Hacker News Reader

Comment by cultofmetatron | Hacker News Reader

I think the ethics here get complicated. for me the line would be if the AI itself was trained on actual CSAM. as long as no one was sexually violated in the course of creating the final image, I see no problem with it from an ethical perspective; all the better if it keeps potential predators from acting on real children. Wether it does or not is a complex topic that I won't claim to have any kind of qualifications to address.

View on HN

> all the better if it keeps potential predators from acting on real children.

The big question is if, those pictures could have the opposite effect.

If there is no proof there should be no ban. What if parent is right (more widespread porn caused people to have less sex after all) ?

This means that a ban caused more harm on real children.

That's a valid and interesting question to ask and study, but I don't think it's relevant to the decision of whether it should be illegal.

It is incredibly relevant. If murder is prevented by having people play violent games and live out their fantasy there, isn’t that a good thing?

I’m not convinced that it would be, but it’s an interesting hypothesis.

The comment I replied to was proposing the opposite equivalent, that fake CSAM (written fiction, AI generated images not trained on real CSAM) could increase risk of action.

I don't think violent video games should be banned, whether they increase or decrease IRL violence (I personally suspect they don't have a significant effect either way). And I don't think "simulated CSAM" (where no actual minors were involved in any part of the creation) should be banned on that basis either (though I don't know enough to guess whether it would tend to increase or decrease actual violations).

I think that's the most, if not only relevant part to base your decision on

And the followup big question is — how do you measure which effect, if any, occurs in practice?

So do you believe violent video games induce more violent crimes then?

The issue is a fair bit subtler than that. The analogous question here isn't "do violent video games induce violent behaviour in the general population?" but rather "do violent video games induce violent behaviour in people who already have a propensity for violence?"

Or, even more specifically, "does incredibly realistic-looking violence in video games induce violent behaviour in people who already have a propensity for violence?". I'm not talking about the graphics being photorealistic enough or anything, I mean that, in games, the actual actions, the violence itself is extremely over the top. At least to me, it rarely registers as real violence at all, because it's so stylised. Real-world aggression looks nothing like that, it's much more contained.

Yep. It can definately go both ways. A game like Doom can be a nice way to put off some steam.

IIRC, violent crime is increased in people pre-disposed to it when they use outlets and substitutes (consuming violent media, etc). That might not translate to pedophilia, but my prior would be that such content existing does cause more CSA to happen.

That's incorrect. There have been studies on this. In a few cases seeing depictions of violence causes an urge to act violently, but in the majority of people predisposed to violence it causes a reduction in that impulse, so on average there's a reduction.

The same has been shown to be the case with depictions of sexual abuse. For some it leads the person to go out and do it. For the majority of those predisposed to be sexual predators it "satisfies" them, and they end up causing less harm.

Presumably the same applies to pedophiles. I remember reading a study on this that suggested this to be the case, but the sample size was small so the statistical significance was weak.

This review [0] is a bit reductionist and overconfident with some of its adjacent claims, but it includes a decent overview of the studies we've done on the topic and references those for further reading. The effect is weak enough at a societal level that it mostly doesn't make sense to consider (and those effect directions are not supportive of your claim of overall reduction if you want to interpret them as strong enough to matter), but when restricted to groups pre-disposed to violence you do see a meaningful increase in violent behaviors.

[0] https://www.mdpi.com/2075-4698/3/4/491

Reply on news.ycombinator.com