What I want to know is why this is so hard to predict? I agree with the conclusion, I've seen power law differences in performance in athletics, in work, and in other domains many times.
I've interviewed and hired a lot of people in my day, and it's an enduring source of frustration to me that discerning who is going to be vastly better than who is a genuinely hard problem that you can't necessarily predict in an interview.
I've read Cowen's Talent, and he definitely has some good ideas in there, and I do try to use more orthogonal interview questions and evaluate the drive for self improvement in their lives, but signal to noise is still *high.*
Internships are about the best you can do. Anyone else have any tips / ideas?
I’ve worked in several public bureaucracies and what you say about productivity is true BUT there are reasons to keep the less productive majority around:
1 Workloads are cyclical and sometimes you need all hands on deck
2 Expertise is lumpy and often the most useless colleague can do a small amount of specific tasks very well
Also there are huge internal and external incentives to keep pay very flat so you get adverse selection into public service by lazy people and into the private sector by those who are useful.
So a normal (Gaussian) distribution is mainly a series of additive factors without feedback. Say a series of many variable genes that are added together to generate a phenotype.
Whereas a "power law" (Pareto) function applies where the function includes a multiplication factor plus feedback. For example, a good poker player or stock investor who combines a higher winning percentage with reinvestment of winnings. Basically, it's like the effect of compound interest.
But it also seems like one would have to be careful to figure out what is causing the positive multiplicative "feedback" effect. It's likely to be a function of opportunity or interest rather than pure ability. For example, actor's success is probably a power law because once an actor is in a big movie, he or she is likely to get offered more roles in other bigger movies. The more movies they are in, the more they get offered. So the profession would get segregated into legions of unemployed unknowns and a few A-list movie stars. But that wouldn't mean that only the "stars" have acting ability.
> In the Swedish case, 61% of violent crime was due to 1% of the population. If this relatively small group of 100k people were effectively prevented from causing problems by means of confinement or strict monitoring, violent crime in Sweden would decrease by over 50%.
This doesn't follow. It assumes that the remaining population of Sweden won't respond to the sudden lack of competition in violent crime by increasing their own output.
> In terms of personnel, it would seem one could cut 50% of the staff and not see much of a decline in the publications. Arguably, this is not because the non-publishing people help out the others, because if they did, their names would be on the papers too.
> Job performance in many sectors is hard to measure objectively, but sports provides a good case: Most points are scored by a small minority of star players.
I think we've gone off track from the argument that individual productivity is independent of the rest of the staff. If you cut the non-point-scoring players from a basketball team, the point-scoring players will instantly stop scoring points.
Even if you just replace the non-scorers with random guys you pulled off the street, you'll see productivity among the guys you didn't replace crater.
What I want to know is why this is so hard to predict? I agree with the conclusion, I've seen power law differences in performance in athletics, in work, and in other domains many times.
I've interviewed and hired a lot of people in my day, and it's an enduring source of frustration to me that discerning who is going to be vastly better than who is a genuinely hard problem that you can't necessarily predict in an interview.
I've read Cowen's Talent, and he definitely has some good ideas in there, and I do try to use more orthogonal interview questions and evaluate the drive for self improvement in their lives, but signal to noise is still *high.*
Internships are about the best you can do. Anyone else have any tips / ideas?
I’ve worked in several public bureaucracies and what you say about productivity is true BUT there are reasons to keep the less productive majority around:
1 Workloads are cyclical and sometimes you need all hands on deck
2 Expertise is lumpy and often the most useless colleague can do a small amount of specific tasks very well
Also there are huge internal and external incentives to keep pay very flat so you get adverse selection into public service by lazy people and into the private sector by those who are useful.
Thanks for the analysis of distribution statistics.
So a normal (Gaussian) distribution is mainly a series of additive factors without feedback. Say a series of many variable genes that are added together to generate a phenotype.
Whereas a "power law" (Pareto) function applies where the function includes a multiplication factor plus feedback. For example, a good poker player or stock investor who combines a higher winning percentage with reinvestment of winnings. Basically, it's like the effect of compound interest.
But it also seems like one would have to be careful to figure out what is causing the positive multiplicative "feedback" effect. It's likely to be a function of opportunity or interest rather than pure ability. For example, actor's success is probably a power law because once an actor is in a big movie, he or she is likely to get offered more roles in other bigger movies. The more movies they are in, the more they get offered. So the profession would get segregated into legions of unemployed unknowns and a few A-list movie stars. But that wouldn't mean that only the "stars" have acting ability.
> In the Swedish case, 61% of violent crime was due to 1% of the population. If this relatively small group of 100k people were effectively prevented from causing problems by means of confinement or strict monitoring, violent crime in Sweden would decrease by over 50%.
This doesn't follow. It assumes that the remaining population of Sweden won't respond to the sudden lack of competition in violent crime by increasing their own output.
The supply of criminals is pretty inelastic though.
Lots of observational and experimental evidence on this.
> In terms of personnel, it would seem one could cut 50% of the staff and not see much of a decline in the publications. Arguably, this is not because the non-publishing people help out the others, because if they did, their names would be on the papers too.
> Job performance in many sectors is hard to measure objectively, but sports provides a good case: Most points are scored by a small minority of star players.
I think we've gone off track from the argument that individual productivity is independent of the rest of the staff. If you cut the non-point-scoring players from a basketball team, the point-scoring players will instantly stop scoring points.
Even if you just replace the non-scorers with random guys you pulled off the street, you'll see productivity among the guys you didn't replace crater.