Exposing Tool Use in AI-Assistant Interfaces: Inspectable Evidence for Industrial Assistance

Open Access
Article
Conference Proceedings
Authors: Andrea Alesani
Abstract

AI assistants use tools to obtain information and perform actions, but users may still need to inspect the basis of an answer or action. This paper explores an interface for inspecting document retrieval and PDF navigation, with tool requests and results shown alongside a conversation and source PDF. Three author walkthroughs established concrete evidence conditions, including incomplete retrieval, in which inspection could help a user check an answer. Three moderated expert sessions on Meta Quest 3 then explored how participants responded to this support. When invited to check an installation answer, all pursued a source check; two identified omitted or additional details, while one could not verify the answer. Participants used different evidence routes and suggested clearer links, optional detail, and task-relevant visuals. The findings motivate further development of accessible evidence interfaces that support users in checking assistant outputs and deciding what to do next.

Keywords: Human-Computer Interaction, Tool Transparency, Artificial Intelligence, Interface Design, Industrial Assistance

DOI: 10.54941/ahfe1008406

Cite this paper
Downloads
0
Visits
1
Download PDF

More from this volume

← The Integration of Eye Tracking and Electrodermal Activity Measures for Assessing Cognitive Load and Mental Fatigue in Engineering and Computing: A Systematic Review (2010-2025)IMAGES-AI: Generative AI to improve products for the US Navy warfighter →
View all articles in Human Factors in Design, Engineering, and Computing →