Trust in Prediction Models: a Mixed-Methods Pilot Study on the Impact of Domain Expertise

Jeroen Ooge, Katrien Verbert

External link (DOI)
View presentation:2021-10-24T15:50:00ZGMT-0600Change your timezone on the schedule page
2021-10-24T15:50:00Z
Exemplar figure, but none was provided by the authors
Fast forward

Direct link to video on YouTube: https://youtu.be/UEcJPJrJ8mQ

Abstract

Users' trust in prediction models can be affected by many factors, including domain knowledge and experience with predictive modelling. However, it is not entirely clear to what extent and why this domain expertise impacts users' trust perceptions and evolutions. In addition, it remains challenging to accurately measure users' trust. We share our results and experiences of an exploratory case-study with a visual analytics system that incorporates a prediction model for time series. Through a mixed-methods approach involving Likert-type questionnaires and semi-structured interviews, we investigate users' trust evolutions and factors that affect their trust in the prediction model. Our results underline the multi-faceted nature of user trust, and suggest that domain expertise certainly affects trust, though it cannot fully foresee users' trust evolutions.