IIeX Behavior 2017, Meta-Analysis on the Use of Biometrics Data to Measure Emotional States Highlights

By Pedro Chaves
Edited by Ana Jovanoski 

The end of 2017 was particularly busy at MindProber – after our trip to IDEAS AMAI 2017, Mexico to present our case study, it was time to head to Chicago. We participated in the 2017 IIeX Behavior event and the plenary session on Behavioral Science in Advertising & Brand Building.

In a conference full of great content dedicated to the measurement of behavior in its multiple dimensions, we decided on presenting a part of the results of our latest research endeavor: (1) a comprehensive meta-analysis on the use of biometrics data to measure emotional states and (2) how MindProber is thriving in bringing scale, speed, and credibility to consumer neuroscience and the biometric media testing space. 

We knew it would be risky to have a particularly technical talk in a market research event, but the overall experience was quite positive. As a matter of fact, we received important compliments and insights on how to effectively communicate science. 

Biometrics data to measure emotional states – Meta-analysis highlights

Although the use of biometrics has been widespread both at the fundamental emotion research level and in the industry, no systematic analysis of its use has been conducted until now. We decided to tackle this problem and make somewhat of a double contribution: to publish a hard science piece with relevant psychological and psychophysiological findings and, at the same time, contribute to the increased awareness and seriousness of the consumer neuroscience practice. 

From an initial screening of over 1000 papers, following a strict meta-analytical approach, combining the results from every paper from 1970 onwards which included biometrics recordings and stimulation with emotional content (focusing solely on video), we aimed at answering a simple yet profound set of questions: 

• What are the most common biometric signals in emotion research? 
• What are the features of the most commonly researched signals and those with the highest discriminative power? 

Finally, following from the above questions,  

• Is there any evidence coming from psychophysiology for the so-called categorical emotional models of emotion? That is, are we able to distinguish from different emotional “states” as described by the classical basic emotions models, such as anger, sadness, or happiness? 

The most widely used research methods

The results were clear: galvanic skin response (GSR) and heart activity are by far the most widely used methods in research, being prime examples of indexes of the activity of the peripheral nervous system. 

Galvanic skin response and heart activity are by far the most widely used methods in research.

From an emotional standpoint, what can the data tell us? When looking at the combined effects from the literature, it’s apparent that we are out of luck if we approach biometrics from a categorical approach – that is, the changes induced by the stimuli are not sufficiently different from one emotional category to another. However, it seems that these same metrics are extremely useful to disentangle different levels of emotional arousal (i.e., how intense a given experience is, rather than how pleasant). And within the metrics battle, GSR comes out as a clear winner! 

This, of course, has a significant impact on how we conceive emotional states and how these seem to be represented in the body, but we’ll leave that discussion to the paper itself. 

Read our comprehensive meta-analysis on the use of biometrics data to measure emotional states on the following link: 

Moreira, P. S., Chaves, P., Dias, N., Costa, P., & Almeida, P. R. (2018). Emotional processing and the autonomic nervous system: a comprehensive meta-analytic investigation.