Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[silverspawn]

I don't think that "if you don't have a gun, you might try to commit suicide and fail" is the important factor though. I think the important factor is "if you have a gun, you will attempt suicide because there is an easy way to do so"

[silverspawn]

... and obviously you also need to be suicidal to commit suicide; you won't do it just because you have a gun. But the point is that I think the difficulty of doing it otherwise provides an obstacle that's genuinely large enough to be the deciding factor in many cases. People aren't rational decision makers about this kind of thing, they act on impulse in the moment.

[BryGuy]

So i noticed the suicide rate followed the population density. When i ran the correlation of these limited ten US states, it was 98.5% for Suicide & Guns and only -89.4% for Suicide and Population Density.

[Awaclus]

I don't think that "if you don't have a gun, you might try to commit suicide and fail" is the important factor though. I think the important factor is "if you have a gun, you will attempt suicide because there is an easy way to do so"

As it turns out, "if you don't have a gun, you might try to commit suicide and fail" is the important factor. Studies show that:

• People who die by suicide are more likely to have had access to a firearm.
• People who have attempted suicide and failed are less likely to have had access to a firearm.
• There is no difference in gun access between people who are suicidal and people who aren't.
• There is no difference in gun access between people who have made actual plans to commit suicide and people who haven't.

This is all consistent with the claim that guns increase suicide success rates, and while it doesn't outright disprove that guns increase suicide attempt rates, there's hardly any evidence for that being true and a striking lack of solid evidence where you'd expect solid evidence to show up if it was substantially the case.

https://www.rand.org/research/gun-policy/analysis/essays/firearm-availability-suicide.html

[silverspawn]

That doesn't seem right to me, but my brain is too tired today to respond, will get to it tomorrow

[silverspawn]

So i noticed the suicide rate followed the population density. When i ran the correlation of these limited ten US states, it was 98.5% for Suicide & Guns and only -89.4% for Suicide and Population Density.

mh yeah that's interesting and not obvious (and according to GPT-4 the correlation survives if you control for gun ownership and politics)

[silverspawn]

I'm extending tomorrow to tomorrow, not in the right emotional state today

[silverspawn]

Ok, time to get to this. So first, the four claims

• People who die by suicide are more likely to have had access to a firearm.
• People who have attempted suicide and failed are less likely to have had access to a firearm.
• There is no difference in gun access between people who are suicidal and people who aren't.
• There is no difference in gun access between people who have made actual plans to commit suicide and people who haven't.

The first is predicted by my model (unsurprisingly). The second is also predicted by my model because I do think some amount of failed suicides are a thing, and this statement picks out those people. Like it doesn't matter whether failed suicides are 70% or 15% of the story (which would be my guess); either way that statement would be true.

#3 and #4 are less obvious; I think 4 surprised me a little. But not much because I don't think it's about planning I think an analogy is to a pot whose water level goes up and down and if it exceeds a threshold, they are then read to commit suicide. Then if there's a gun there, they probably do it, whereas if there's not, they have to figure out another way, and by the time they do, the water level has probably gone down again

[silverspawn]

Then the article hmm it's quite long, but I'm interested in this topic, maybe I should read it.

And actually GPT-4 thinks that your position reflects the academic consensus better, i.e., the largest factor is due to failed attempts.

[silverspawn]

Ok I'll read listen to it, but one thing I want to mention first is that the boundary between both hypotheses may be kind of fuzzy. If someone who doesn't own a gun has an impulsive moment and swallows a couple of sleeping pills, this would probably count as an attempt, but it probably doesn't have much chance of working.

[silverspawn]

But I guess if you restrict it to serious attempts it kind of becomes vacuously true

[Awaclus]

The thing about failed suicides is that people intuitively understand (correctly) that guns are an exceptionally effective method, so they tend to use guns if they can, but people vastly underestimate how difficult it is to get yourself killed in general, so people with no access to guns usually assume (incorrectly) that they can just pick some other method and still succeed with a high level of certainty.

[silverspawn]

Yeah, even having not read the article yet, I think I'm ready to concede the point just based on realizing that the spectrum of attempts is so broad. I was initially just thinking of more grand attempts like, idk, not jumping from a high enough building? Actually I didn't really have anything concrete in mind.

[faust]

jumping from a high enough building? Actually I didn't really have anything concrete in mind.

Seems like concrete is exactly what you had in mind!

[LaLight]

so, well, this happened

[LaLight]

we're fine for now

[silverspawn]

what's "this"? Should I be watching the news?

[LaLight]

what's "this"? Should I be watching the news?

oh, if you don't know, i would probably not recommend doing this, just for sanity sake

[silverspawn]

sounds good

[Awaclus]

what's "this"? Should I be watching the news?

I know what it is!

we're fine for now

No need to doxx yourself but are you living somewhere acutely unsafe?

[LaLight]

what's "this"? Should I be watching the news?

I know what it is!

we're fine for now

No need to doxx yourself but are you living somewhere acutely unsafe?

we are in the north.

[silverspawn]

This is a pattern I've noticed a bunch recently. Since GPT-4 is a stochastic parrot, it can happen that tells you one thing and then realizes that its own output is wrong. It figures stuff out as it goes.

Also props to me, I did think of the golden ratio, but it's defined with square roots, so it can't be transcendental.

Also also, almost all numbers are transcendental, but only two of them are defined in non-ugly ways. Typical math stuff.

[faust]

Also also, almost all numbers are transcendental, but only two of them are defined in non-ugly ways. Typical math stuff.

Fun sidenote: I recently learned about the concept of a computable number (which is basically a real number x for which there is a finite, terminating algorithm that can decide whether y > x or y < x). Turns out every algebraic (i.e. non-transcendental) number is computable, and so are pi and e. However since you can enumerate them with Turing machines, they still form a countable set and thus almost all real numbers are non-computable (but we don't really have any good examples of those).

[silverspawn]

Neat, so the ordering goes

Rational ⊂ Algebraic ⊂ Computable

but even the computable set is still coutable

The thing about non-computable numbers that I immediately think of is that a specification of a number usually gives you a blue print of how you would construct such a turing machine (this is also the case for Liouville's constant I think). So the concept of a good example for a non-computable transcendental number is a little paradoxical. You'd have to specify it in a weird backwards way. Whenever you know a sequence of rationals that converges, I think maybe that's already enough to make it computable? Mabye you also need some property of how fast the seqeunce converges, not sure.

[silverspawn]

Probably just having a convergent sequence isn't enough since you don't know how far you need to go to make the remainder sufficiently small to decide if it's bigger or larger than y. But with Liouville's constant you can upper-bound the remaining error.