Profiling and the Tyranny of Stochastic Data

Frame to Left from

Profiling as extra-constitutional and immoral

Profiling is a strongly biased attribution of guilt and a targeting of individuals for adverse action purely based on group membership.  We will discuss profiling and it’s antecedent in the misreading of stochastic data (data with variability or noise as is the case with social data)  – both in cases highlighted in the media and in aggregate data.

Profiling places the profiled victim on aggressive intrusive scrutiny which can be indefinite, prolonged, or intermittent and unexpected. It can lead to considerable trauma to the victim as the invasion of privacy leaves him/her vulnerable to the whims of some, usually unknown, empowered aggressor. This aggressor is often sneaky and paranoid. One is not sure what will or when it will get you – a cause of ‘rage’ against which you can only flail your arms. It is deadly. There has been one too many cases of innocent profiled African Americans killed by law enforcement. Black lives matter. There were Japanese Americans incarcerated during World War 2.  Muslim communities in New York City were monitored after 9/11. Others who have been viewed with suspicion and profiled, likely include, among others, members of groups based on immigration status, ethnicity, political positions such as pro-life/pro-choice/far-right/far-left, sexual orientation, mental health and gun ownership. The lives and rights of people belonging in such ever expanding nets are in peril. Bloated, powerful, self-perpetuating surveillance organizations with an exaggerated sense of their importance and relevance could snuff out our rights and freedoms.

The abrogation of the civil rights of the profiled victim through this aggressive intrusive scrutiny is tantamount to a prison sentence. A prison with no walls – where the victim may neither stay in nor move out. Unlike the usual criminal sentencing, the victim is not told the ‘crime’ he had been tried for. Usually the victim is held captive for the possibility of a crime based on some perceived marginal increment in the crime rate associated with his social group and more likely based on the prevalent biases and stereotypes about the social group. When the aggressor is a government agent then the agent attempts to gain some legitimacy, sometimes, by seeking consent from a judge. In the criminal justice system this would amount to a kangaroo court with sentencing without a jury, without the defendant being read the charges against him and without the defendant being provided a counsel strongly advocating his/her innocence and contesting all arguments by the aggressor. This is both immoral and unethical. Shouldn’t there be a private independent counsel paid for by the transgressor if there is no crime?  Even such a pretense of propriety is unnecessary for vigilante private party aggressors.

Profiling and the tyranny of stochastic data

Let’s look at triggers of profiling. Much of it is justified using stochastic data which feeds existing biases. Stochastic data summarized using a large data set, can trigger small or moderate associations in the aggregate, between group membership and some misconduct. Such associations, after inferential quantitative analysis, are sometimes labelled ‘significant’, a statistical term which, in such contexts, typically translates to a few large proclivities for crime, mostly very slight proclivities or none and even some negative proclivities for wrongful acts for individuals in the social group. This is not the strong sense of something noteworthy that the word ‘significant’ conveys.  Even stronger group level associations do not justify action on an individual in a group before any crime is conceived or committed by the individual. What is usually done in media reports about wrongful acts is to reveal the group identity of the perpetrator and sometimes bring in aggregate data about the group. A recent newspaper report for instance identified a suspected wrongful actor as an illegal immigrant and then went on to report that crime rates for illegal immigrants were less than those in the general US population. Which helps somewhat, though what would the reader infer if there were an aggregate association with increased crime for that group. Readers need to know about the tyranny of stochastic data, through its excessive leverage in influencing opinion, both in the aggregate and through individual event reports, in curtailing the rights and freedoms of individual members of many social groups. The use of individual event reports to justify action on other group members has no logical support as any links to behavior by others in the group is even more tenuous than that coming from group level associations. To me these are just events with an agenda.

Events with an agenda

I like to use this phrase to refer to events which strongly support, at a visceral emotional level, a dubious prevalent political position. Think of instances such as that of Willie Horton, who committed assault, armed robbery and rape when out on a weekend furlough from prison, one among many others for whom similar furlough programs aided in rehabilitation. This case got a lot of press and helped a presidential campaign. There was much reporting in the 1970’s about a ‘welfare queen’ – a Ms. Linda Taylor who was eventually convicted of defrauding the welfare system of $8000.  Something which dismantled a safety net which we might need again given the churn in the job markets due to rapidly changing technology and markets, and globalization. Referring to such incidents as ‘events with an agenda’ almost implies that the events were staged to support such political positions. Given what is said and done with such events in the backdrop one might as well think of these as staged events.

You read an article about a multiple homicide. Some reads from such stories include “Gun-owner shoots and kills”, “Mentally ill person goes berserk’, “Person with this political affiliation acts out’, and ‘Religion/race based terrorism’.  None of this is predictive of behavior in other members of the social groups identified – they are going about their business peaceably pretty much as they did before the reported incident. A bulk of gun-owners, for instance, likely keep guns away from children and use appropriate safety features such as trigger locks, gun safes or gun cabinets. Though some controls such as limits on access to military style weapons and sensible checks at purchase may be justified.  Much more dubious is the claim that ‘good civilians with guns can prevent violence by bad people with guns’ made by a number of political luminaries based on one instance where a gun-owner did limit damage by a shooter.

I am actually partial to gun ownership by those who are trained on how and when to use them – though not any ‘civilian’ – good or bad. This comes off an experience as a child growing up in India – perhaps an ‘event with an agenda’ which has overly influenced my world-view. My father worked as a scientist for the government and we lived in this walled government compound. A little sheltered from the world around us. Some rumor in the city had triggered a communal riot between Hindus and Muslims. A number of people of both faiths had been killed. The air was thick with the smell of burning tires. And there was a rabid crowd of rioters, purporting to be Hindus, at the gates of our government compound armed with knives, daggers, spears and hockey sticks. They wanted to get to the Muslim families living in our compound with murder and mayhem in their minds. The gates were locked but it didn’t look like they would hold up for too long. It was then that a Sikh man who had served in the Indian Army pulled his rifle out and rushed to the gate. A pretty rugged tough man for whom it wasn’t too difficult to put up an act of aggression and convey potential harm to the rioters if they didn’t leave. That and a few rounds of shots over the heads of rioters saved a number of people that day.

Extra extra read all about it!

Let’s get back to the out-sized effect of such events in a few other contexts. Most people deemed mentally ill are either under medical therapy, and/or in counselling or self-directed approaches, psychological or spiritual. Some may be in distress leading often to withdrawal rather than the kind of engagement which would lead to violence – it is reported that individuals with mental disorders are more likely to be victims of violence then perpetrators – at a rate more than 11 times higher than the general population. It is a rare case where a pro-life activist fire-bombs an abortion clinic. Anti-fascist activists are very vocal but instances of violence are rare. Immigrants, for good reasons, maintain a low profile despite some very active discrimination against them in employment and pay. Yet any incident involving an immigrant leads to a replay of all the ugly characterizations of them made in 2016 to win an election. All things to remember as the media cries out ‘extra extra read (or see or hear!) all about it’.

Any profiling for crime yet to be committed can only be justified by an almost sure predictive model at an individual level, without using information which we typically wouldn’t use, for instance, when we hire a worker. Stochastic data in the aggregate or case reports are entirely inappropriate in justifying profiling. You may recall a horrific mass murder of attendees at a music concert. I will headline that as ‘Casino aficionado commits mass murder’. We know one person with casino associations who said he could shoot someone in plain sight and get away with it.

Profile Donald Trump!