Facial Imaging's primary market is advertising – its technology consists in mapping 143 points on the face, activated by 43 facial muscles
When it comes to testing new products, most of us have been through the focus group experience. You sit behind a one-way mirror and watch a handpicked group of people dissect your new concept: a magazine redesign, a new website or a communication campaign. It usually lasts a couple of hours during which the session moderator does his best to extract intelligent remarks from the human sample.
Inevitably, the client – you, me, behind the glass – ends up questioning the group's relevance, the way the discussion was conducted, and so on. In the end, everyone makes up their own interpretation of the analyst's conclusions. As usual, I'm caricaturing a bit; plus I'm rather in favour of product pre-tests as they always yield something useful. But we all agree the methods could be improved – or supplemented.
Now consider Focus Group 2.0: To a much larger sample (say a few hundreds), you send a mockup of your next redesign, a new mobile app, or an upcoming ad campaign you better not flunk. The big 2.0 difference resides in a software module installed on the tester's smartphone or computer that will use the device's camera to decipher the user's facial expressions.
Welcome to the brave new world of facial imaging. It could change the way visual designs are conceived and tested, making them more likely to succeed as a result. These techniques are based on the work of American psychologist Paul Ekman, who studied emotions and their relation to facial expression. Ekman was the first to work on "micro-expressions" yielding impossible to suppress, authentic reactions.
The human face has about 43 facials muscles that produce about 8,000 different combinations. None of theses expressions are voluntary, nor are they dependent on social origin or ethnicity. The muscles react automatically and swiftly – in no more than 10 or 20 milliseconds – to cerebral cortex instructions sent to the facial nerve.
Last month, in Palo Alto, I met Rick Lazansky, a board director at the venture capital firm Sand Hill Angels. In the course of a discussion about
advertising inefficiencies (I had just delivered a talk at Stanford underlining the shortcomings of digital ads), Rick told me he had invested in a Swiss-based company called Nviso. Last week, we set up a Skype conference with Tim Lellewellyn, founder and CEO of the company (Nviso is incubated on the campus of the Swiss Federal Institute of Technology in Lausanne where Dr Matteo Sorci, Nviso's chief scientist and co-founder, used to work.)
Facial Imaging's primary market is advertising, explains the Nviso team. Its technology consists in mapping 143 points on the face, activated by the 43 facial muscles. Altogether, their tiny movements are algorithmically translated into the seven most basic expressions: happiness, surprise, fear, anger, disgust, sadness and neutral, each of them lasting a fraction of a second.
In practice, such techniques require careful adjustment as many factors tweak the raw data. But the ability to apply such measurements to hundreds of subjects, in a very short time, insures the procedure's statistical accuracy and guarantees consistent results.
Webcams and, more importantly, smartphone cameras will undoubtedly boost uses of this technology. Tests that once involved a dozen people in a focus group can now be performed using a sample size measured in hundreds, in a matter of minutes. (When scaling up, one issue becomes the volume of data: one minute of video for 200 respondents will generate over 100,000 images to process.)
Scores of applications are coming. The most solvent field is obviously the vast palette of market research activities. Designers can quickly test logos, layouts, mockups, story boards. Nviso works with Nielsen in Australia and New Zealand and with various advertisers in South Korea. But company execs know many others fields could emerge.
The most obvious one is security. Imagine sets of high-speed cameras performing real-time assessment at immigration or customs in an airport; or a police officer using the same technology to evaluate someone's truthfulness under interrogation. (The Miranda Warning would need its own serious facelift …) Nviso states that it stays out of this field, essentially because of the high barrier to entry.
Other uses of facial imaging technique will be less contentious. For instance, it could be of a great help to the booming sector of online education. Massive Open Online Courses (Moocs) operators are struggling with two issues: authentication and student evaluation. The former is more or less solved thanks to techniques such as encoding typing patterns, a feature reliably unique to each individual.
Addressing evaluation is more complicated. As one Stanford professor told me when we were discussing the fate of Moocs, "Inevitably, after a short while, you'll have 20% to 30% of the students that will be left behind, while roughly the same proportion will get bored …" Keeping everyone on board is therefore one of the most serious challenges of Moocs. And since Moocs are about scale, such a task has to be handled by machines able to deal with thousands of students at a time. Being able to detect student moods in real-time and to guide them to relevant branches of the syllabus's tree-structure will be essential.
These mood-analysis techniques are just nascent. Besides Nviso, several well-funded companies such as Affectiva compete for the market-research sector. The field will be reinforced by other technologies such as vocal intonations analysis deployed by startups like Beyond Verbal. And there is more in store. This story of Smithonian.com titled "One day, your smartphone will know if you are happy or sad", sums up the state of the art with mobile apps designed to decipher your mood based on the way you type, or research conducted by Samsung to develop emotion-sensing
smartphones. As far as privacy is concerned, this is just the beginning of the end. Just in case you had a doubt …
