People often confuse and conflate “facial coding” with “facial recognition.” That's understandable, as both rely on AI and computer vision technology.
However, they are not the same thing, and it is important to understand why.
Facial coding is the science of measuring human emotions through facial expressions. Pioneered by Paul Ekman in the 20th century, facial coding employs either trained human annotators or computer algorithms. In the latter, automatic detection occurs via primarily front-facing cameras with opt-in audiences. There are many applications and societal benefits for this technology.
At the same time, media interest in electronic surveillance and facial recognition technology has reached an alarmist threshold, with many reports speaking of AI advances in facial coding and surveillance in the same breath. But the former rarely gives way to the latter, and concerns around many applications of AI are unfounded.
The defining characteristics between Facial Detection, Facial Expression Recognition, Facial Coding and Facial Recognition
- Facial Detection: To detect facial expression, the technology needs to first detect the location of a face. The output is a simple bounding box created by the coordinates of the face’s position.
- Facial Coding: The process of measuring human emotions through facial expression recognition, where appearance of certain facial expressions can be indicative of different emotional reactions. Facial expressions are classified by facial muscle movements such as the corners of the mouth, eyebrows etc. Context in which a facial expression is displayed also plays an important role in this process.
- Facial Recognition: Unlike facial expression recognition and facial coding, facial recognition uses vectors that map out facial characteristics and compare the data from multiple facial images to identify a single person.
Society sometimes judges rapid technology advances with a certain level of suspicion, though the fact remains: Facial coding and emotion AI have the power to transform industries and, when employed responsibly, improve people's lives.
The Responsible Application of Emotion AI
Although fictional scenarios depicting artificial intelligence (i.e., Terminator or HAL from 2001: A Space Odyssey) have dominated popular culture for a generation or more, many people don’t consider the difference between general artificial intelligence—a synthetic consciousness, or a simulation of one—and more innocuous AI applications. Much of the commercial use of emotion AI today falls squarely into the latter category.
The ability for computers to make intelligent predictions based on facial coding is not the stuff of science fiction, nor does it portend a dystopian future. It has real commercial applications today, and they’re based on real human feelings. In a 2018 report, Gartner predicted that 10 percent of personal devices will have emotion AI capabilities by 2022. In 2019, Gartner also predicted that by 2024 AI identification of emotions will influence more than half of the online advertisements you see.
Automated facial coding, specifically, has huge potential in a variety of applications, such as:
• Healthcare - pain detection, depression detection, tools for aiding people with autism, enhancing focus
• Education, recruitment and training, adaptive learning programs based on attention
• Gaming and entertainment apps - adapting to emotions of players, creating better UX
• Marketing and media - ads and content diagnosis, prediction of view-through rates and social sharing
• Vehicle industry - personalizing setting and experiences to individuals, driver safety warning and detection
And in this time of COVID-19, facial coding has higher relevance than ever before because it can work on remote devices with front-facing cameras, like PCs, laptops, tablets and phones -- with user permission. More than ever, people are adapting to life with more eyes on screens, with cameras activated for group communication and other utilities.
Automated facial coding can be both practical and ethical if it adheres to a few simple parameters:
- Opt-in: The people who participate in facial coding studies or real-time detection should overtly opt-in to having their emotional reactions measured.
- Transparency: Anonymize emotion measures by default, and only associate measures with individuals when there is a clear purpose and consent.
- Adult participants: Exclude children from facial coding, unless there is consent by a parent or guardian.
As with all new tech, we must set expectations for the proper use of AI in society. But we mustn’t paint with a broad brush and lump highly valuable, innocuous applications in with questionable surveillance tactics. The potential of facial coding to improve human experience and well-being is great, so it is only responsible to embrace and apply the science.