Work

What shapes perceived safety in autonomous mobility? Cross-cultural evidence from large-scale online eye tracking

Year: 2026

Type: article

Abstract: As autonomous vehicles increasingly perceive and navigate the world through AI-based sensing systems, understanding what people pay attention to and how they interpret these same environments has become equally important for socially accepted autonomous mobility. Public acceptance depends not only on how safely autonomous systems operate, but also on how people interpret the environments in which these systems act. While increasing the visible representation of diverse social groups is widely assumed to promote inclusion, safety, and trust, empirical evidence remains limited. We address this question using a novel combination of large-scale online eye-tracking ( n = 1272; ~ 4000 AOI-level gaze indicators) and survey data collected across 22 cities on 5 continents. We examine how people visually attend to social diversity in autonomous mobility environments and how these attention patterns relate to perceived safety and security. The findings reveal that gaze patterns towards social groups were structured and remarkably stable across AV and non-AV mobility scenarios. Yet this stable pattern of visual attention was only weakly associated with perceived safety and security. Instead, perceptions of safety and security were more strongly associated with trust in automation, minority status, and broader socio-cultural context. Our findings therefore show that visual attention and safety evaluations represent distinct dimensions of public responses to AI-mediated mobility, challenging the assumption that visible representation alone is sufficient to foster inclusion, trust, and safety. The study demonstrates the value of eye tracking for AV ethics, provides the first cross-cultural evidence supporting the distinction between functional safety and perceived safety in autonomous mobility. It suggests that socially legitimate autonomous mobility depends not only on technical performance, but also on the socio-psychological and institutional conditions through which trust and safety are constructed.

Source: AI & Society

DOI: 10.1007/s00146-026-03278-1

Authors: Anu Masso, Ahti-Veikko Pietarinen, Mergime Ibrahimi, Sadok Ben Yahia

Institutions: Tallinn University of Technology, Hong Kong Baptist University, University of Southern Denmark

Language: English

Topic: Human-Automation Interaction and Safety

Subfield: Social Psychology

Field: Psychology

Domain: Social Sciences

SDGs: Sustainable cities and communities, Reduced inequalities

Open Access status: hybrid

Estimated APC paid: $3,190