Researching AI Legibility through Design

As part of our work on the Uncanny AI project (funded by the PETRAS Centre for Excellence for Cybersecurity) we recently published a significant paper titled Researching AI Legibility through Design, which explores how to make AI more legible and acceptable to users. In this work we use speculative icon designs to unpack the complex issues around privacy, ethics, and trust which arise from applications of AI and the data which make them work.

Our prototype designs have been produced as part of a deck of cards. By playing a simple game, which involves matching the icons to their descriptions we have been running workshops with various stakeholders (including the fabulous MA students in Digital Management at Hyperisland, pictured) to develop the next iteration of the designs.

This work utilises Design Research methods to combine and build upon research from multiple disciplines and aims to drive forward practical and usable approaches to making AI more legible for users and exploring the Implications for Adoption of emerging technologies.