Maybe there is hope for 2020: AI that 'predicts criminality' from faces with '80% accuracy, no bias' gets in the sea Springer Nature has decided against publishing a paper describing a neural network supposedly capable of detecting criminals from their faces alone. Word of this decision comes as top boffins signed a letter branding the study harmful junk science. The missive, backed this week by 1,168 researchers, students, and engineers, …
It might very well be that the AI got good at distinguishing mugshots taken for police purposes from whatever the authors used for their "non-criminal" set. And there's a thing: how do you know that random face images do *not* represent a criminal? Maybe the AI can distinguish photos of criminals that get caught...
Yes, probably their shell suit collars were visible in the pics.
Reminds me of when I was a student. One day the cops came in to our mess hall and were going round apparently randomly talking to people, then they came over to me. Turns out they were doing an ID parade that evening, and wanted people who were roughly a certain age, height and colouring to take part alongside the suspect.
When we turned up, the guy's solicitor, with the suspect were allowed to pick which of us they wanted in the parade. The suspect had turned up in a nice suit, obviously trying to look as uncriminal as possible, then gets put in a parade against a bunch of scruffy students. When asked if he had any objections, he did say that he stood out because we all "look like a bunch of fucking students".
I remember getting 20 quid for couple of hours doing nothing except sitting round and standing in a line briefly, and the police station had a bar so we got a couple of rounds in between performances and had a great time.
The three different witnesses all picked the suspect out. He was definitely guilty, because when he was standing next to me and they were about to bring in the last one he whispered "this fucker will get me for sure because he's an ex-cop".
"how do you know that random face images do *not* represent a criminal?"
Maybe a simple study? Load the systems with images of MP's. Wait a month and investigate as necessary.
My hypothesis is that we would either catch a few criminals or significantly reduce the support for AI and CCTV. Maybe even both? And if I'm wrong, we turn over the contents to the tabloid press to allow for a more thorough review.
I mean if its good enough for the plebs it must be good enough for those that make the rules.
I was just thinking about retro-phrenology. In the same way as the man with the little hammers could bring up bumps to improve character, will plastic surgeons be able to rehabilitate criminals or is it something that's too difficult to change, like their eyes being too close together?
>Those who are treated as criminals, have nothing to loose by being one.
This pretty much sums up the problem of being a young, black male in the US. What's amazing, given the circumstances, is that as many as two thirds of this population somehow manage to escape the net.
(Its not just the US, either. Remember the sorry tale of the bank manager (Dale Semer) who's obviously guilty because, well, just look at his hairstyle and jacket. Real bank managers don't loiok like that....)
It also gets worse. As it might even be seeing the tell tale effects of stress and other factors of poverty, illness and addiction that lend to making someone vulnerable to being involved in crime.
It's no excuse to those who are criminals, but often the system can be setup to cause more trouble, so it's easier to give reason to discriminate and take in the larger social setup.
Unlike the Victorian times this is not based on any theory. Just simple pattern matching.
Humans do it all the time. Sometimes we make mistakes and are surprised when a “criminal looking” person is found to be a violin player, but in majority of cases we find ourselves to be correct.
If it does not fit your nice ideology, it does not mean it is wrong.
This type of research is still based on the wrong premise. You should create an AI that determines whether someone is a criminal. It'll have 100% accuracy, obviously. You can then proceed to virtually prosecute those criminals, virtually sentence them, and virtually lock them away. Do it all in the cloud and it'll be much less disruptive than traditional (offline) policing.
There are perfectly reasons to write a paper on facial recognition of criminals but what it should be about is researching the confounding factors that have given an AI the ability to pick out pictures of criminals. There was a Chinese AI project on this which was shown to have developed the ability to recognise official police photographs of criminals because they included them in the training set. That would be good research as it would support understanding of the impact of the photographer on AI analysis of any pictures.
This was exactly what I was speculating about in my earlier post, not knowing about the AI project you mention - do you have a link?
I'm sure that the US AI trainers must have done things like eliminate the height marker along the wall behind the face image. Didn't they...?
