Slashdot Mirror
Microsoft Developing a Tool To Help Engineers Catch Bias in Algorithms (venturebeat.com)
Microsoft is developing a tool that can detect bias in artificial intelligence algorithms with the goal of helping businesses use AI without running the risk of discriminating against certain people. From a report: Rich Caruana, a senior researcher on the bias-detection tool at Microsoft, described it as a "dashboard" that engineers can apply to trained AI models. "Things like transparency, intelligibility, and explanation are new enough to the field that few of us have sufficient experience to know everything we should look for and all the ways that bias might lurk in our models," he told MIT Technology Review. Bias in algorithms is an issue increasingly coming to the fore. At the Re-Work Deep Learning Summit in Boston this week, Gabriele Fariello, a Harvard instructor in machine learning and chief information officer at the University of Rhode Island, said that there are "significant ... problems" in the AI field's treatment of ethics and bias today. "There are real decisions being made in health care, in the judicial system, and elsewhere that affect your life directly," he said.
Correctly read as: "Microsoft is developing a tool to help developers detect wrong bias in their algorithms."
to detect bias in algorithms, be used in an attempt to insert bias into algorithms, without detection?
Just spit-balling here.
The main problem with this endeavor is that the "bias" they are trying to suppress is actually the opposite of bias. They seek to treat people differently on the basis of identity politics instead of on their actual behavior. The AIs will naturally be confused by being disallowed to latch onto the strongest signals in the data.
... sewing machines.
It little behooves the best of us to comment on the rest of us.
Of course it is. From what I understand, in nearly all cases the algorithms that make decisions about routine stuff don't even have access to information about the person's race, nationality, gender, etc. If so, how is bias even possible? It sounds like the individuals it disfavors may have some kind of adverse event in their history that was fed into the algorithm. I.e. missing down payments, drove 50mph over the speed limit, did 2 years in Virginia for possession of fentanyl, etc.
Except in the case of car insurance, where gender is given, and is very biased against males, and for good reason. Bias against any other identifiable category, no matter how good of a reason, and the court of public opinion will summarily issue a guilty verdict, and then Hank Johnson will introduce a new bill banning algorithms.
From the article:
Northpointe’s Compas software, which uses machine learning to predict whether a defendant will commit future crimes, was found to judge black defendants more harshly than white defendants.
So that was an existing algorithm that judged somebody on how they were born rather than their individual behavior.
It's turtles all the way down.
Eliminating Bias from AI means discarding facts and data that violate SJW principals.
Remember, Citizen: Equality means including an equal number of every ethnic and minority group, no matter their relative numbers in society.
Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
I've been reading stories in removing bias from algorithms but still don't get it. What is an algorithm bias? If the results don't have perfectly flat distribution across sex, race, religion, and other protected groups?
If you are developing algorithms to predict let's say possible criminal behavior and it ultimately predicts higher crimes among those who actually commit more crime then you you have one of three choices 1) Keep it and use it responsibly or 2) Throw it away and eat your development costs or 3) Neuter it to the point of it not working, thus you fail.
Caution: Contents under pressure
I think we have to be a little more formal with terminology. The summary and most articles these days use "algorithm" and "AI" interchangeably. You can use an algorithm to train a machine learning model, but the model isn't really an algorithm in the classical sense.
The trained model can definitely have bias based on the training data. The classical example is, train a word2vec or glove model on the texts of wikipedia, then find the vector representations of doctor and nurse. You'll find that nurse is considered a female term while doctor is male.
This may be acceptable for trivial things like advertising or movie suggestions, but machine learning is now being used for important things like job application screenings. Many times the model can be very opaque and this bias may not seem obvious. Even worse, it seems every company now wants to have AI in their product, and may have half-rate data scientists that graduated from a data science bootcamp.
The research I've seen on this subject is serious work. In the case of the doctor/nurse vector representation, the goal would be to make the occupation gender neutral. The tricky part is that you'd still want the model to retain certain qualities, like mother being female and father being male.
Police software that predicts and detects looting and a riot in a poor part of a city.
The political SJW want that set to be a detected as a human rights demonstration. Free speech and locals who doing an airing of grievances.
When the owners of property protest against all the looting, crime and damage, thats "riot".
The SJW see the system as a bug.
