that serve to form judgements about the 'riskiness' of a system and how much it can be trusted.Risk, defined as a function of uncertainty and consequence, is therefore not simply a property to be calculated but a message that must be communicated and perceived within the interaction.The design of these risk communication cues can shape whether users fail to register a risk, underestimate it, or develop appropriate risk perceptions.This thesis empirically explores the design of risk communication across varying AI systems and its effects on user perception.Drawing on four articles spanning empirical work across online experiments, a design workshop, and laboratory studies, we examine how communicative affordances, and the design mechanisms deployed within them, can encode risk information, and how these shape risk perception and system trustworthiness.
Author: Schömbs, Sarah Victoria
Institution: The University of Melbourne
Language: English
Subfield: Human Factors and Ergonomics
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7224977470
