[joelegner]

Another interpretation is that the portraits communicate essentially zero information about the subject. No information means no way to differentiate beyond the absolute basic facts: male, age such and such, seems to be wearing western clothes, things like that. No information means no way to shade probabilities. You might as well flip a coin.

Or this way: Serial killers look like random people, just as programming language inventors look like random people.

[neilv]

If that were the goal, wouldn't a diversity of obviously nigh-saintly people in the mix... be better than focusing on a historically maligned group?

[TylerLives]

Facial asymmetry, testosterone (thick eyebrows, baldness), weight, piercings, tattoos, unnatural hair colors etc. surely have some correlation with being a serial killer?

[Jensson]

Also to being a coder. Coders and serial killers are both very male groups, so likely share many of the same signals.

[RugnirViking]

do the research and report back. I'd be surprised by "unnatural hair colors" personally, but "surely" isnt the way for that sort of thought

[psychoslave]

Any measurable characteristic will allow to establish some correlation, but it doesn’t necessarily will be a high correlation.