“Biometrics” in Market Research – The UK’s Information Commissioner’s Office (ICO) posted a communication about the need for companies to assess the claims and scientific evidence produced by biometric vendors, as not all those claims have enough scientific support behind them.
There are a couple of really important points to discuss here.
1 Lack of clarity on the terminology – “Biometrics” in market research
The first is the lack of clarity on the terminology we use in the field about these techniques, and what each company does.
The actual reports (you can find them here) illustrate this clearly. They use the UK GDPR definition of “biometrics” as “personal data resulting from specific technical processing related to the physical, physiological or behavioral characteristics of a person, which allow or confirm the unique identification of that natural persona, such as facial images, or dactyloscopic [fingerprint] data”.
They then refine it to “Technologies that process biological or behavioral characteristics for the purpose of identification, verification, categorization or profiling”.
This is not “biometrics” as market research looks at it. Nor are these uses of facial coding, implicit associations or (neuro)psychophysiological monitoring as they are used in the field.
However, and, somehow, in a more problematic way, they then extend to (actually, not limited to) “microexpression analysis, keystroke analysis and physiological data for classification and inferences…”.
This is problematic as it now may include what we typically think of as “biometrics” in market research, puts all techniques in the same bag, and bundles these applications with the former (identity verification, etc.). This is a conceptual mess that does not help the field in any way.
Drop the term – MindProber example
We should be doing the exact opposite exercise: demarcating the uses of “biometrics” for market research from the uses for personal identification/verification. And within the industry clearly separate what each technique is, what it does, how it’s backed by actual science as well as state its limits.
This includes dropping the term “biometrics” altogether, towards accurately describing what we do. At MindProber we have fallen into the trap of using the term biometrics, as this is what the field talks about. We are now trying to move towards the less sexy, but more accurate psychophysiological monitoring term (and referring to each psychophysiological index in particular).
An important distinction
This distinction is important. For instance, the report clearly indicates that the shaky science is on the side of “Emotional AI”. According to the report, the “underlying science of emotion detection is highly contested…”.
I fully agree with these.
We have been advocating for dimensional vs. categorical models of emotion with the clients who care to listen to us since the inception of MindProber.
However, if you look at Annex A in the Biometrics Insights Report, MindProber does not fall into the Emotion AI bucket (which would mostly be populated by facial coding companies), but into the GSR one (we’re just scalable GSR with a neat platform to do media measurement/analysis).
There is no scientific dispute around the use of GSR as a measure of arousal in response to external stimuli. The science is absolutely clear here (and not disputed in the report). While we are guilty of not making this distinction loud enough, putting it all in the same bag creates confusion.
2 Consumer Neuroscience (or “Neuromarketing”)
Consumer Neuroscience (or “Neuromarketing”) is a really problematic field.
This is somewhat related to the above. During my tenure at MindProber, and having come from Academia, I have been many times shocked by the types of unsubstantiated claims that are made on a daily basis. We hear from somewhat innocent, but altogether irrelevant nonsense such as “90% of decisions are unconscious” to blatant abuses of the limitations of each technique in the name of Sexy NeuroBuzz.
I can’t count the number of times prospective clients have been underwhelmed when we explain the limitations of the techniques, or when we tell them that problems are complex and signals are noisy, or that there are no simple answers.
Many just turn to the company that will tell them what they want to hear (and that’s ok if they want more marketing than neuro, but they need to be conscious that they are drinking the Kool-Aid).
Others don’t, and I think they get a lot of value (but no easy answers) from the data we provide.
What clients need to know
The informed client must understand that each technique is different, has different technical characteristics, different degrees of scientific backing, and is appropriate for different research questions/purposes. We must show this and back up each of our claims beyond buzzwords. I admit this is not easy.
For instance, our recent debates around deciding whether to talk or not about “attention” show just this: we know clients want measures of attention, but in client meetings, it is really hard to move beyond the sexy buzzword and embrace the complexity of the phenomena in all its aspects.
So, even if the ICO reports on biometrics in market research were just around the use of techniques such as facial coding, implicit associations, eye-tracking, or (neuro)psychophysiological monitoring for the purposes typically seen in market research (and even if Emotion AI was all of this), scrutinizing claims would be a most welcome exercise.
This involves analyzing not only the validity of what is measured but also the inferences and claims companies make when selling, interpreting, and presenting results. I don´t think this would be a pretty exercise, but it is one the field needs.
This article was originally published on LinkedIn on October 26, 2022 by Pedro R. Almeida, CEO at MindProber
[Article updated: December 2022]
What do you think about the term ‘biometrics’ in market research?
Let us know in the comments below or send us an email – we’d love to hear from you.