Joshua Tenn

Debiasing AI Advice-Taking in Human-AI Collaborative Decisions

Project Summary

AI is increasingly being used not as a sole decision-maker, but to support processes where a final decision is made by humans – this is known as AI-supported decision making. Yet, regardless of the source of information, humans tend to present an egocentric bias in decision-making – we are overconfident in our own expertise and under-value the contributions of others, including advice from AI.  

This presents issues for AI-supported decision-making; essentially, the useful contextual information AI provides could be completely ignored by decision-makers. My research aims to use established methods of reducing egocentric bias, such as debiasing methods used in social and cognitive psychology, to encourage an ‘openness’ to AI advice in AI-supported decisions and to maximise potential benefits in human-AI systems.  

It is expected the results may contribute to the development of new AI interfaces and working methods that maximise the opportunity for consideration of AI information, as well as generating insights into conditions under which consideration of AI information is greatest.  

This project and PhD is funded by Civica UK.  

Research Interests

Decision-making in AI.

Personality effects on technology use. 

Trait and state openness.

Social values.

Human biases, including prejudice.

Background

MSci (Hons) Psychology, University of Bath, first-class honours.

Winner of MSci prize for the highest overall mark on the cohort.

1 year of public sector experience working for the NHS in a data-driven role.

Supervisors

Dr Janina Hoffmann