Scott Page’s The Difference

A lot of people think of calls for diversity as fuzzy headed liberalism at its worst. If you’re one of them, please keep reading. Or you could click here and just buy Scott Page’s book and read that, which is what I’d like to convince you to do.

This is a book about problem solving. He starts with a set of observations about how we see the world, and how different people bring different approaches and perspectives to the same problem. His approach is mathematically grounded, although you can skip the math or delve into it. He talks about how bringing different perspectives, heuristics, interpretations and predictive models to a problem can result in super-addative results, as one person helps another overcome blockers.

From there, he looks at how groups compare to experts, and looks at those situations where a group will do better than an expert, even when no member of the group is as sophisticated or broad as the expert. He also looks at those places where averaging over the crowd can get you better results–that if the perspectives are different (and relevant) then a crowd may well have a more intricate model than any one expert.

He also talks about differences between instrumental and fundamental preferences. (We should walk to the park, we should bike to the park, versus we should go to the park or the movies) and how diversity in the latter doesn’t always lead to better results.

He doesn’t make the point that such fundamental diversity of preferences should lead us to prefer liberty. I’m somewhat surprised by this, because it ties to his main points so well. If we want very different things, then we gain a lot by allowing people to make their own choices: some good, some bad, but reducing coordination costs.

It’s been a fascinating read, and I think it will have substantial long-term impact on my thinking. Thanks to Jon Pincus for the pointer. Also, I’ve decided to experiment a bit with Amazon affiliate links, and wanted to disclose that before Threat Level got revenge.

3 thoughts on “Scott Page’s The Difference

  1. Another effect of diversity is illustrated by the MIT effect.
    MIT became the leader in computing for one singular reason, Harvard, Yale, Cornell and the rest of the Ivy lead all imposed quotas on Jewish students, most did not hire Jewish faculty.
    If you assume that talent in computing is evenly distributed amongst the population, that 10% of the population belong to minorities targetted by the discriminatory policies, that the discriminatory policies extend to 95% of the institutions and that talent follows a normal distribution, the 5% of institutions that reject discrimination are going to attract double their share of the discriminated-against talent.
    MIT’s transformation from trade school to first rank powerhouse was largely due to Alfred Lee Loomis. Loomis was appauled by anti-semitism long before WWII and he was instrumental in bringing many of the founders of the Manhattan project to the USA. Loomis was responsible for establishing MIT at the forefront of the US war effort.
    Once you have a nucleus of double the normal amount of exceptional talent it is not difficult to attract more. But the effect is more marked in science and engineering because 1) the measurement criteria are considerably more objective than in the arts 2) emerging fields tend to attract more people from minority groups.
    This effect is possibly more pronounced in commerce and engineering than science: the victims of discrimination tend to think about changing the world, the beneficiaries are more likely to be content to passively observe which is the function of science and the arts.

  2. Yes, you should read Tuxedo Park for the history of Loomis and MIT.
    The bit about a normal distribution is just simple math. I worked it out with Excel. Perhaps I should write a paper or something.
    The anti-semitic schools are choosing the best 100 students out of 900, the equal opportunity schools are choosing the best 100 students out of 1000. That alone gives them an edge, their recruits will be 1.28 standard deviations from the mean or better while the discriminating university is 1.22 sds from the mean or better.
    The effect was more pronounced in the 50s when MIT was one of the very few equal opportunity top rank schools. It was the only game in town. Under those circumstances an equal opportunity recruiter will recruit an almost entirely minority faculty that is 1.64 standard deviations from the mean or better.
    Now Jews are much less than 10% of the population but Harvard and the rest of the Ivy league were discriminating on more than just one front. normsinv (0.09) is a lot less than normsinv (0.05).

Comments are closed.