There is a petition to force Springer to pull an article apparently in the publication pipeline. It appears Springer has already caved. The article’s title is “A Deep Neural Network Model to Predict Criminality Using Image Processing”. I could not locate it online; its university host has already taken it down. Since I can’t read the article, I won’t be able to discuss its methodological merits — but that won’t be necessary, since the petition is entirely ideological. Its categorical tone brooks not dissent:
Data generated by the criminal justice system cannot be used to “identify criminals” or predict criminal behavior. Ever.
Ponder that statement, and take a moment to appreciate what a colossal betrayal it is by “expert researchers and practitioners [in] statistics, machine learning and artificial intelligence” of their respective fields. (Expertise devaluation is very much in vogue.)
Let us go on record as emphatically rejecting their categorical claim. Data generated by the criminal justice system can certainly be noisy or biased. If, say, the police, for some unfathomable reason, are fond of arresting and roughing up peaceful, law-abiding redheads, then statistical algorithms trained on this data will incorrectly predict criminal tendencies for redheads. There are ways to mitigate* such noise and biases — that’s pretty much what the field of statistics is all about! To say that data generated by process X can NEVER be used to make predictions regarding process X is barmy nonsense.
This wasn’t written by scientists. It was written by grievance-studies drones:
This crisis stems from the fact that machine learning scholars are rarely trained in the critical methods, frameworks, and language necessary to interrogate the cultural logics and implicit assumptions underlying their models. Nor are there ample incentives to conduct such interrogations, given the industrial incentives that are driving much machine learning research and development.
“Interrogate”. Did we mention that there were demands? Are such petitions ever issued without demands?
To reiterate our demands, the review committee must publicly rescind the offer for publication of this specific study, along with an explanation of the criteria used to evaluate it. Springer must issue a statement condemning the use of criminal justice statistics to predict criminality and acknowledging their role in incentivizing such harmful scholarship in the past. Finally, all publishers must refrain from publishing similar studies in the future.
As @cringelabcoat said on twitter, “This is really evil and you are ruining statistics.” To which they might respond, “Yes, but I’ll preside over the ruins.”
* Needless to say, if the redheads are being harassed by the police for no good reason, a reform is in order. But that’s beyond the scope of this post.