I would not bet on it. A facial gaydar neural net was tried with photos with and without the the faces blanked out. It gave the same answers both times so whatever it was making decisions on (clothes? background?) it wasn't the faces. An interesting test of this Mr Cudoogo detector would be to try it with and without blanked out faces.
I think it sort of depends.
If your wearing a police uniform and your out and about and you see somebody that sees you, then snaps their head around to their friend and says something, and the pair then quickly stuff things in their pockets, turn around and start walking in different directions I think it's probably going to ring alarm bells for the cop because it's going to look as if they've seen you and are then trying to split up to avoid you, which probably has a reason. Some of those reasons may be innocuous, but may indicate that they are doing something dodgy.
But that's the behavior and body language that's kicked off alarm bells, and not actually the appearance of the person, so shown a static image how could the AI tell that there was an issue?
Suspect A: "Oh look there are cops, I'm scared let's get out of here"
Suspect B: "Don't wave your phone around, put it away or they will get upset"
Suspect A: "OK, I'll post on FB later, let's head over to the coffee shop"
Cops: "Those blacks (I say it nicely) are acting suspiciously, arrest them"
"be arrested for an arrestable offense"
Oh, I think recent videos have demonstrated that the police are quite capable of going from "what are you doing" to "assume the position motherfuggler" in about three heartbeats. All they need to do is escalate to the merest suggestion of violence (like you putting your hands up and saying "what the hell, man?") and there you go, an arrestable offence. Easy peasy.
I have been "stop and searched". Actually, technically the officer said "Excuse me sir, could I have a look inside your bag?" inside a no alcohol zone that was being enforced, probably to check that our gate guards were actually enforcing the license conditions of hiring the place.
I was actually one of the event organisers, and the bag contained a bunch of high vis jackets and radios for the other committee members. The chap looked into it, realised it was nothing he cared about and then produced a stop and search pad and started writing out a thing that he was probably supposed to give me, I just asked if we were done, he said yep just need to give you... I wished him a good day and walked off.
Not sure if he bothered finishing writing out the stop and search ticket I was probably supposed to take (thereby potentially doing my part towards skewing the statistics) but it's not as if I wanted it for anything. But anyway, my point is that if i'd have hid the bag behind my back and told him that he couldn't look in it and screamed "fuck da police" to look tough to my mates then yes, he'd have nicked me because I would have created a situation requiring him to.
My personal experience with the police is that they are just human beings trying to do a job and if you let them do their job they will let you get on with your life without any trouble whatsoever.
"Simply because to get arrested you have to be arrested for an arrestable offense, [...]"
IIRC in England the Blair government decided to remove the threshold that made something an arrestable offence. Police officers had too many laws and penalties to remember which constituted an arrestable offence. The Blair government also created 3000 new criminal offences.
It seems to have become possible in England for someone to be arrested purely to effect a house search - when a search warrant could not be obtained due to lack of evidence. A useful vague offence is "conspiracy to..." - which if true could carry a high penalty in itself.
This post has been deleted by its author
It's harder than it looks, apparenty: https://www.youtube.com/watch?v=waWugm79mQo.
but on a personal note, I find some people do look a bit dodgy from their features (nothing racial here, I'm white and I'm referring to white people). I don't know if it is exactly their facial features or their expressions or use of eye contact or whatever, but I'm automatically wary about certain people - the wife also tends to be wary of exactly the same people. There was a gang of 5 violent burglars on a BBC one programme yesterday and they all had "the look".
When certain people knock on your door, before they open their mouth, you automatically suspect they are up to no good; just something about them. You are ready for their scam before they claim to be from the gas board or that you have a slate missing on your roof.
Certain TV and movie actors also have "the look" and tend to be picked for parts where they play the baddy or psycho. Some have made a long career out of it. As to what they are like in real life, who knows. This is the problem with generalisations. I'd guess that there is maybe a bias of people who have the dodgy look who actually are criminals compared to the rest of the population but probably not enough to go down the rabbit hole of trying to make a science out of it. There are of course plenty of people who don't have "the look" who are dodgy.
I share the view that there are people you meet first time, passing, etc, and your internal bell just rings for no apparent reason (sometimes quiet, but persistent, background noise to be weary). On one hand, I'd love to see research into this and intuition as well, on the other hand, re. research results, I hear this quiet, but persistent, background noise that tells me the outcome won't be used for anything good.
