Well, yes, but it doesn’t necessarily mean very much, since facial recognition software doesn’t necessarily work anything like the human brain.
It’s not even designed for the same goals. The human brain is designed to learn around fifty faces out of a population of a few hundred (most Stone Age peoples lived in bands of fifty or so, and most bands never had contact with more than a handful of other bands), and recognize them quickly and accurately. Facial recognition software is designed to learn thousands of faces, out of a population of hundreds of millions, and doesn’t need to be perfectly accurate, since (at this stage) there’s always a human operator double-checking the match. Plus, the population that the software has to cope with is a lot less uniform than the population that your average Stone Age human was exposed to; it has to be better at overcoming racial differences.
On the other hand, the software is the product of human brains; some or all packages might be hard-wired to recognize the features that their designers learned as babies.
All in all, examining how well software recognizes faces tells you more about software than it does about faces.
I have a work friend … “Susie”. She is good buddies with “Jen”, who I work opposite shifts with. (I work mornings to her evenings, I work evenings on her days off). We never are all at work at the same time, never hang out together at the staff room. The three of us are all of British/Northern European extraction, pale complected and vaugely blonde. (Susie is dirty blonde, going grey. I am sort of dark blonde, with more auburn tones. Jen is bleach platinum blonde. Susie is 6 years older than me, and I am five years older than Jen. Our height varies maybe by an inch or two, Im not sure. and we are all somewhat overweight, ranging from a few extra pounds to obese.
The majority of our co-workers are Asian, and unless they work with us directly, they get us confused. We get called “the triplets”, which confuses the other Caucasian staff members, since none of them think we look even remotely alike.
Occasionally people will insist they reported something to me, and it will turn out they reported it to one of my “sisters”
Indeed, it does. Because of its potential value as a security tool, facial recognition technology has been extensively tested. We know from published results of those tests that “males are easier to recognize than females and that older people are easier to recognize than younger people.”
Umm, OK, wrinkles and facial hair may be helpful in distinguishing faces, but it is easy to concoct a reason for the effect when we already know its direction. The report also states that “… results based on the limited amount of data suggest that people of Chinese origin are easier to recognize than people born in Mexico.”
Would most of us have predicted that outcome? (OK, Juan and Ling; you’ll get your turn.)
In all, the available data seem to support Cecil’s main point: that the definition of “looks alike” is a lot more complicated and subjective than many of us had guessed.
Before you can decide if one race looks more alike than another you first need to define “race”.
Race is a pretty slippery concept. The dividing line between races isn’t clean and are set more by custom than any hard measure.
The White race has quite a lot of variation in things like hair color because of how we decided to define “White”. We could just have easily decided that blonds and brunettes belong in different races.
I related a story in another thread about how an Asian former co-worker of mine confessed that he couldn’t tell what I meant when I identified a green-eyed, freckle-faced redhead as “the Irish woman,” because, me said, “You all look alike to us.”
It may seem that “race” is a pretty dicey concept, but the vast majority of individuals living today have no problem identifying themselves and others as belonging to specific ethnics groups; also, these identifications are generally consistent. Since racial groups are identified by and have an effect (albeit not always conscious) on people’s reactions and behaviors towards others, these “races” have to be recognized as genuine and important social constructs.
In fact, it seems more than a little disingenous to state that we must define race with perfect clarity to be able to say whether one race is more “recognizable” than another. When studies were done (referred to in my initial post) showing that black subjects recognized black faces more quickly than those of whites, and that the reverse was true of white subjects, there was no difficulty in finding subjects who were clearly self-identified and easily identifiable as being white or black. While such studies have been criticized, the idea that race is too artificial a construct to be studied has not been one of those criticisms.
Heck, when there were complaints about “Friends” showing only white people living in NYC, no one tried to defend the show by saying, “But they all have different hair colors and bone structures!”
At any rate, while the “One Race: Human” is admirable politically and sensible in terms of genetics, it is not currently an adequate approach for studying the social effects of race, studying racism and/or figuring out what to do about racism.
I’d like to suggest that this thread stick to recognition within races, and that a discussion of whether races exist, or how to define them, be opened in a different forum.
