Can Facial Expressions Designate Presence Info?
from the seems-tricky dept
There are some customer support phone systems these days that claim they can recognize if someone is angry or raising their voice (or simply cursing at the hold music) -- but that's only the tip of the iceberg for having more advanced systems recognize your state. For many years, people have talked about how "presence" was going to change everything. If you could somehow let your computer know if you were busy or tired or free or whatever, perhaps it could help better prioritize everything that might interrupt you. While things like instant messenger do have some amount of presence information built in -- it's still mostly reliant on the user setting things (or noticing that you haven't done anything for a while and setting you to "idle"). However, there's been a lot of research done over the years about having computers and phones better recognize when you can and can't answer a call. Of course, it's quite a difficult problem -- and some new researchers are tackling a related issue by trying to teach a computer system to recognize what your facial expression means. Rather than using this system to keep you from being bothered when you're in a pissed off mood, though, it sounds like they're looking to pitch you depression medication when you're feeling down. Forget behavioral advertising, it's time for emotionally targeted advertising. However, they are looking at some additional applications, such as having a camera that monitors your face while you drive and tries to keep you alert if your face starts to appear bored. Of course, last time we heard about a similar plan from IBM -- about five years ago -- the system was supposed to tell you jokes and squirt water on your face (yes, you would have an automated Bozo the Clown riding shotgun). It's not clear that much has really changed over the years -- as these projects always seem to be continually in the future.
Reader Comments (rss)
(Flattened / Threaded)
Useful in ER rooms?
I had to go to the ER on Sunday to check for a broken toe (I fell, stayed painful 4 days later). Luckily for me, the ER was pretty empty. However, there were 2 other patients in the ER that were borderline violent -- some black woman who was admitted for endometriosis pain, had gotten into a fight with another woman, and was very angry, so the policeman nearby was ready to take action. It sounded like she had a long history of psychiatric problems. Also some white teenager with a bandanna over her head, very distraught, and I wouldn't have been surprised if she started cutting herself with the nearest sharp instrument.
Since all ER patients get those wrist bracelets now anyway, how about a detector that detects the warning signs for violent patients? Even "calm" patients could become violent if they are made to wait for 5 hours.
p.s. the toe bone was fine.
(reply to this comment) (link to this comment)
Unreliable data
Doesn't work so simply Dorpus. One of my mates who's a cop and another who works with very violent disturbed teenagers tell me similar stories - the outward emotinal state of a person has limited bearing on predicting their actions. Asking someone to "calm down" when they are expressing apparently "angry" emotion can also be an extremely counter-productive trigger that exacerbates the situation. A guy could be smiling and laughing and then pull a weapon on you just when you think everything is safe and sound.
The mind is much more disconnected and independent from physiology than many believe. Our bodies make many actions based on conscious or subconscious processes yield no relevant or useful information to an observer, in fact they can be entirely misleading.
You should study "The lie behind the lie detector" and the long term statistical evaluation of FACS pertaining to the discredited practices of polygraphy and facial expression decoding, which many law enforcement people have studied and realised are largely useless in practice outside a statistical laboratory based experimental setting.
That doesn't mean ALL physiometric data is useless to decoding internal states, just not precise cognitive/emotional states. A car steering system which monitors whether the driver seems tired is a great example of a technology that can work with great accuracy, since tiredness is a global parameter of the organism reflected in rection times and so on.
btw, I broke a toe once so you got my sympathy - people take the piss because it's "just a little toe", but I've been in many scrapes, blown up, electrified, thrown out of moving vehicles and I tell you nothing ever hurt as bad as that toe did.
(reply to this comment) (link to this comment)
Re: Unreliable data
I hear what you're saying. On the other hand, the majority of ER patients are not seasoned criminals who are expert at hiding their emotions. Might they be more predictable?
(reply to this comment) (link to this comment)
Very interesting...though, I would argue there might be some sort of dependendances on cultural behaviour. Surely, raising one's voice in some cultures is considered as a sign of anger...while in others, it is just a way to state a point.
I remember once while working in very small hotel in London, and at the same time there were workers from Portugal who when they chatted amongest themselves...to us, it seemd that they were arguing...but they weren't...it was just their way of stating their points.
(reply to this comment) (link to this comment)
VERY tricky!
Sounds like very promising research that may be useful in predicting trends. I agree with "I for one" in the inability of determining precise states of mind. I have been dating one woman exclusively for over 4 months, and I STILL have no idea what will happen next! Guys, back me up here! :-)
Oh, think that pinky toe hurt? Try a kidney stone my friends! Many women have likened it to child birth.
(reply to this comment) (link to this comment)
could be annoying
Although it may be nice to have a computer sense our mood/tiredness/attitude and tailor its interactions with us - I am sure that after a day or two when our computer hangs up that all important work call, or IM from our boss because we have the flu or a hangover or are just pissed because we accidentally hammered our thumb rather than the nail. What if we were in the middle of a monster Medal of Honor or Halo sesh and the computer sensing our frustration at being fragged for the hundredth time decides that our current activity is too stimulating and annoying and shuts off our network connection.
(reply to this comment) (link to this comment)
Re: VERY tricky!
Owwww! You win!
Yea sorry, no contribution to the discussion.
(reply to this comment) (link to this comment)
Why not...
instead of trying to prgram a detector that can tell if you're mad about waiting in line for 5 people just work to speed up the service system that is so slow that it makes people violent.
(reply to this comment) (link to this comment)
i can just see corporations spending millions a year on this product instead of hiring another 100 CSRs and then claiming that their "anger factor" on inbound calls is at "industry standard levels" although call times have doubled....
(reply to this comment) (link to this comment)
Why
WHy bother spending so much money on something that is silly? I won['t trust a system like that. If I am too busy to take a call, then its off or on vibrate. If you are wanting to filter your calls, then you have caller id. You can then select which is more important, the work you ware doing, or that cal from the boss. Why would the computer care if I am tired?
I will shut it off. If I am at work and I am tired.. there is the coffee pot that lurks not so far away.
(reply to this comment) (link to this comment)
Long term Dependancy & Inertia
There is just too much potential danger in becoming too dependant on these hi-tech wonders.
They sound great at first - but when you analyze the long term consequences of dependancy, and virtual elimination of casual human contact in the services industry....makes you become really concerned about using them
(reply to this comment) (link to this comment)
Add Your Comment