Douglas_Knight comments on The Amanda Knox Test: How an Hour on the Internet Beats a Year in the Courtroom - Less Wrong

42 Post author: komponisto 13 December 2009 04:16AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (632)
Tags: applied bayesian rationality crime current_affairs evpsych practical psychology

Comments (632)

You are viewing a single comment's thread. Show more comments above.

Comment author: gwern 28 December 2009 12:53:24AM *  9 points [-]

It's clearly absurd to say that pretty girls never murder people. But allowing for the normal hyperbole and inexactitude of conversational English, I don't think that's what SforSingularity means, rather, 'pretty girls are one of the demographic least likely to be responsible for a brutal murder'.

This isn't too unreasonable.

  • First off, the number of murders so ascribable are small: females make up half the population and if we limit pretty to the top 5% or so (a reasonable guess at % for 'pretty enough that a guy will actively note and think "pretty!"'), we're already down to less than 2.5% of the population.
  • Second, women in general commit far fewer violent crimes than men. http://www.ncjrs.gov/html/ojjdp/jjbul2002_10_1/page4.html mentions that for juveniles, at one point, the male:female ratio was 22:1. Let's be conservative and put the young adult ratio at 5:1; now we're down to 0.4%.
  • Third, attractiveness is correlated with IQ, and IQ is well-known to correlate with lower crime rates. (see Wikipedia for a few links; IIRC, WP understates the case but I can't be fashed to dig up the stronger correlations). Let's cut another 10% off the rate, down to 0.36%.
  • Fourthly, attractiveness correlates to higher socioeconomic status through multiple mechanisms, which cuts down violent crime even further. (I don't think I need to adduce any citations for that!)

And so on. I've missed many factors (eg. maybe happier & less stressed people are more attractive, and that too is correlated to less propensity for murder; maybe pretty women have fewer violence-inducing mental diseases and substance abuse habits; etc.). But I've already knocked their murder rate way down. I suspect it has much further to go in a true reckoning.

Does his comment still look satirical to you?

Comment author: Douglas_Knight 28 December 2009 09:22:41AM 3 points [-]

First off, the number of murders so ascribable are small: females make up half the population and if we limit pretty to the top 5%..., we're already down to less than 2.5% of the population.

That's like saying very few murders are committed by people named Amanda. That's OK, if you're very careful, but you have to, at least, weigh it against the very few murders by people named Rudy. 2-4 are OK, though.

Comment author: gwern 03 April 2013 06:00:51PM 0 points [-]

Reference class tennis. Being an attractive female with a (formerly) well-off family is far more important a reference class than being named Amanda; and the corresponding reference class for Rudy would be being an unattractive man who is a poor African immigrant.

Comment author: Douglas_Knight 03 April 2013 11:44:24PM 1 point [-]

Reread your comment. What is the point of your your first bullet point that pretty girls are rare? That tells you nothing, just as the fact that Amandas are rare tells you nothing. Points 2-4 about the relative propensity to murder are relevant. But I'm explicitly talking about point 1 in isolation.

Comment author: gwern 03 April 2013 11:53:42PM 0 points [-]

What is the point of your your first bullet point that pretty girls are rare? That tells you nothing, just as the fact that Amandas are rare tells you nothing.

Yes, it does. If there is any sort of inverse quasi-linear relationship between prettiness and propensity to murder, as one would expect, we would expect the reduction in murder rates compared to the average to be the largest at the extremes - that is, for rarely pretty girls we will expect rarely large effects.

Comment author: Douglas_Knight 04 April 2013 12:18:20AM 0 points [-]

That's not how you used it in your post. Seriously, just read your post. What do these numbers mean?

Comment author: gwern 04 April 2013 01:37:40AM 0 points [-]

Seriously, just read your post.

I rather think I did, since I wrote it.

What do these numbers mean?

It means that we're talking about an extreme part of the population.

Comment author: ESRogs 04 April 2013 03:12:22AM 1 point [-]

My own reading of the bullet points in the post is something like this:

1) Group X is a small fraction of the population.

2) Reason A why group X is disproportionately unlikely to commit murder.

3) Reason B why group X is disproportionately unlikely to commit murder.

4) Reason C why group X is disproportionately unlikely to commit murder.

In the great-grandparent comment above you list an additional reason why pretty girls would be disproportionately unlikely to commit murder, but that wasn't clear at least to me from the original post. So, I agree with Douglas_Knight that bullet point 1 seems to serve a different purpose from points 2 through 4.

Powered by Reddit Powered by Reddit