AUTHOR=Goto Nobuhiko , Lim Xue Li , Shee Dexter , Hatano Aya , Khong Kok Wei , Buratto Luciano Grüdtner , Watabe Motoki , Schaefer Alexandre TITLE=Can Brain Waves Really Tell If a Product Will Be Purchased? Inferring Consumer Preferences From Single-Item Brain Potentials JOURNAL=Frontiers in Integrative Neuroscience VOLUME=13 YEAR=2019 URL=https://www.frontiersin.org/journals/integrative-neuroscience/articles/10.3389/fnint.2019.00019 DOI=10.3389/fnint.2019.00019 ISSN=1662-5145 ABSTRACT=

Recent research has shown that event-related brain potentials (ERPs) recorded while participants view lists of different consumer goods can be modulated by their preferences toward these products. However, it remains largely unknown whether ERP activity specific to a single consumer item can be informative about whether or not this item will be preferred in a shopping context. In this study, we examined whether single-item ERPs could reliably predict consumer preferences toward specific consumer goods. We recorded scalp EEG from 40 participants while they were viewing pictures of consumer goods and we subsequently asked them to indicate their preferences for each of these items. Replicating previous results, we found that ERP activity averaged over the six most preferred products was significantly differentiated from ERP activity averaged across the six least preferred products for three ERP components: The N200, the late positive potential (LPP) and positive slow waves (PSW). We also found that using single-item ERPs to infer behavioral preferences about specific consumer goods led to an overall predictive accuracy of 71%, although this figure varied according to which ERPs were targeted. Later positivities such as the LPP and PSW yielded relatively higher predictive accuracy rates than the frontal N200. Our results suggest that ERPs related to single consumer items can be relatively accurate predictors of behavioral preferences depending on which type of ERP effects are chosen by the researcher, and ultimately on the level of prediction errors that users choose to tolerate.