As to mugshots shown on media though, I'd be VERY careful. Whoever is paraded for a mugshot opportunity, they won't be in a good mood, and our own bias tells us that they must be bad, bad people. And even if they grin stupidly (because one moment they were having a good night's sleep, the next they found themselves in a police cell and what the f... just happened?!) to us they're not just bad, they're evil monster with an evil grin, etc. Media illustrate text with a shot that is put there to reinforce the message, which is manipulation, but there's hardly any news that isn't.
I share the view that there are people you meet first time, passing, etc, and your internal bell just rings for no apparent reason
Confirmation bias means you don't remember the times you got it wrong.
Some decades back I worked with a guy who everybody thought was gay - man bag, thin leather tie, seriously fashionable clothes and he could accessorize(*). Nope, wife, three kids, straight as a very straight person. Looks can be very misleading.
(*) I'm trying to remember which film that joke came from.
When I lived in a big city the difference between people that shopped in the posh shopping areas compared to the poor was obvious.
The AI may be differentiating between poor and rich people (hairstyle, clothing...).
Other factors may be that the AI picks out the pale pallor and bony face of drug addicts.
In the UK, differences between Carribbeans and Asians are pronounced thanks to the 1970s expulsion of Asians from Uganda, Kenya and Tanzania (they were largely self employed entrepreneurs then sent their offspring into professions because self employment sucks) while many people from the Caribbean came over to fill employed (rather than self employed) menial positions.
Then we've got data sets. Is the AI working out the differences between mugshot photos and, say, passport style photos, mood (generally neutral in passport photos, for some reason generally less than neutral in mugshots...)
The whole area seems fascinating. A complete waste of time for distinguishing criminals but lots of lessons about history, bias and the importance of good data sets.
>When I lived in a big city the difference between people that shopped in the posh shopping areas compared to the poor was obvious.
That's the advantage of this technology. Human police were only able to spot criminals by simple clues such as skin colour. But this will be able to scan all the posh people in suits coming out of Liverpool st station in the morning and detect which ones who are insider trading or rigging libor.
Instead of the police grabbing a few minor drug dealers or car radio thieves they will be able to prevent the next financial meltdown.
"You probably think people with that "look" are criminals because that's the way they are portrayed in movies."
A friend of mine is an actor, he gets to play a lot of roles as a "villain", usually on TV, occasionally in films. He looks like a complete thug onscreen, but that seems to be a "look" he can create.
One of the many YT vids I've watched over the last few weeks was a flim-buff-lisicley piece which pointed out that Henry Fonda played his first villain in "Once Upon a TIme in the West" - up until then he'd only been cast as a good guy and Sergio Leone had to convince him that he could make the change. Didn't do too badly as I recall, so being a villain can certainly be "acted" (although I have no idea how well he acted as a good guy!).
I'd say you're able to tell when (most people) are forcing an expression, and it doesn't look right - so there goes the alarm. I bet there are some damn good fakers that would pass mustard, but that's the arms race I guess.
Although I've also had the same when in foreign countries and I haven't learned the cues from childhood so I don't think this intuition is anything to be relied on, quite the contrary.
Also, if someone is (or appears to be) either well above or well below you in a social hierarchy, that may trigger a fight or flight response, in my opinion.
Whether those responses are justified is case by case, of course. This research is a joke.
nah its just your prejudice based on social conditioning and fear of things you dont understand
During my "goth" phase as a teen families would cross the street to avoid the scary looking person dressed in black scowling (doing nothing and wanting nothing to do with anyone), usually into crowds of alchopop fuelled tracksuit clad teens who became known as chavs causing trouble in town centre.
Similar thing happened when i shaved my head (fund raiser for charity) i went from long haired hairy biker looking bloke to skinhead thug and the reaction was telling, especially people assuming i was pro brexit and anti migrant (couldn't be further from the truth)
fact is you cant judge a book by its cover
That said if any one knocks on my door and isnt delivering something i ordered then they are cold calling and i automatically dont care and send on way dont even bother to look at the face to know i dont want it lol
Also if you want to recreate the "look" its one part stubbing toe, 2 parts trying to remember pi to 60 decimal places, with a smidge of docile cow vacant thousand mile stare added for additional intensity, its a fine line to walk so you dont look like your severely constipated
I must say that both my kids and my wife tells me I look "menacing" when in public. It's funny, because I'm generally good natured towards everybody, and I mean no harm to anyone. It is true though, that I have zero patience for people who don't seem to hear or understand that "no", means no. And if they don't... well.
