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...

Slight tangent: The article cites the ProPublica study on the Northpointe software in which journalists (not statisticians) reported the software as biased. What they left out is that an independent study found this study showing bias to be wrong.

Source: Flores, Bechtel, Lowencamp; Federal Probation Journal, September 2016, "False Positives, False Negatives, and False Analyses: A Rejoinder to “Machine Bias: There’s Software Used Across the Country to Predict Future Criminals. And it’s Biased Against Blacks.”", URL http://www.uscourts.gov/statis...

In fact the ProPublica analysis was so wrong that the authors wrote: "It is noteworthy that the ProPublica code of ethics advises investigative journalists that "when in doubt, ask" numerous times. We feel that Larson et al.'s (2016) omissions and mistakes could have been avoided had they just asked. Perhaps they might have even asked...a criminologist? We certainly respect the mission of ProPublica, which is to "practice and promote investigative journalism in the public interest." However, we also feel that the journalists at ProPublica strayed from their own code of ethics in that they did not present the facts accurately, their presentation of the existing literature was incomplete, and they failed to "ask." While we aren’t inferring that they had an agenda in writing their story, we believe that they are better equipped to report the research news, rather than attempt to make the research news."

The authors of the ProPublica article are no longer with the organization, but this article shows up in any news article about AI bias. The fake story just doesn't want to die...

With all that said, I have some hopes that algorithms will help make truly race-blind decisions in criminal justice. It's easier to test them for bias than humans, and decisions are made in a consistent, repeatable manner.

Interesting. Thanks for sharing. The decisions of linear models (like the ones used for sentencing, credit scores etc.) can indeed be explained. The credit bureaus sell explanations of the score and tips on improving the score as an extra service, for example. That doesn't mean that people will like the explanation... If the EU will indeed require explanations for algorithmic decisions, then models will be limited to simple linear models.

What's more objective than using an algorithm for sentencing that is using data that has been shown to be predictive for recidivism? Indeed it's not about disliking somebody but about public safety and deterrence.

Frank Pasquale left out a couple of details in his opinion piece. First, these algorithms are only used in determining sentences, not to determine guilt. At that point in the trial guilt has been determined beyond a reasonable doubt. At the sentencing phase rules of evidence do not apply anymore and almost anything goes. That's why the prosecutor puts crying victims on the stand. There have been two supreme court cases (Malenchik vs Indiana and Loomis v. Wisconsin) that challenged the use of algorithms in sentencing and both upheld (in the later with some minor restrictions) that these algorithms can be used in sentencing. The conditions in general are that the algorithm has to be scientifically sound. That was the case in both cases that were challenged as there is existing peer-reviewed literature that examined the algorithms. Frank neglected to mention that. The secrecy of the algorithms is a consequence of patent and copyright law btw. The algorithms in these cases are a scoring function. Math is not (and should not) patentable or copyrightable. In this case the consequence is that the manufacturers only recourse is to keep it a trade secret. That could be solved better, but in my opinion people shouldn't get their hopes up that there's some exploitable loophole in the algorithm or something.

We can debate whether assessments (actuarial prediction instruments) should be used in sentencing or criminal justice. I’m very much in favor as it does reduce bias and leads to reproducible results. It’s much easier to control for biases in decision making with statistical methods than it is to control or fix bias in humans. Does anybody believe that human judgement is less biased? You can read up on the work of Paul Meehl who spend his lifetime showing that even simple assessment tests outperform the judgment of trained clinicians. Part of the sentencing is taking into consideration how likely the perpetrator is to commit a new offense. Humans suck at making predictions and estimating probabilities. This is no different in criminal justice.

Let me end this with pointing out some of the positive change that systems like this have brought: early release from incarceration. Low risk prisoners are more frequently released early (not just from overcrowded California prisons when ordered to do so by a federal judge), and then put on probation/parole. And work out well it did: http://time.com/4065359/califo... The expected crime wave from federally mandated early release didn't materialize. In my opinion thanks to these prediction models.

There are many things wrong with policing and criminal justice in the US, but the move to what’s generally referred to as “evidence based practices” (incl. actuarial prediction instruments) has been pretty positive. The great part is that both Dems and Reps are behind the idea of risk assessments so we might actually see some change for the better.