[tsco77]

True. But agency is just how we rationalize responsibility being "fair", no?

The legal system is designed with "agency" in mind, but it's true purpose is to create law abiding citizens. We hold people responsible, because doing so alters behavior, mechanism of altered behavior is just window dressing.

[scooble]

> True. But agency is just how we rationalize responsibility being "fair", no?

I doubt this is a description many people would agree with - that agency doesn't exist but we just pretend it does so we feel better about locking people up to modify their behaviour. Aside from anything else, if we are probabilistic systems, we presumably can't actually do things like rationalise, or design things like legal systems. Similarly, I'm not sure it makes sense to talk about the legal system as having a goal if it is a (presumably slightly more complex) probabilistic system.

Of course, just as with materialism we can bite the bullet and accept that our typical descriptions of human life are strictly false and the real story is just a scaled up version of billiard balls knocking each other around (mutatis mutandis for the probabilistic picture). But if we want a picture that lets us keep some of our current view of human life, it is not clear that bringing probabilistic systems into the picture helps any.

[state_less]

> if we are probabilistic systems, we presumably can't actually do things like rationalise, or design things like legal systems.

Why not? Computers can reason and have constructed proofs. Probably a computer could construct legal arguments too. They might have some probability for errors (e.g. cosmic rays) that need to be accounted for and corrected, just like humans.

> But if we want a picture that lets us keep some of our current view of human life, it is not clear that bringing probabilistic systems into the picture helps any.

It might help people understand why people murder, or commit crime. Criminals might have errors that need to be corrected rather than focusing on a crime that needs to be punished.

[scooble]

Rationalising is a kind of ethical reasoning - constructing an account of something that shows it to be consistent with a wider body of ethical beliefs. So in your example, the idea of agency is a way of making the idea of responsibility mesh with our belief that things should be fair.

It is not clear to me that a computer can hold ethical beliefs, or that it would be troubled when some other beliefs are inconsistent with those beliefs and motivated to find a rationalisation that makes them compatible.