[sent from a soft-padded, windowless cell]
I have been told on three separate occasions by different people with no connection to each other that I look like a rapist.
This is oddly specific and fairly distressing and if I knew what was causing it I'd work to change it, because, damn. (*)
Also I'm a fat white dude and on the two occasions I've been out (in my earlier years) with friends and the cops stopped us, I was the only one put in handcuffs while they made sure we weren't doing anything illegal.
(* and here I've been spending all my time trying to cultivate the Evil Santa look! Whatthehell, man.)
"they all had 'The Look'..."
That generalisation might work with violent crime... broken nose, cauliflower ears... erm, sorry, that's a rugby player!
How do you identify a fraudster on your doorstep? Polite, mild-mattered, nice haircut... erm, sorry, that's a Mormon/Jehovah's Witness
"The people you're describing are misguided but they're not fraudsters because they believe they speak the truth."
Those people are the mules. They are exploited by people higher up the chain who know what they are doing is trying to pull in more followers who will contribute time and money. JW door-steppers often are an older couple plus a younger woman who seems quite vulnerable.
On one memorable occasion my "post" alarm sounded so I opened the door. An elderly man had a JW leaflet in his hand - and it was obvious his female companions were pushing the same through other doors. He looked at me sheepishly and said - "I'm sure you don't want this" and walked away.
Apparently they have an exclusion list - which they occasionally test to see if a new occupant has arrived.
The Mormons presumably still charge their followers a tithe of their earnings - and sell them the dogma allowed beverages etc?
Some Christian churches in London have been selling Covid-19 "cures" to their congregations at high mark-ups
Our local Catholic Secondary School has selection criteria for their religious quota of entrants that aim to establish how well parents donate to the local church.
I've been told that some people had referred my as "that guy with the shifty eyes".
Yes -- It's called a nystagmus, it's uncontrollable and besides making me look nervous it makes it really hard to look people in the eye when I'm talking to them. The stereotypical " criminal look" is exactly that: a stereotype.
"[...] it makes it really hard to look people in the eye when I'm talking to them."
IIRC the police expect innocent people to not look them in the eye when being interviewed as a defendant. Apparently the guilty often do the opposite - thinking it makes them look confident and innocent.
Downcast eyes are an apparently intrinsic trait of acknowledging an authority figure.
That's all just bias,.. And conditioning.... Men in three piece suits and $100 hair cuts ripp people of every day everywhere... They start wars and fuck others countries over for profit.
It's always the "he seemed so "normal" pillar of the community that up and kills his ex wife and kids, Catholic priests on any of the data set?
It's actually a trait of psychopaths to blend in and be "normal"....
ML gets another black mark against it because people who lack real intelligence think its a magic bullet, the fuckwits who thought research like this had merit deserve a kick in the balls.
AC> When certain people knock on your door, before they open their mouth, you automatically suspect they are up to no good;
What bollocks. You can't tell someone's predisposition to criminel acts from their face. Every fool knows you need to measure the bumps and lumps on their head! Science. SCIENCE!
Wait for it...
"We found that 99.9% of people walking into the Ritz did not have criminal features, compared to 85% of people walking into a benefits office."
Programmed by people that didn't add the faces of tax evaders, embezzlers, and stock market fraudsters to the database, naturally.
This is basically the same as the 'craniometry' that 19th-century 'scientists' used to try to explain why women were inferior to men and blacks were inferior to whites. Seriously, they took detailed measurements of skulls and skull-shapes, and estimated brain volumes and all sorts of things. They used their findings to 'prove' whatever they wanted so long as it fitted in with the social narrative of the day. Scientists believed it because it felt sort of sciencey. Even at the time though, some scientists dismissed it for the quackery it was.
