The Supreme Court Is Allergic to Math | Hacker News Reader

The Supreme Court Is Allergic to Math

fivethirtyeight.com

36 pointsangpappas10 comments

I'm a numbers guy, but I believe the court's skepticism around statistics is warranted. Many concepts in statistics appear to be simple, but have lots of subtleties. There are lots of hidden assumptions in many statistical numbers, and it's not rare to see profesionals making mistakes.

If the court has to decide in the case of a bridge collapse, there will probably be a good amount of advanced mathematical modelling presented as evidence, and the justices will ask for some expert advice, they would't try to grok the math themselves. Similarly with statistics, it would be very prudent for them to defer to the experts, and even more so, considering that statistics appears deceivingly simple on the surface, while it's so easy to get it wrong.

This isn't a good argument for ignoring statistical evidence though, it's a good argument for augmenting the judiciary with a technical/legal expert staff.

Expert witnesses don't really help the situation, as both sides will produce experts that make technical arguments.

Excellent suggestion. The article actually mentions this idea "McGhee, who helped develop the efficiency gap measure, wondered if the court should hire a trusted staff of social scientists to help the justices parse empirical arguments. " I think it's a very good idea.

As for my original argument, I didn't advocate for the court to ignore statistical evidence. I think courts in general should weigh all the evidence. They just should exercise caution with statistical arguments, it's only too easy to get it wrong.

The real punchline to this article occurs near the end:

"But maybe this allergy to statistical evidence is really a smoke screen — a convenient way to make a decision based on ideology while couching it in terms of practicality."

As a data scientist and former attorney, I can tell you that these people understand the math. This isn't about understanding, it's about giving up power. Judges do not want the law to act mechanically on important issues---if it did, who would need them? They want subjective terms in the law, subject to interpretation, so that they have some freedom to change outcomes that seem unjust to them. They want that power. And if they start ceding it to algorithms, their role will be diminished.

It seems more like a structural flaw in how the law is written. Arguably if a case makes it to the SC then it means the law was ambiguous to start with. For something like boundary allocations an algorithm could have been written into the law directly by lawmakers. Why does this even end up in front of the Supremes?

Switching to a fair districting algorithm will never benefit the party in power, since by definition the existing law put them there. The only change they're going to want to make is in the direction of further bias.

The UK is successfully going through a re-districting process at the moment. So I don't think it's impossible.

Not just the US. Supreme Court.

I believe most judges and jurist in the world are allergic to Math just because they had to choose between a Science or Humanities curriculum as if it were opposite one of each other.

This is a hugely important issue for not just any specific area but American society in general. One word I don't see come up in Mr. Roeder's otherwise interesting article directly is "emergent", as in "emergent effects", but I believe that is at the core of many debates around issues like privacy/surveillance. Certain problems only show up with large enough numbers, ie, they exhibit non-linear scaling effects despite looking linear at small scales, and courts cannot simply draw a straight line and get an accurate reading. Privacy and security are ultimately based on economics, so there are cases where empiricism is unavoidable.

To take a single example, the "expectation of privacy in public": historically it's been overall accepted (and seems like "common sense") that an individual fully out in public has minimal expectation of privacy. Anyone randomly going by might see them. And in turn it makes sense that someone (including law enforcement) could review any single point source (like a security camera) without issue. However, as a practical matter there has still historical been a sort of emergent privacy due to resource limits. Nobody can track every single person, and dedicating the resources to surveil any single target/organization imposed natural limitations. Additionally, courts exercised some control via warrants for certain levels of surveillance, including new technology. For example, in Torrey Dale Grady v. North Carolina a few years ago the SCOTUS ruled putting a tracking device on a person or vehicle is a search under the 4th Amendment.

But in a world of ubiquitous recording and sufficient storage and computing resources, a lot of point sources can become effectively the same thing as putting trackers on everyone despite each being within existing jurisprudence individually. The nature of "voluntarily" giving up data gets called into question. Combined with other data or alone, where people go, who they meet, and at what times can be used to draw statistical inferences about a tremendous amount of information any reasonable person would believe private. But it's only the number of data points that have changed and the resource feasibility for storing/utilizing that data, not their individual nature. There is no whizbang new tech needed, just "more", "cheaper", and "better math".

Are American courts ready for those sorts of questions? What about apportioning responsibility for future externalities whose effects only show up statistically? It'll be a big deal in terms of restraining the new powers major entities are acquiring.

  > Are American courts ready for those sorts of questions?

In short, no, and they never will be. Courts aren't supposed to be in the business of solving social issues. Their fundamental function is to adjudicate disputes where there has been concrete, identifiable, and remediable harm to one individual by another individual.

Remember that in a gerrymandering case the fundamental question is whether a _specific_ plaintiff has been _specifically_ harmed by gerrymandering. The best statistical model in the world isn't going to answer that question; at best it gets you halfway there.

The executive and legislative branches are supposed to be the organs of government that grapple with these issues; the ones that enlist legions of bureaucrats to aggregate data and decide on the best aggregate policies. The real issue is that we're asking too much of our courts because the other two branches of our government--especially the legislative branch--has become ineffective and derelict. And, yes, that's party because of gerrymandering. It's quite a dilemma, but that still doesn't mean that the courts should jump into the fray.

I'm skeptical of conservative arguments that courts should stay completely out of gerrymandering disputes. It's obviously self-serving, not to mention hypocritical in light of the fact that all the conservative justices wanted to strike-down non-partisan redistricting commissions as unconstitutional. But as a clear legal rule that approach has significant benefits. Among other things, it would force communities to return to the ballot box rather than resort to the courthouse.

Reply on news.ycombinator.com