Domestic spying is now "Benign Information Gathering"
Meanwhile criminals will use "biased" algoritms.
It's "principles". FFS, if you can't even spell derp right...
The whole point of Big Data is finding connections that humans wouldn't have noticed.
Finding connections to justify future scruitiny is one thing, but making a decision about someone's future based on connections alone is not.
The number of storks nesting on Danish houses is (famously) positively correlated with the number of children who live in those houses. You could imagine an algorithm which discovered this connection adjusting a family's health insurance risk by counting the number of storks on their roof. A moment's thought reveals that, despite what you've been told, storks don't cause children, but rather larger families tend to have bigger houses with more roof space.
Data is informative, but there is no substitute for human judgment when decisions must be made about people's lives.
sub f{($f)=@_;print"$f(q{$f});";}f(q{sub f{($f)=@_;print"$f(q{$f});";}f});
SJWs simultaneously complain that black people are being arrested more, and that an algorithm that predicts higher recidivism for blacks is improperly biased and should return the same risk for whites.
Explain this to one, and you'll get a blank stare followed by an accusation that you're a racist.
First example you cite has been shown to be based on flawed statistics, i.e., the algorithm was shown not to produce biased results on the data. Bad things happen when journalists try to do statistical analysis.
Reference: Flores, Bechtel, Lowencamp; "False Positives, False Negatives, and False Analyses: A Rejoinder to âoeMachine Bias: Thereâ(TM)s Software Used Across the Country to Predict Future Criminals. And itâ(TM)s Biased Against Blacks.â", Federal Probation Journal, September 2016, You can find the article here: http://www.uscourts.gov/statis...
To put it clearly: COMPASS expects black people to commit further crimes twice as often as they really commit further crimes, while at the same time expects white people to commit only half as often further crimes than they really do. So COMPASS is much more likely to predict a black person to become a career criminal than they really are, while it does not expect the same from white people. This is a clear bias, and it has nothing to do with the input parameters per se, but with the way they are interpreted by COMPASS.
Apparently, COMPASS ranks a combination of factors quite highly, that is more common with black people (especially a poverty background without stable families and poor education), even when each of the factor is only slightly elevated, while it does not put too much value on single factors even if they deviate highly from the norm (e.g. it strongly underestimates the recidivism rates for certain sexual offenses).
So COMPASS has a strong social bias, but does not look into individual traits too much. And because the social conditions in the U.S. are highly correlated with, it basicly punishes black people for being poor.
So maybe recalibrating the whole thing so that the predictions exhibit minimal error compared to the outcomes is perhaps an option?
Ezekiel 23:20
So now I'm a racist because SJWs are complaining that black and white recidivism isn't equal, when they should instead be complaining about the error rate? And yet they wonder why more progress isn't being made...
I agree with your assessment of the algorithm, but that's the criticism serious people are making. SJWs are shrieking about the higher chance, and don't care about distinctions like yours.
Are you really that ignorant? That doesn't prove COMPASS is racially biased. It proves that it, like most AI initiatives, is snake oil. We have tons of data on repeat offenders and career criminals. Checking it for simple accuracy should be thing 1 to do.
Now, let me turn that around, because people like you have been telling us for decades that poverty breeds crime and therefore criminal behavior needs to be combated with massive social programs that somehow never end poverty but we need more and more of them. Now you say this algorithm uses your side's ideas as a predictor, probably because whoever created it listened to you , and now you say it's wrong. Well, if it is wrong, what does that say about the premise in the first place?
Just as scientific racists hoped that science would justify and enable their racism, technoracists hope that technology will justify and enable theirs. Technoracism is only a couple of years old (about as old as this article), it's only arisen following recent advances in machine learning. The technoracists hope to exploit layered neural networks' inherent ability to launder and obscure the human biases they were trained on, and portray the results of this GIGO effect as being purely logical and therefore somehow justifiable.
The way to pull the rug out from under both scientific racism (which has been enjoying a renaissance recently) and technoracism is use the ethical problems inherent to prejudging people based on immutable traits to argue for why we should never engage in such activities on ethical grounds, no matter how scientifically rigorous or even statistically predictive they may be.
