Combining a conjoint experiment and machine learning model to include end-users in a constructive technology assessment: The case of seasonal thermal energy storage

Abstract

By definition, the mapping of technological development occurs in the context of uncertainty and with a risk of failure. Decisions taken at the onset of the development phase are significant because the technology has not reached the market yet. The technology loses its malleability during the development phase because of entrenchment. Thus, the objective of a constructive technological assessment is to embed social aspects – additional perspectives – in the context of technological developments. In line with extant literature, a pivotal perspective is the (future) end-user of the technology. Hence, market acceptance remains a constraining factor in technological development, notably for renewable energy technology. However, existing studies have not included (future) end-users in development phases. They have always taken place after the technology has reached the market.

To bring technological development and market acceptance closer, this study develops an innovative method that combines a conjoint experiment with a supervised machine learning model to predict the behavioural intention of end-users to accept a new (energy) technology. This study illustrates the proposed methodology through a conjoint experiment on seasonal thermal energy storage. The ML model uses input from a survey with a conjoint experiment to predict end-users’ market acceptance. With N = 12,096 (1008 participants exposed to 6 x 2 conjoint tasks), the ML model predicts behavioural intention with an accuracy of RMSE = 1.55 on a 10-point scale. The model is then used to predict hypothetical acceptance from mock end-users’ profiles and varying technological features. Thus, this study opens new perspectives for constructive technology assessment and conjoint experiment literature, and furthers discussions on social acceptance of energy technologies.