Home > Research > Publications & Outputs > Researching and Designing Uncanny AI to Legible AI

Electronic data

  • Researching_and_Designing_Uncanny_AI_to_Legible_AI_reduced

    Accepted author manuscript, 122 KB, PDF document

Links

View graph of relations

Researching and Designing Uncanny AI to Legible AI

Research output: Contribution to conference - Without ISBN/ISSN Abstractpeer-review

Published
Publication date13/11/2020
<mark>Original language</mark>English
EventInternational Transdisciplinary Conference - University of Dundee, Dundee, United Kingdom
Duration: 13/11/202015/11/2020

Conference

ConferenceInternational Transdisciplinary Conference
CountryUnited Kingdom
CityDundee
Period13/11/2015/11/20

Abstract

The sociotechnical landscape has become more reliant on artificial intelligence (AI) operating via the Internet of Things (IoT). Technology is typically designed to ease users’ interactions, consequently concealing functionality, leading to various misconceptions regarding AI and how data is processed (Pilling and Coulton, 2019). However, obscuring functions and the incognito collection of data is profitable for companies (Bryson and Winfield, 2017; Mosco, 2014). Though, this intrusion of technology is questioned by users, concerning the security,
privacy and the unknown implications of use (Bostrom, 2014; Bridle, 2018). Labelling AI as uncanny and illegible to users, either purposefully or as a result of AI’s unintelligible coded decision logic (Burrell, 2016). Using a Research through Design methodology, we attempt to make AI operations legible through iconography (Lindley et al., 2020; Pilling et al., 2020). We are currently gathering research through workshops, enabling us to test the rigour and disrupt the current iteration of iconography to provoke further iterations. The participants of the workshop complete game-like tasks using cards, which depict the icons and their associated text labels separately, as a medium to engage—in a tangible manner—the intangible and functions of these technologies. We have already had success in participants using with ease our initial iconography system, while also highlighting and exposing icons of interest; which are worth further research and discussion.