> This is basically the same as the 'craniometry' that 19th-century 'scientists' used to try to explain why women were inferior to men and blacks were inferior to whites.
I was just going to post about the same exact thing - Craniometry. Two consequences of that particular brand of massive stupidity were Eugenics and Scientific Racism.
It turns out that some really stupid ideas refuse to die.
Yes - the Nazis used specific measurements to try to determine if you had any Jewish blood, Roma blood or whatever they wanted to hate... and then other specific measurements to determine how 'germanic' you were etc.
It does seem like racism can be dressed up as science well enough to fool even people who are respected as scientists. The sad fact is that we all know that science doesn't care what you think - facts are facts. But it doesn't stop you calling thing science when they aren't. Just because you take lots of measurements and write formal-sounding papers on it, doesn't make in any less false.
A neural network might be able to tell violent criminality by looking at a face, in the sense that it does better than guessing. You could look at whether there are facial injuries consistent with being in fights, like reset jaws, etc.. That would pick out a lot of innocent people as well, but also pick up a fair few violent criminals.
Also, and this is more speculative, iron and B12 deficiency is linked to aggression in boys. If it also happens to have some impact, however slight, on how the boy's face matures during puberty, then it could be possible to produce a statistically significant result over the course of millions of faces.
But I'm only talking about doing better than chance, not getting anywhere near the levels talked about in this paper.
I don't get the impression, from this report on El Reg, that the paper had any kind of rational way to define criminality. Without that the AI has no rational starting point., even before you start on trying to recognise it (whatever "it" is) with AI. And I don't think there ever could be one. Because it is defined by the laws that are broken Would such an AI, for example, be able to include rich people who hide their company's income and profits through a whole string of shell companies? Would the AI change in some magical way when a dodgy practice that enables tax evasion is outlawed and the actions become suddenly criminal. Or reverse itself when an activity is decriminalised. Or when a person crosses a border between localities when any given activity is or is not legal. Is the 18 year old who legally drinks in one US state somehow changed when they visit one where 21 is the age? Does the change occur as they cross the border? Or only if they lift a glass of alcohol to their lips?
i.e. There is no objective thing called criminality. Criminality is a breach of the law, whether wilful or not, whether the law is justifiable or not, whether the law is significant or not. And breach of that law remains criminality even if the law is later repealed as unjustifiable. It's a status, not a state of being.
And at what point does criminality make one a criminal . One offence? Two? Three?
I'm a criminal, I definitely drank before I was 18. I once tried an illegal substance, I'm sure I accidentally went through a red light once. I've definitely walked out of a shop with a newspaper and forgot to pay once. I frequently, as a kid and within the age of criminal responsibility, jumped off a bus without paying. And I've got on and off a bus while it was still moving!!!
And I'm equally sure that I've breached a few other laws (when was the one about archery practice repealed?).
Indeed!
Crime is not an innate physical characteristic of a person* so it's flawed to use that as a predictor. Of course, crime is experienced and managed in a social context, and those contexts are often biased.
There is also a huge variety of crimes, of varying severity and with varying consequences. And as you noted, there's no universal list of crimes.
Although it's unfit for profiling future crims, I'd love to see this study used to understand our social biases.
*Some characteristics are criminalized, like race, gender, sexual orientation, disability, but that's deeply unpleasant and I hope it wasn't the point of this paper.
Ah, Nature vs Nurture. How we act is a product of both. For example, one may possess a gene which codes for some type of 'bad' behaviour (like violence), but will only be expressed when certain environmental conditions are met. Many CEOs, MPs, scientists etc. might have been gangsters* had they been born in abject poverty in a slum. Similarly, many people now languishing in prison could have been CEOs, MPs, scientists etc. had they been given the same starting conditions as those who did achieve such status.
To say that a computer, or anyone, can tell by someone's face that they're criminal is a big load O' shite.
*(some might say they are anyway)
Having lived in Pennsylvania, I commend the researchers attention to the nearby state legistlature, long known as a repository of grifters, the kind who hold unannounced Saturday evening sessions to annoint themselves with a pay raise.
Another advantage for the veracity of the project is their photos should be publicly and voluntarily available. The disadvantage being the high percentage of positives.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
