Can Software That Predicts Crime Pass Constitutional Muster? | Hacker News Reader

Can Software That Predicts Crime Pass Constitutional Muster?

npr.org

13 pointsbountie7 comments

Wikipedia:

> Reasonable suspicion is a legal standard of proof in United States law that is less than probable cause, the legal standard for arrests and warrants, but more than an "inchoate and unparticularized suspicion or 'hunch' "; it must be based on "specific and articulable facts", "taken together with rational inferences from those facts".

I don't see how software changes this. Police have been concentrating their patrol time (and hence stop frequency) on high-crime areas since time immemorial. This is just a more accurate tool for that. Actual detainment will still require justification based on the particular circumstances and actions by the individual, just like they always have.

As long as they're not investigating any particular person or entering any private space uninvited, I can't imagine how the constitution would limit where the police choose to patrol.

With regard to constitutional issues, the article refers explicitly to stops (temporary detainments) which require 'reasonable suspicion'.

If it's based on crime statistics, this software would be heavily oriented toward black communities. My guess is that will be regarded as racist / profiling. I think even more than a Constitutional issue, you're going to run into social sensitivities around profiling.

The thing about these statistics is that they are usually statistics on where reported crimes occur, and reported crimes of certain types.

Crime is essentially infinite these days. I could walk down the street in any neighborhood, and in the vast multitude of federal and state statutes, I would feel comfortable that i could arrest every single person i met for something.

Maybe they wouldn't get convicted. But i would legally be able to arrest them.

So then the question becomes "what kinds of crimes do we want o prevent, and where do these occur", that is what gets you into profiling, because it is actually profiling. You are making value judgements about what kinds of crimes you want to be looking for. Because of socioeconomic and other statuses, the offender profiles for these crimes differ, etc.

Even if you only make the value judgements, and do so blind to race, it's still profiling, just "offender profiling" rather than "racial profiling".

It might be less heavily oriented toward black communities than gut instinct, though.

If it's based on crime statistics, this software would be heavily oriented toward black communities.

So what? On what else than crime statistics do you propose to build a predictive system for crime? It doesn't matter which communities it picks as risky, people in those communities will scream "bloody murder" no matter what community we'd be talking about.

Reply on news.ycombinator.com