Beyond Words: Predicting Preference Strength From Vocal Cues
(A2026-134111)

AUTHORS

Hauke Roggenkamp, ETH Zurich

ABSTRACT

Consumers increasingly make purchase decisions through voice-enabled technologies. In voice-mediated interactions, consumers not only communicate their choices through what they say but also through how they say it We demonstrate that so-called vocal paralinguistic cues, such as a person’s vocal tone and intonation, provide an unexplored window into consumers’ underlying states at the moment of a decision. Using gradient-boosted decision trees, we demonstrate that vocal paralinguistic cues reliably predict consumers’ preference strength and behaviorally relevant outcomes. We further demonstrate that the approach generalizes across choice settings (including product preferences to beauty ratings). This research provides firms with a novel lens into consumers’ preferences and decision states at the point of purchase. We also make our model publicly available and provide a Shiny app to use the model as a feature-extraction tool.