Sorry C K, you’re absolutely right. I apologize for going so far off track – what can I say, I’m easily distracted by bright, shiny object (and also by the chance to get on a soapbox). I’ll do my utmost stick to the original topic here, I promise.
Your information about Aspergers and facial recognition is incorrect. I am a pediatric neuropsychologist and i see a fair amount of kids and teens with Aspergers. They do not have prosopagnosia (the medical term for inability to recognize faces) as a rule. You are correct that they have a terrible time with nonverbal aspects of communication, whether it be recognizing facial expressions, tone of voice, body posture and position, physical interactions between people, etc. They do not have difficulty recognizing the physical features, unless there is another reason for it.
I felt that it was important to correct this because i don’t want people to have a misconception about the nature of Aspergers syndrome, which is being diagnosed more and more.
[Warning: Wildly OFF-TOPIC post – skip if you aren’t interested in the Asperger’s Syndrome part of the original post]
Dear Karen, I beg to differ, as would Simon Baron-Cohen and a few others.
My main occupation is autism eduction, so I have spent tremendous time researching Asperger’s Syndrome and high-functioning autism (HFA) myself, as well as dealing with a good number of autistic people of various levels of functioning. The question of whether Asperger’s includes difficulty recognizing faces seems to depend more on what the individual therapist regards as the “cut-off” between Asperger’s and HFA is. That cut-off is, as you likely are aware, a subject of much debate among the top experts in this field.
Certainly I have found a lot of AS and HFA people who report that they fail to recognize people who have drastically changed their hairstyle or clothing, and parents of HFA kids who report their children fail to recognize a favorite character on TV if the characters shows up with different hair or clothing than usual.
However, as you no doubt could point out in support of your statements, even the most “classically autistic” people have tremendous brain activity in rxn to a family member’s face. The difficulty in recognition does not seem to happen with people who are seen constantly.
Essentially, difficulty with facial recognition and in “reading” faces both seem to stem from the same neurological factor: people on the autism spectrum process facial images in the inferior temporal gyrus, which is used in the “normal” brain only to recognize inanimate objects.
At any rate, I’d love to discuss this further, so let’s take it to a new thread.
P.S. Please don’t shoot me, C K – it’s relevant to the original post, isn’t it? <Big innocent puppy eyes, plantive tone.>
Also off-topic… Jennifer I’d like to read more about the neurological processes involved in facial recogniction by persons with ASD, especially with reference to the inferior temporal gyrus. Could you help with some specifics? Thanks.
Since there is some interest in problems in facial recognition and/or face reading in autism-spectrum disorders, I’ve started a new thread in the “General Questions” area entitled “Autism Spectrum Disorders and face-reading/face recognition”. I hope anyone with questions or comments on the topic will find time to go over there and post (or repost) their thoughts.
Two things… first, a link to a very amusing (and rather difficult) quiz at www.alllooksame.com that might very well validate some of the discussion here, particularly as it applies to Asians.
Second, hopefully nobody will be offended, but as a Caucasian American who has lived in China for nearly 20 years, I thought I would share one of my favorite jokes:
Q: What is tragic about being born in China?
A: You have straight black hair, brown eyes, and even if you are a one-in-a-million kind of a guy, there are at least ten-thousand other people out there who are exactly like you.
I think a fundamental misassumption we are making is that there is such a thing as “white people.” Perhaps we would see more homogeneity if we looked at the real ethnic groups within this artificial category which we call simply “white”. And, maybe we don’t distinguish these ethnicities because they are now so interbred that it is no longer possible to figure out what the original tribes looked like.
It seems rather obvious to me that before this discussion can even begin, we must define what we mean by “white people”, “black people”, etc. The idea that “white people” look less alike than people of “other races” stems from a more inclusive (by ethnicity and geography) definition of “white people”.
The reason that anime characters are drawn as they are is not because the Japanese are all some inhuman homogeneous mass. It’s because the artistic conventions of SOME STYLES of anime demand that distinct natural traits be simplified to the point wherein they simply disappear. There are examples of anime where a far more delicate hand is used by the artists, and they do not resort to these conventional and unnatural markers for identity.