"When information is power, privacy is freedom" - Jah-Wren Ryel
Comment removed based on user account deletion
Comment removed based on user account deletion
It does not. Take a look at this Washington Post article
Note the first graph. For each risk score, chance of recidivism is approximately the same between blacks and whites.
What ProPublica showed is the reverse, that black defendants who do not reoffend are more likely to receive a high score than white defendants who do not reoffend. Given that black defendants as a whole are more likely to re-offend, this is unavoidable without making the predictor biased against whites instead. The Post article goes into this.
if race is not even a factor in compass, how could it be biased racially?
have you seen my sig? there are many others like it but none that are the same
if race isnt used as an indicator, but the output shows race XX more likely than race YY to do something..... maybe just maybe it means that there are other reasons why those results came out.
the correct answer isnt to disregard the bias, or to inject explicit bias to "correct" the "problem", it is to research what else could be the cause of those results.
have you seen my sig? there are many others like it but none that are the same
The only slippery slope is humanizing one group of animals (MS-13) and then other well know violent groups, until you humanize the general criminal to the point that it is impossible to uphold the rule of law and have a civil society.
Perfect example is in the UK you will get more time in prison defending your home with lethal force than the criminals that broke in and attacked your family with the intent of doing harm.
In a sane world you put those people down like the feral animals they are. Sorry you had a bad childhood and grew up on the shitty side of town, but that is neither my fault, and until you attacked me and mine, my problem. Criminals and bleeding hearts to do not get to have an input on how we deal with people once they make it our problem.
Ten years of prison, therapy, education, and then a high double digit chance of recidivism and back to jail after they hurt people. Better to just put 2-3 rounds of hollow point center mass and take care of the problem and get on with life.
Last time I checked it has never not been open season on bands of violent people that terrorize the community.
You've clearly never checked then.
humanizing one group of animals (MS-13)
By dehumanising them you're immediately refusing to attempt to understand them, their motives and why they persist against your efforts to eliminate them. Which means you'll fail.
until you humanize the general criminal to the point that it is impossible to uphold the rule of law and have a civil society
Humanising the general criminal is a sign of a civil society.
Dehumanise them and you're no longer civil.
Perfect example is in the UK you will get more time in prison defending your home with lethal force than the criminals that broke in and attacked your family
Complete and total lie. You will get no time in prison for defending your home with lethal force against a criminal you believe is using lethal force against you.
Of course, if the criminal is merely trespassing then yes, you'll go down if you kill them. As you fucking should.
In a sane world you put those people down like the feral animals they are.
Nice oxymoron. In a sane world you don't treat people like feral animals. Are you sane? The evidence isn't looking good.
Sorry you had a bad childhood and grew up on the shitty side of town, but that is neither my fault, and until you attacked me and mine, my problem.
Further demonstrating that you have no wish to live in a civil society.
Criminals and bleeding hearts to do not get to have an input on how we deal with people once they make it our problem.
Yes, they do. Everybody makes and agrees the laws society obeys, and enacting your own excessive violence against people makes you a criminal too.
Ten years of prison, therapy, education, and then a high double digit chance of recidivism and back to jail after they hurt people. Better to just
..fix the currently broken justice system. Not..
put 2-3 rounds of hollow point center mass and take care of the problem and get on with life
..because you too are a criminal and I'm betting you haven't put 2-3 rounds of hollow point centre of your own mass.
Hypocrite.
Comment removed based on user account deletion
Never heard anyone claiming skin color was the cause. Often it's just the best distinguishing feature.
Comment removed based on user account deletion
test this algorithm ?
5 out of 6 people enjoy Russian Roulette & 6 out of 7 Dwarfs are not Happy
And you're still talking about something else entirely. What I'm talking about are the countless people who think the algorithm is racist because it predicts the recidivism rate for black is higher than it is for white. Whatever other aspect of the algorithm you want to talk about, that's another issue. What my post is about are the people who see black>white=racism, when that's reality. The recidivism *is* higher. Sorry, that's the level of understanding most SJWs have, and the complaint they're making. The intelligent people talking about the real problems are a minority. If they only ever talked about the problems you're describing, I wouldn't be complaining since it's an entirely valid point. But that's not reality, and you're falling into the same trap: You want your reality to exist, so just act as if it does regardless of fact.
