Calibrating user trust in AI products
Manú Bartlett & Ewa Koc at UX Brighton 2024
About this talk
In this session Ewa (UX Designer) & Manú (UX Researcher) will talk us through a variety of case studies from AI powered products which have successfully built calibrated trust in their users.
Taking their own experience building an AI search tool for seasoned experts in patent intelligence as the starting point.
Exploring three key elements for building user trust:
How to explain how your product works
- It’s not about how an AI makes a decision, it’s why
- Case study: Derwent Innovation’s AI search
How to ‘humanise’ your AI product to build trust
- Case study: ChatGPT
- How to harness good bias to personalise your AI: Feedback, training on the fly & data sources
How to teach users to use the Al product correctly
- Responding to actions
- Error messages
- Interactive onboarding
About Manú & Ewa
Manú Bartlett
Manú has worked as a UX Researcher for over 6 years, starting off with microfinance users for a fintech inclusion organisation in Mexico in 2017, then working as an in-house researcher for the British Government, followed by Opencast UK Tech Consultancy and last year she joined the global leader in data products: Clarivate Analytics.
She is an Anthropologist who won the prestigious Fulbright scholarship to study for a masters degree in Design & Anthropology at Parsons Design School, the New School in NYC.
She also holds a BA in Economic Geography from the London School of Economics (LSE).
Her work unites the iterative process of design thinking with in-depth ethnographic & quantitative research to understand the needs and desires of users in complex contexts.
Ewa Koc
Ewa is a senior UX designer with a background in linguistics. Formerly at Roche Diagnostics, a global healthcare leader, she now works at Clarivate, focusing on IP Intelligence Solutions.
She completed her master’s degree in UX/UI at Ramon Llull University of Barcelona and has previous experience in marketing and technical writing.
Ewa is particularly interested in accessibility and data-driven design, enabling her to make a meaningful impact in the digital world.
Related reading
Books, research, ideas and people referenced in this talk.
Ad · Book titles link to Amazon. As an Amazon Associate I earn from qualifying purchases.
Projects and resources
- Derwent Innovation00:31Derwent Innovation is the patent search tool that the speakers used as their case study for building trust in AI.
- 2024 Edelman Trust Barometer03:23The speakers use the 2024 Edelman Trust Barometer to compare how much different populations trust AI.
- Duolingo Max11:05Duolingo's changing bird icon is used as an example of human-friendly language and reinforcement around an AI product.
- Magic Studio12:11Canva's personalised onboarding is used as an example of teaching users how to use an AI product.
Papers and articles
- Humans and Automation: Use, Misuse, Disuse, Abuse04:50Parasuraman and Riley's 1997 paper is cited for the relationship between a user's expertise and trust in an automated system.
- Calibrating Trust in Human-AI Collaboration05:35This paper provides a research reference for the calibrated-trust concept, where trust matches an AI system's actual capability and risk.
Concepts
- Human-in-the-loop AI10:21Human-in-the-loop is the approach the speakers describe for letting people provide feedback when an AI prediction is uncertain.
