Machine Learning Methods for Estimating Preferences: Adaptive Elicitation, Heterogeneity, and Meta-Attributes



I will discuss a family of methods for eliciting and estimating people’s preferences based on recent developments in machine learning and optimization theory. I will then focus on the problem of dynamically designing elicitation questions for estimating risk and time preference parameters of people - a type of "conjoint analysis for behavioral economics applications". The proposed method dynamically (i.e., adaptively) designs such choices to optimize the information provided by each choice, while leveraging the distribution of the parameters across decision makers (heterogeneity) and capturing response error. Finally, I will discuss work in progress on a method to estimate sparse meta-attributes that may capture the way people make choices using “simple rules”. Results from online studies of some of these methods as well as preliminary results of the recent meta-attribute estimation methodology will be discussed. Applications that used the proposed methodologies with various populations will be shown, and the potential benefits of the proposed approach for research and practice will be discussed.

This research seminar is organised by the Erasmus Centre for Marketing of Innovation (ECMI).

Contact information
Dr. G. Liberali