"But still, COMPASS shows a strong bias, as it overestimates the recidivism of black people by a factor of two while at the same time, underestimates the recidivism rates of whites by about the same amount."
That is completely false. Did you even bother to read the article you're posting about? It includes this sentence which might interest you: "The predictive accuracy of the COMPAS recidivism score was consistent between races in our study – 62.5 percent for white defendants vs. 62.3 percent for black defendants." And this one: "Across every risk category, black defendants recidivated at higher rates."
What the article said that you apparently misinterpreted was this: "These contingency tables reveal that the algorithm is more likely to misclassify a black defendant as higher risk than a white defendant. Black defendants who do not recidivate were nearly twice as likely to be classified by COMPAS as higher risk compared to their white counterparts (45 percent vs. 23 percent). . . . The test tended to make the opposite mistake with whites, meaning that it was more likely to wrongly predict that white people would not commit additional crimes if released compared to black defendants."
Your error was also pointed out to you by another poster. You are completely misstating what the article actually says when you say things like "COMPAS expects black people to commit further crimes twice as often". "Twice as likely to be classified as higher risk" does not mean "expects them to commit further crimes twice as often." It means what it actually says.
I suggest you read the article before commenting further.
Humanising the general criminal is a sign of a civil society. Dehumanise them and you're no longer civil.
Well, la-di-da. It must be nice to be so holy.
In a sane world you don't treat people like feral animals
He said from that dangerous space behind his keyboard, where no such people lurk.
Well, this is the thing. I rarely encounter physical danger because I live in a civil society.
It's not a coincidence.
As I said, holy. Where do you live that there is no crime and no feral humanity, I wonder? Nowhere on earth, I'm thinking.
You're strange. Most of Western Europe has low crime and no feral humanity. The rest of Europe may too, I'm just not up to speed on their statistics.
It's quite easy. You just act civilised instead of resorting to primal reactions all the time.
When MS makes a product that doesn't suck...they'll have bought a vacuum cleaner manufacturer.
The whole point of AI categorization systems is to uncover bias. We want the thing to make a decision for us, after all.
This is basically saying that MS is trying to create tools to make AI that doesn't work. I give them an high probability of succeeding.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
A more useful analysis would determine if skin colour was the only factor in the difference in rates, or other factors. (Hint: it's other factors)
About the only heavily socialist nation left is Cuba.
Yeah sure. Low crime rates. Japan has a low crime rate. So does Singapore. But UK, Norway, Italy, Ireland, France, and Sweden have nothing to brag about other than having somewhat lower rates than the USA. Maybe look up the numbers before spouting off in your oh-so-precious mode next time. https://www.numbeo.com/crime/r...
Ah, nice. You quoted an arbitrary index that isn't even based on actual crime.
The European countries you quoted all have very low crime rates. There are communities and areas within them that don't, but there are also communities and areas within them where the monthly crime rate is less than one per thousand people, and that's including harassment, graffiti and being noisy leaving the pub.
Oddly enough the US isn't entirely dissimilar. That's because many parts of the US are also civilised. Really, you should give it a try.
You're backing away from your original statement, which I suppose really you had to do, when presented with actual facts that sort of take you off the pedestal and place you back in the real world.
My original statement is that treating people inhumanly is stupid and counterproductive, and nothing I've said since contradicts that.
You lack reading comprehension as well as humanity.
I don't lack either. I also don't lack the ability to perceive reality for what it is, as you do.
And to help you out, the statement you backed away from regards "no feral humanity". Now it's "Oh, in some places there might be." Pretty soon maybe you'll make it to the truth.
What the mothering fuck? You're the idiotic cunt calling people 'feral humanity', not me. No, there is not an issue with feral humanity in Europe.
There are some criminals, but not many, and the justice system needs to assure that they are appropriately managed.
Just how fucking big are the blinkers you're wearing? You must be looking at the screen through a fucking pinhole. Fuck this, not replying to you any more because you are demonstrably incapable of understanding what I'm saying.
Actually someone else originally called them feral in the thread, but please, keep telling me how I don't know how to read. https://www.dailystar.co.uk/ne... Public attack on a policeman with a sword. Yeah, nothing feral there. And that's one example. I won't get started on grooming gangs raping underage kids; your brain might not be able to handle that part of reality.