Face Value: Remote facial expression analysis adds predictive power to perceived effectiveness for selecting anti-tobacco PSAs

Abstract

Perceived effectiveness (PE) is a validated tool for predicting the potential impact of anti-tobacco public service announcements (PSAs). We set out to evaluate the added predictive value of facial expression analysis when combined with PE in a remote (online) survey. Each of 302 tobacco users watched 3 PSAs and allowed transmission of webcam videos from which metrics for “attention” (head position) and “facial action units” (FAU) were computed. The participants completed scales for their subjective emotions, willingness to share on social media, and intention to quit smoking using the Tobacco Free Florida website. Based on PE, both ready to quit (RTQ) and not ready (NR) respondents favored the same PSAs but RTQs assigned higher PE scores. Negative PSAs (“sad” or “frightening”) were more compelling overall but RTQs also favored surprising ads and were more willing to share them on social media. Logistic regression showed that the combination of Attention + FAU+ PE (AUC = .816, p < .0001) outperformed single factors or factor combinations in distinguishing RTQ from NR. This study demonstrates that on-line assessment of facial expressions enhances the predictive value of PE and can be deployed on large remote samples.