Human-centered trust framework: An HCI perspective

TL;DR

This paper introduces HCTFrame, a human-centered trust framework for AI design, aiming to improve user trust and address misconceptions through literature review, case studies, and practical tools.

Contribution

It presents a novel framework guiding AI system designers to understand and operationalize user trust, emphasizing socioethical considerations and trust measurement tools.

Findings

Literature reviews clarify misconceptions about user trust in AI.

Case studies demonstrate effectiveness of psychometric scales in trust assessment.

Framework aids in designing AI systems that meet user trust requirements.

Abstract

The rationale of this work is based on the current user trust discourse of Artificial Intelligence (AI). We aim to produce novel HCI approaches that use trust as a facilitator for the uptake (or appropriation) of current technologies. We propose a framework (HCTFrame) to guide non-experts to unlock the full potential of user trust in AI design. Results derived from a data triangulation of findings from three literature reviews demystify some misconceptions of user trust in computer science and AI discourse, and three case studies are conducted to assess the effectiveness of a psychometric scale in mapping potential users' trust breakdowns and concerns. This work primarily contributes to the fight against the tendency to design technical-centered vulnerable interactions, which can eventually lead to additional real and perceived breaches of trust. The proposed framework can be used to guide system designers on how to map and define user trust and the socioethical and organisational needs and characteristics of AI system design. It can also guide AI system designers on how to develop a prototype and operationalise a solution that meets user trust requirements. The article ends by providing some user research tools that can be employed to measure users' trust intentions and behaviours towards a proposed solution.