Potential for Decision Aids based on Natural Language Processing
(A2023-114322)

AUTHORS

Maximilian Witte, University of Hamburg; Jasper Schwenzow, Universität Hamburg; Mark Heitmann, University of Hamburg; Martin Reisenbichler, Vienna University of Economics and Business; Matthias Assenmacher, Ludwig-Maximilians-Universität München

ABSTRACT

Decision aids help consumers navigate the growing complexity of the choices they face. The emergence of large language models could spark a new generation of consumer decision aids based on vast amounts of unstructured text data. We evaluate the feasibility and social acceptance of such interactive decision aids in the context of political voting. In this research we (1) develop a new generation of interactive decision aids that enables human-like interaction to support individual voting behavior and (2) demonstrate the potential of such decision aids to adequately reflect complex political position and reach large parts of society. First empirical results indicate that AI-powered decision aids are particularly useful for specific sub-groups. We discuss practical challenges and solutions for building AI-powered decision aids based on free user input.