Hiring and judging character
the connection between hiring people for a job and judging the character of people fascinates me (e.g. for finding good friends, or for romantic relationships).
both are clearly important: you need to hire people to get tasks done, and you want to find good friends/partner(s)
both can take quite a bit of commitment to accurately judge
both can end disastrously (due diligence)
in both scenarios, there can be multiple people doing the judging, so there can be some gossip among those “in the know”
sometimes it can be really hard to tell the difference between someone who is an extremely good fit, and someone who is a terrible fit.
How do you distinguish 10X software engineers from the rest?
Yishan Wong:
Look to referrals of referrals. Look for the best person according to the best person according to the best person you know (yes, that’s two levels). If there is consensus among several of your best people, this is a fairly reliable indicator. It is difficult to go more than 3x per “link” though - people are typically unable to discern the difference between someone 9x as good as they are from someone who is 3x as good as they are; this is why you have to do the “best person you know” referral twice. First, you hire all the best people you know, work with them for awhile to find the best amongst them, and then find all the best people they know. It is a multiple-stage approach; many people don’t know this, and just assume that “refer the best people you know” is the end of it.
Similarly, stalking someone’s friends might be useful if you want to determine their character.
Similar is a post by Eliezer Yudkowsky:
People can’t distinguish better or worse within more than one standard deviation above their own level. I can’t either: if there were two superintelligences or just >Eliezers arguing with each other, to me they would both sound reasonable. (The “plus one standard deviation” part isn’t a standard result, just me trying to eyeball the phenomenon.)
Also similar is a post by Caroline Frances Hubert. In comments Pasha Kamyshev writes:
There are only a few reliable signals that someone can attempt to use. A lot of verbal signals can be easy to fake. The strongest signal that can’t be faked is the ability to predict the future. Another good one is the ability to pass the “Ideological Turing Test” and explain both sides of an ideological argument. Another one is your general inability to predict individual actions of the person, while still roughly seeing the end-result outcome as positive.
See also “He Who Pays the Piper Must Know the Tune” by Robin Hanson.
Also interesting is the IQ gap reported in leader–follower relationships.
From Handbook of Organizational and Managerial Wisdom (p 310):
The modest relationship of leadership and intelligence may occur because it might not be intelligence per se that matters in leadership; instead, it might be the relative intelligence of leaders to their followers (Bass, 1990; Stogdill, 1948). Researchers have suggested that leaders can be most successful when they are slightly more intelligent, but not too much more intelligent, than their followers. As Gibb (1969) put it, “The evidence suggests that every increment of intelligence means wiser government, but that the crowd prefers to be ill-governed by people it can understand” (p. 218). This proposition was supported by Simonton (1985), who proposed a curvilinear relationship between intelligence and a person’s influence over other group members (an indication of leadership), such that there would be a high correlation between the group mean IQ and the IQ of its most influential member, with a leader–follower IQ gap of between 8 and 20 points, depending on organizational level, and with smaller leader–follower gaps at more senior levels of an organization’s hierarchy. Thus intelligence may set the stage for leadership, but many other factors are involved in successful leadership as well.
many more Quora questions about judging character (many by Alex):
How to evaluate someone’s ability to be an exceptional judge of character
decent answer: Sridatta Thatipamala’s answer to How can you become a good judge of character?
How do I judge the character of a lady/girl?
- observing ppl under stress might be important
one important question: does the person have the same life goals as me?
“Are you at peace with your past romantic partners? Do you feel at peace with their joys though the joys may be mutually shared no longer?” (source)
Yishan Wong(source):
the lesson here is that sometimes “10x people” are so good that they don’t really look for jobs in the same way other people do, so you kind of have to find them via unorthodox means […] [Y]ou often have to make your job posting indicate a bit of quirkiness in just the right way so that it happens to stick out to them when they’re casual perusing job ads
[…]
the main idea is to visualize the personality and character of the type of people you want to hire, and then put a message out publicly that only those types of people will respond to or provide a correct answer to.
GiveWell has The process of hiring our first cause-specific Program Officer. Talking about senior hiring, they list some general principles:
- The best way to evaluate someone is to work with them. The ideal way to make a senior hire would be to have a longstanding relationship with someone, perhaps as a part-time consulting arrangement.
- Interviews are highly unreliable. The evaluation process should be designed as much as possible to mimic working together.
- References are extremely important, and it can be useful to talk to many people about the candidate (including people who weren’t specifically offered as references).
Also:
When a candidate didn’t mention a major aspect of the criminal justice reform field, we would ask about it and see whether they were omitting it because they (a) had strong knowledge of it and were making a considered decision to de-prioritize this aspect of the field; (b) didn’t have much experience or knowledge of this area of the field.
(So the general lesson here might be “do your homework” before trying to hire people.)
From “Projects, People and Processes”:
I generally consider it very hard to evaluate people, and don’t know of any reliable and reasonably quick way to do so. Our experiences recruiting have generally left us feeling that the only good way to evaluate someone is to work with them for an extended period of time, and we’ve heard similar sentiments when seeking advice from other organizations.
From “The Most I’ll Admit”:
My ability to discern human nobility is markedly worse than I thought in 2015. I’ve probably always been this bad, but 2016 helped me see my limitations clearly.
https://www.facebook.com/satvik.beri/posts/610742700734
https://www.openphilanthropy.org/blog/reflections-our-2018-generalist-research-analyst-recruiting
https://www.lesswrong.com/posts/HCssSWMp7zoJAPtcR/you-are-not-hiring-the-top-1
https://80000hours.org/2019/08/how-replaceable-are-top-candidates-in-large-hiring-rounds/
Something I find unfortunate is that most people require an overwhelming amount of evidence before recognizing someone’s greatness. The more time one has to think and the more one interacts with someone (or just reads their content), the easier it is to tell their greatness. But it seems like someone people have the skill/ability to recognize greatness given a very small amount of evidence. The rest of humanity requires such an overwhelming amount of evidence of greatness that it becomes impossible to ignore it (and such evidence often only becomes available once the person in question has died). (See Eliezer Yudkowsky’s posts about Solomonoff induction and the difference between science and rationality for related ideas. “I sometimes say that the goal of science is to amass such an enormous mountain of evidence that not even scientists can ignore it” (source).)
On false positives and false negatives: if people are very careful to only “approve” of people who are very similar to themselves or have concrete accomplishments, there will be lots of false negatives but few false positives. If people can be tricked by confidence/credentials/social status given to a person by others, then we can expect lots of false positives. I think many “great minds” are great in somewhat unique ways, which makes it pretty hard to detect their greatness (if one is using some sort of similarity metric to judge how great someone is). Here’s like a worst-case scenario in terms of judging someone: the person is great in a really important but difficult-to-detect way, they don’t like to accumulate credentials, they haven’t been detected by other people (so there’s no social status floating around), they don’t signal confidence, and they aren’t similar to other great people who have been discovered so far. The chance of a random person being like this is very low, but I do wonder how many people like this we miss out on because our judgment mechanisms aren’t good enough (i.e. have too many false negatives).
Eliezer once left a comment on one of Scott Sumner’s posts about how he (Eliezer) was able to identify Sumner as a “correct contrarian”, despite at first not having much subject knowledge. It was something about the rhythm of the debate and the kinds of rhetoric Sumner used.