Hacker News Reader

> Let's query all phones with activity showing X among men in this geography and find ourselves a suspect.

Afaik this technically impossible with Apple Health at this point.

> Many people with limited funds for expensive lawyers will plea bargain for 2 years

Plea bargains rarely (if ever) happen in the Netherlands, and I suspect they don't in Germany either. The reasoning behind this is probably that when the state sues they should go for justice, not victory. Plea bargains may incentivize strong arming, bluffing tactics and money politics that don't belong in a court room.

Plea bargains (Verständigung in Strafsachen, § 257c StPO) happen in Germany, but not in murder cases, though a sentence might be reduced for a cooperative accused. They are more common in complicated financial crimes or small scale crimes like theft. There’s also a limited frame in which the judge decides on the punishment, it would be extremely odd that a potential sentence of 30 years is reduced to 2 years. Having less extreme prison sentences and a right to a paid lawyer reduces the pressure on the defendant.

> A more troubling use of the data, not mentioned in the discussion so far, would be query driven usage, for example: Prosecutor: OK, lets query all phones with activity showing X among men in this geography and find ourselves a suspect.

That was long a problem with DNA data.

I don't know how completely they test DNA nowadays in criminal cases, but in the early days and for many many years afterwards the tests did not determine a unique individual. It just narrowed it down. There could be a large number of people in the world that would match the sample.

There were two correct ways to use DNA with that level of matching.

1. After you have narrowed down your list of suspects through ordinary, non-DNA police work to a small number of good suspects, then check their DNA. If exactly one suspect matches it is extremely likely that is the correct suspect.

2. It can be used to nearly conclusively eliminate suspects. (I say "nearly conclusively" because you are almost never sure that the DNA you are testing from the crime scene is actually from the criminal. There are almost always ways someone who was not the criminal could have left DNA at a crime scene).

Then there was the completely wrong way to use it, which, unfortunately WAS often used:

• Keep a database of DNA samples from a small subset of the general population, such as people who have been arrested. When you start investigating a crime and think you have a DNA sample of the criminal, check that against your database, and if you get a single match conclude that person was involved.

This approach of starting by looking up DNA from the crime scene in your database and focusing everything on a single match is only valid if:

A) the database includes everyone who could possibly be the criminal, or

B) your DNA test is comprehensive enough that it does identify people uniquely.

We don't do A yet, and as I said I don't know if the test they use nowadays are good enough to satisfy B.

Isn’t this misuse of DNA data literally called the “prosecutor‘s fallacy”?

The DA says, “there’s a million-to-one chance it wasn’t this dude.”

Without any other evidence linking the purported perp to the crime, the reality is that the DNA only shows he’s one of 350 people who could have done it (assuming a population of 350m).

> A more troubling use of the data, not mentioned in the discussion so far, would be query driven usage

To be clear, this would not be possible on iOS, because the data is stored locally on the device (They had access to it). Health data cannot be remotely queried.

I have understood how this is acceptable in USA and in many parts of the world. This concept of plea bargains and not having to prove anything beyond a certain limit. I mean isn't it clearly the notion that miscarriage of justice is accepted?

I accept that a lot of this idea in my head is from films set in the USA but then again I have read on the subject and plea deals seem to be not only accepted but encouraged.

How a person, who seemed to have murdered someone, ought to be given only M years of sentence just because that person agrees for a plea deal where if properly tried and proven guilty the person might have been sentenced for N years which usually would be more than M? Just because the prosecution have to work less the culprit gets a lesser sentence? And as you say many who are arrested and haven't done anything but just wants to live with a lesser sentence just agree to such deals because they can't afford a lawyer, let alone a good lawyer.

Is it not questioned? Is it something in the constitution? How does it work? The "deal" itself - is it verbal or plea deals are actually signed and verified?

It is not in the Constitution. Some have argued it is in conflict with the Constitution. See https://mic.com/articles/11949/supreme-court-must-strike-dow...

The deal is in writing.

As you allude, though, it's a terrible system. The DA deliberately files charges one or two tiers higher than needed as a scare tactic to encourage defendants to plea out.

The motivation seems mostly to keep the number of actual trials down to match the capacity of the overall system. And it's one of the reasons (there are others) that our incarceration percentage is higher than anyone else's.

Especially considering how difficult it is, even for white collar professionals, to grasp a basic confusion matrix and adjust their probabilities for false positives.

For DNA testing, for example, the testing labs don't publish, and the courts do not allow questions regarding, their base rate false positive rate. They're worried that juries wouldn't be able to reconcile the "0.5% of DNA tests are false positives" with "beyond a reasonable doubt".

If you don't know the false positive rate, and someone comes in with a positive test result, you know almost nothing conclusive.