In The Face Of Danger
Artificial Intelligence To Keep Us Safe?
In the mid-1990s, a friend of mine was attending graduate school in Chicago. One night, she was asleep in her bed when she woke up suddenly to find a strange man crawling in her bedroom window. What she did next might have saved her life. She sat up and calmly, firmly said, “Who are you and what are you doing in my house?” Within seconds, the man mumbled about someone named Jose Garcia and clamored out. It was obvious, she said, the man was dumfounded by her reaction. I hate to think what might have happened had she panicked, screamed in fear, or even attempted to assault him. Was there something in her face that exuded confidence and power, enough to ward off a potential attacker? Or was she simply able to mask her fear enough to fool him? To what extent can our human intelligence “read” people?
We think we’re pretty good at it, most of the time. Most of us have internal radar that can help us sense people who are untrustworthy, for example. But how much can we detect? And could a computer potentially do it better than we do?
You may be familiar with the term “Artificial Intelligence,” a branch of computer science that studies how smart our computers can be—relative to human intelligence. As it turns out, a computer that can analyze our behavior is not all that farfetched. What if our machines could know more about us than we do? What if computers could recognize and even predict behaviors based on facial expressions?
In 2008, Steve Wilkins, the forensic services supervisor for the Pierce County Sheriff’s Office in Tacoma, Wash., used facial recognition software, a photo taken from a security camera at an ATM, and a database of prisoner mug shots taken over a 16-year period at the Pierce County jail to identify a suspect in a forgery and theft case. At that time, the office was one of two in the country pioneering the use of a program called Morphoface, which analyzes images and compares them against a database of available possible matches (Mulick).
According to Mulick, the technology itself has been around for a while, and it’s gotten some criticism for potentially violating privacy laws. In 2001, she says, it was used at the Superbowl in Tampa, Fla., to identify felons as fans made their way through the turnstiles into the stadium. More recently, in Toronto, the Ontario Lottery and Gaming Corporation is getting ready to unveil facial recognition software in 27 casinos. The system will scan everyone who comes into the casino and compare those photos to a database of self-exluded gamblers—in other words, people on a voluntary ban list. If they return to the casino, the system will flag them (Robson).
But what if this technology could recognize way more than just appearance? What if it could analyze the intricate movements of a person’s face to determine things like mood, state of mind, or even intent? What if the technology could not only identify criminals, but also the potential for criminal behavior?
A 2010 study published in the journal Advanced Imaging explores this possibility. According to the study, humans have approximately 40-90 unique facial muscles that create about 5,000 expressions. And it is from those expressions that we decide things like whether or not we trust a person, or whether or not that person is happy.
Using sophisticated computer technology, researchers at the Machine Perception Laboratory, Institute for Neural Computation (University of California, San Diego) were able to code and analyze facial expressions as never before—and, while the technology is not at the level and accuracy of actual human analysis, it can identify, through video streaming and its imaging database, emotions such as anger, contempt, disgust, drowsiness, pleasure, etc.
The technology works by focusing in on what are called “micro-expressions,” or the hundreds of involuntary facial movements that indicate mood or intention. If you’ve watched programs like The Mentalist, you get the idea. Our face, according to this study, gives signals even we don’t know about. I have to wonder—was my college friend’s face twitching involuntarily with extreme fear, at the same time that she was consciously disguising it? Is it easier to fool each other than it might be to fool a machine?
The applications of this technology are interesting, to say the least. Technology in automobiles can detect if a person is going to fall asleep at the wheel, indicate dishonesty on a polygraph more accurately, tell doctors if a patient is in real or “fake” pain (Nelson), or even suggest when a prison riot might break out, and who will start it (Lohr).
The possibilities are mind-blowing. Could a computer detect a crime before it happens? Could a personal camera placed outside our home, or carried on our person, alert us if a person is potentially dangerous?
This technology, of course, is based on the idea that Truth is absolute, that somehow the truth of situations and human behavior exists independent of what we choose to believe about them, that deviant behavior might be categorically defined, reduced to a specific set of criteria.
As Lee Nelson describes, “The truth always is present in some form. Trying to control one’s facial muscles can send mixed signals, leading others to be unsure about causal intentions. Changing how we feel, inside, alters our expression on the outside. Thus, even the best liars eventually give themselves away.”
Copyright 2011, Melissa Dereberry
****
Images: Microsoft.
Mulick, Stacey. “Facial Recognition Software Gives Pierce County Help in Tough Cases.” Tacoma News Tribune. Dec. 22, 2008, web edition: http://www.seattlepi.com/local/393308_computercrime23.html).
Lohr, Steve. “These Computers—Watch US and Learn.” The Virginian Pilot [Norfolk, Va.] January 2, 2001, Sunday Edition: A10.
Nelson, Lee. “Machine vision and your face: the Facial Action Coding System is a comprehensive inventory of the muscles and their movements that form frowns, glares, grimaces, and smiles.” Advanced Imaging 25.3 (2010): 12+. Academic OneFile. Web. 1 Feb. 2011.
Robson, Dan. “High-Tech Facial Recognition To Keep Gambling Addicts Out of Casinos.” Guelph Mercury [Ontario, Canada] January 12, 2011, Final Edition, Local: A4.
