Is facial coding a valid means of collecting emotional state?
Cheaper, faster, better. This is what we want and we want it now. But what is best is not always what is fast or cheap.
Facial coding is a popular method for collecting the emotional reactions in “neuromarketing” work and was recently featured in a WSJ piece, The Technology that Unmasks Your Hidden Emotions. The article focuses on the uses of facial coding in assessing consumer emotional response as well as potential abuses of the technology concerning privacy and public safety. These are definitely important issues. But it’s even more important to take a step back and think about what the technology is and should be useful for.
Can it do it? Yes… sort of.
Facial coding or FACS (facial action coding system) was first developed by Carl-Herman Hjortsjö and then later adapted by Paul Ekman as a means to record and categorize facial expressions to describe emotional response. Originally, these responses were coded and interpreted by people, making for very subjective data. Ekman’s work used 3 HD video cameras to get clean readings from people’s faces in controlled experimental settings. However, more recently computer automated programs analyze video-captured responses, as used by companies like Affectiva and Emotient. So it may be a good option so long as you don’t mind having to throw out a significant amount of bad data and over-recruit to make up for it (one must wonder the importance of the data points that get thrown out and if that skews the results).
Being able to capture consumer responses via quick videos (like surveillance in stores) and webcams is certainly an attractive idea to marketers. It’s cheap and it’s fast. But is it better than other methods? The WSJ article missed this important question, instead stating that it’s the best.
Facial coding certainly has its advantages for certain situations, consumers or questions, but definitely not for all. There are, in fact, a lot of different methodologies and technologies that can be used to measure consumer emotional response and a lot of situations where it may not be the best choice at all.
In fact, the very theories that facial coding is based on are still hotly debated. Ekman claims that there are 6, maybe 7, universal emotions: anger, disgust, fear, happiness, sadness, surprise, and contempt. Contempt, as a universal emotion, is less clear in the research and debate over the universality of the other 6 are also in question after several new papers came out this past year questioning the theory.
The real problem comes in when thinking of the practicalities of its use in market research. One already mentioned is the data throw-out rate which brings into question the validity of the data collected. But further, given the categorical nature and variable technology (webcams, single low-definition video), the sensitivity of facial coding may not be as good as other methodologies.
This can mean that if you want to check the performance of a product, you may not be able to tell the difference between reactions from two similar products, like fragrance or color of a lotion. Other methods, like fEMG (facial electromyography), measure the electrical changes in the facial muscles, heart rate, and skin conductance directly and are therefore more sensitive to subtle changes in emotional expression that may not be visible via camera. Further, using multiple methods to measure experience gives a more complete insight into the emotional experience. Being able to differentiate products and experiences in this way allows for meaningful and actionable conclusions.
Ekman says “Emotient’s software is highly accurate, but the accuracy of the system hasn’t been independently tested," as stated in the WSJ article.
Additionally, we know from our own research that people do not experience all the possible emotional reactions at the same time. Emotions develop over an experience and that development can be just as important, if not more important, to a product designer or brand manager than the global or initial response alone. As stated in the article: "Ken Denman, CEO of Emotient, says his company makes a point of discarding the images of individual faces within seconds after it has logged the sentiment they express. ‘There’s very little value in the facial expression of any individual,’ he said.” Unfortunately for this approach, humans simply aren’t this simple.
Example: Fragrance companies are great at making well-liked fragrances. So if you ask people to smell them and rate them, they will say they like them all. Now if you have 6 well-liked fragrances but need to choose only 1 for your product improvement, which do you choose? By using a multidimensional approach you can diagnose the differences (one may be slightly more arousing, or more comforting, etc.). Furthermore, by tracking those changes over time with a well designed psychological paradigm, you can get rich results that help a brand manager understand why one is better than another.
And yes, it may be argued that each has its best place. At HCD, we strongly believe that many validated technologies exist that have advantages and disadvantages depending on the question being asked. Facial coding may be easily used for gauging global "emotional" reactions to commercial watching or group behavior where accuracy loses to speed and large groups.
Other restrictions may also apply with any technology that is used, such as how well the technologies work together. In most situations, more than one technology is required to understand the consumer experience and so those technologies must work well together. For example, if you want to gauge the reactions of a person walking through a store, we would recommend using eye tracking to see what the consumer sees and focuses on in the experience. In that situation (walking, interacting), you would need eye tracking glasses, obstructing facial coding video collection, making this approach difficult – if not impossible.
It’s really a question of quality over quantity and the old saying that you get what you pay for. One thing you will notice when looking at the results of neuromarketing studies is that most of those "findings" are things you could have either gotten from your traditional measures (simply asking) or by rational hindsight generalizations or educated predictions. Recent headlines have included revelations like “sex sells” and “puppies make people happy”. We don’t find that kind of insight to be useful to clients. They need something more than that for the cost of the work to be worth it as well as useful.
“Neuromarketing” or consumer neuroscience shouldn’t be about “reading minds” or replacing traditional methodologies. Real and thoughtful applied consumer neuroscience is about using the right combination of sensitive physiological measures from psychology and neuroscience so we can get at the "why" of consumer behavior and this is something that can be most useful for making better products.