Every time I poke my head over here you've got another 4 posts about me. It's like the quarantine zone for my 3 haters.

I'd be interested to read a rebuttal or write-up from you on what Pollack is getting wrong.

There's a lot more in this episode:

- How they built these tools in-house instead of outsourcing
- When involuntary commitment can increase risks for the seriously mentally ill
- Why anti-addiction drugs fail
- Why screening algorithms aren't more common in government

Full conversation is here:

Worth noting the algorithm doesn't determine whether kids get removed from homes, or override human decision-making.

The department has developed a screening tool for CPS: When a call comes in about a kid who's potentially at risk, the tool generates a "risk score" based on the integrated data.

That score helps a caseworker decide whether to do a site visit or not.

About 1/5th of parents involved in the child welfare system have a diagnosed opioid use disorder.

You're often dealing with the same people in these programs, just in different contexts. So it makes a lot of sense to integrate the data.

This problem doesn't just manifest in deaths.

A relatively limited number of drug addicts are disproportionately likely to be booked into jail, be homeless, or be involved in a child welfare call.

Like many mid-size Rust Belt urban regions, Allegheny County deals with high overdose death rates, largely from opioids.

Here's Allegheny County: think Pittsburgh plus.

This fries me: We architect these intricate pandemic response systems, but we don't ever test them.

Compare that with the military, which "live fire tests" weapons and systems constantly.