Another approach to (or rather away from) slippery slopes is to see the entire slope as a single thing à la TDT. Gandhi, contemplating his willingness to make the trade to become 95% Gandhi, can also foresee that 95% Gandhi would make a similar trade to 90% Gandhi, and so on. So his first decision is acausally linked to the whole of the slope, and to decide to take one step is to decide to go all the way.
The concept predates explicit TDT and can be found in popular wisdom: how often have I heard “there is no just once” in fiction, whether a policeman asked to break the rules just this once, an alcoholic offered just one drink, etc. Kant’s Categorical Imperative is similar.
Cf. the maxim “Everything you do is a decision about who you want to be”, or the outside-view version, “The way a person does one thing is the way they do everything.”
So his first decision is acausally linked to the whole of the slope, and to decide to take one step is to decide to go all the way.
(emphasis added)
No no no. His first decision is causally linked to the whole of the slope. If you draw out the DAG of causation, there’s an arrow going right from “became 95% Gandhi” to “became 90% Gandhi”, and an arrow going from “became 90% Gandhi” to “became 85% Gandhi”, and so on (with some intermediate nodes depending on resolution).
Thanks for noting the connection to Kant; ADT’s minimally-metaphysical yet deontological approach, i.e. its Kantian approach, is its hallmark and triumph. (TDT goes heavier on the metaphysics and is weaker for it.)
Note that Kant actually claimed that he was not preferring one consequence over another, he was finding a self-contradiction in one consequence, and no self-contradiction in the other.
That is, “you should not steal” because, at the end of the slippery slope, there is a self-contradiction, something like “if everybody ought to steal, what does theft even mean?”. (I’m trying to give Kant a fair shake, though I think he’s wrong.)
I’m really not very familiar with Kant, but I guess I always thought to frame it like, after an infinity of time certain norms will defeat themselves, and other norms will reinforce themselves, and if we are to support a self-defeating norm then that’s like choosing to become counterfactual. Then endorsing a self-defeating norm seems metaphysically but not phenomenologically possible, and since Kant didn’t care about metaphysics, it’s straight up impossible. Or something, agh. But anyway it looks ADT-esque, except that Kant had a hard time with meta levels for some reason.
I don’t think it’s right to say that Kant didn’t care about metaphysics. Given that his work on ethics was Metaphysics of Morality and that was based on the more-metaphysical Groundwork for the Metaphysics of Morality. Whether one can choose a norm that is inconsistent, is a question for the noumenal world, not the phenomenal world.
Another approach to (or rather away from) slippery slopes is to see the entire slope as a single thing à la TDT. Gandhi, contemplating his willingness to make the trade to become 95% Gandhi, can also foresee that 95% Gandhi would make a similar trade to 90% Gandhi, and so on. So his first decision is acausally linked to the whole of the slope, and to decide to take one step is to decide to go all the way.
The concept predates explicit TDT and can be found in popular wisdom: how often have I heard “there is no just once” in fiction, whether a policeman asked to break the rules just this once, an alcoholic offered just one drink, etc. Kant’s Categorical Imperative is similar.
Cf. the maxim “Everything you do is a decision about who you want to be”, or the outside-view version, “The way a person does one thing is the way they do everything.”
(emphasis added)
No no no. His first decision is causally linked to the whole of the slope. If you draw out the DAG of causation, there’s an arrow going right from “became 95% Gandhi” to “became 90% Gandhi”, and an arrow going from “became 90% Gandhi” to “became 85% Gandhi”, and so on (with some intermediate nodes depending on resolution).
I get the impression that “acausal” is an applause light here.
I think “acausal” (or logical) linkage was used as a generalization of “causal” linkage, as ‘dependences that TDT would take into account’.
Or possibly a typo.
I doubt that, given that he also said TDT instead of CDT, etc.
You are correct.
[de-jargoning for newcomers]
TDT, CDT, ADT := models of Decision Theory
DAG := Directed Acyclic Graph
Thanks for noting the connection to Kant; ADT’s minimally-metaphysical yet deontological approach, i.e. its Kantian approach, is its hallmark and triumph. (TDT goes heavier on the metaphysics and is weaker for it.)
It’s a funny sort of deontology that is justified by the agent’s preferences over its consequences …
That’s true, it is a funny sort of deontology.
Note that Kant actually claimed that he was not preferring one consequence over another, he was finding a self-contradiction in one consequence, and no self-contradiction in the other.
That is, “you should not steal” because, at the end of the slippery slope, there is a self-contradiction, something like “if everybody ought to steal, what does theft even mean?”. (I’m trying to give Kant a fair shake, though I think he’s wrong.)
I’m really not very familiar with Kant, but I guess I always thought to frame it like, after an infinity of time certain norms will defeat themselves, and other norms will reinforce themselves, and if we are to support a self-defeating norm then that’s like choosing to become counterfactual. Then endorsing a self-defeating norm seems metaphysically but not phenomenologically possible, and since Kant didn’t care about metaphysics, it’s straight up impossible. Or something, agh. But anyway it looks ADT-esque, except that Kant had a hard time with meta levels for some reason.
I don’t think it’s right to say that Kant didn’t care about metaphysics. Given that his work on ethics was Metaphysics of Morality and that was based on the more-metaphysical Groundwork for the Metaphysics of Morality. Whether one can choose a norm that is inconsistent, is a question for the noumenal world, not the phenomenal world.