Los Angeles County's use of predictive analytics on at-risk children | Hacker News Reader

Los Angeles County's use of predictive analytics on at-risk children

psmag.com

29 pointspavornyoh8 comments

Combine this with California's vast DNA database populated by their Newborn Screening Program [0], add a dash of sale to third parties [1], and we're truly venturing into scary territory.

[0]: https://www.cdph.ca.gov/programs/nbs/Pages/NBSProgrOVforPare...

[1]: http://sanfrancisco.cbslocal.com/2015/11/09/dna-data-from-ca...

Predictive analytics can't even figure out what ads to show me.

I don't have any hope of it determining when a young person will become a criminal.

To me this just means that certain kids will be treated like they're going to become criminals, and because they're kids they will want to become criminals, and then the people behind this farce will say, "Wow look at all the criminals we predicted!"

Anyone else want to put down bets on the "at risk" kids here being right along the racial lines?

Well, a major risk factor for crime is poverty, and poverty is correlated with race. I would guess that it will be along racial lines, and that will be fairly accurate if we don't address the root cause.

The article asks "Should Los Angeles County Predict Which Children Will Become Criminals?"

That is a bunch of hooey. We can't predict individuals' behaviours. We can't even predict the behaviours of whole populations with anything approaching reasonable accuracy. And even if we could somehow predict future personality development we would still not be able to predict whether the person would survive long enough to become the criminal they're predicted to become. Noone in their sane mind claims to be able to predict things like accidents, diseases, etc. a few years down the line.

Let's be absolutely honest here. This "predictive analytics" is no more accurate than reading tea leaves. The people selling this technology are charlatans, fair and square. They are obviously well aware that their tech's ability to "predict" anything is no better than chance, but they are also aware that they are selling it to people who can't tell the difference.

>> “They looked at these public-health models and said, ‘This isn’t rocket science.

Well then it must be Common Sense.

It's outrageous that anyone might spend money on this (probably overpriced) software instead of on resources that would actually support children's development. That being said, I suspect that predictive analytics in this case is just a buzzword for actuarial science [*].

https://en.wikipedia.org/wiki/Actuarial_science#Actuaries_in...

I think it is quite possible to predict that some people are more likely to become criminals. For example, take a young kid and perform a traumatizing mock arrest on him. It helps if this kid is from a community where public-police trust has been historically low, and doubly so when general public-police trust is at an all-time low. At best, he will be sociologically labeled a criminal and tend towards such. At worst, he'll consciously rebel against the inhuman system that has prejudged him as a mere problem to be managed.

There was an interesting overview talk on predictive policing at 32c3: https://www.youtube.com/watch?v=wIQ2Xhov7D4

Reply on news.ycombinator.com