The Facial Analysis Problem
I have two beliefs which collide in the facial analysis space:
It is critical to measure event impact
Events need to deliver emotional impact
Over the past 4-5 years, facial analysis (and other) technologies have claimed the ability to measure emotional responses. I loved it. Finally, a way to actually measure if event content delivered resonance!
And while not perfect, experimentation with this technology did yield interesting data - recognizing when people were smiling or were frustrated, etc. It even read facial expressions in early post-pandemic times with many audience members in masks.
The technology and resulting data also, in my view, was somewhat problematic.
Here’s why:
It’s too good
Smart cameras, powered by artificial intelligence (AI), can sense not only what people are feeling, but can tell us a lot about them: age, gender, body type, ethnicity and more.
This is a problem. And as has been widely reported, AI has no sense of ethics (values need to be programmed in.) Data collected by smart cameras needs to be carefully filtered.
Many of the companies offering facial analysis solutions for event marketers claim to approach their work with values at the center. But I know from personal experience that they are reporting the age and gender of attendees. This probably makes sense for B2C marketers with a specific demographic target, but less so for B2B events.
So, in my B2B work, I generally try to ignore this data. I find it worrying and kind of scary. I don’t want age and gender data to bias my reporting or future event planning.
It’s not good enough
Smart cameras can also make mistakes, particularly amongst non-white people, as recently reported when a facial recognition camera misidentified a Black woman, leading to her arrest.
And it gets even dicier when you consider fluid gender identities and multicultural audience members who may disagree with how the AI classifies them.
The AI is just not good enough yet, in my view, to be trusted.
Facial analysis and the attention economy
In the world of events (and all marketing, really), a lot of attention is being paid to ‘attention’ - meaning, marketers want to know if their messages and content have our attention. It’s a reasonable consideration. If marketing spends a LOT of money on marketing and we ignore it…did it even happen?
I agree that getting and keeping people’s attention is important, and I spend a lot of time thinking about how to create stuff that is worth paying attention to.
Measuring attention, though, is far trickier. Passive attention measuring technologies (like facial analysis cameras) tend to assess if people are paying attention by their body language and where they’ve trained their eyes. Presumably if someone is looking at something they are paying attention to it. We daydreamers may know otherwise.
The problem with this approach is it doesn’t consider the diverse ways in which people pay attention. For example, if someone is autistic, their facial expressions, eye placement and body language will be atypical and likely misread by smart camera technology which looks at common patterns. I want to make sure everyone is counted…because everyone counts!
I pay attention with my ears much more than my eyes. I’m a listener. I’ll often not look at the screen or speaker so that I can pay better attention to the content being presented. The tech, however, thinks I’m focused on other things. If people are hunched over their phones, are they paying attention? Sometimes I’ll play a (simple) game on my phone to occupy the visual part of my brain so I can pay better attention to what I am hearing.
My wife is a doodler - scribbling on a notepad when she needs to pay attention to content. This got her in a lot of trouble at school with teachers who didn’t appreciate the method of her madness (until she aced their exams).
Facial analysis technology is…too good at putting people into categories, and not good enough at recognising their diversity, and doesn’t recognize that people pay attention in a myriad of ways. I’m still interested in seeing how the tech evolves but for the moment, I believe our industry has a facial analysis problem.