A group of faculty at the University of Pennsylvania have published a vision for the future of the institution, in the form of a proposed constitution. The document has a large number of signatories, some of whom have chosen to remain anonymous, but whose affiliations have been verified before posting.
First, it seems like the bar of evidence that should be met to justify racial preferences should be high. In a situation where there was nothing to be gained or lost from differential treatment of racial groups, it wouldn't be neutral to discriminate, it would be actively wrong. I take from that that the burden of proof should rest pretty strongly with those in favor of discrimination.
There are some suggestive pieces of evidence about the interaction between diversity and merit, some of which you highlight here. But the evidence seems kind of spotty and disunified, and I'm not sure how it stacks up compared with the very reliable correlations one can find between standardized test scores and objective-looking measures of performance.
Second, it's somewhat unusual for college students to do group activities, and when they do the goal is normally in service of individual learning. I think there's a stronger case for diverse working groups in research or industry where the actual output is the thing that matters. Also, the Page model is about diversity of methods rather than demographic categories.
On the issue of whether to focus on merit vs value-added to more average students, there's a question about impacts here that needs more systematic investigation. In some fields it seems likely to me that the world benefits more from the very top tail-end of the distribution of skills... like training one Terry Tao or Einstein is much more impactful than training a thousand average mathematicians or physicists. In these fields, admissions lotteries could limit opportunities to develop skills for some of the students with the greatest potential, which would be a great loss. How many fields are like this? I'm not sure but it could be a lot of them.
There's also the question of where to draw the line. I think all can agree that eg the bottom quartile of US high school students wouldn't benefit much from top-flight university education in the present day. A much better argument can be made for lottery-type admissions at the primary and secondary ed level. But once kids are college-age, my impression is that a lot of general cognitive skills are already "baked in" and hard to train up further. Maybe a "top five percent" system instead of top ten would be ideal.
The points you're making are the right kind of considerations to think about, though. I agree that the "pure merit" planks of this platform are too controversial to belong on a university's constitution.
Many good points here Dave, thanks for taking the time.
Regarding the issue of group activities, I don't think colleges are concerned primarily with how students do while enrolled, they are much more concerned with what happens later (your examples of Terence Tao and Einstein illustrate). Here's a bit from my paper in JEL (link in post):
"While it is important to measure and quantify performance in relation to institutional mission, this is especially difficult to do for accomplishments made long after graduation—the development of vaccines, the invention of products, the founding of companies, the winning of awards and honors, the leadership of foundations, election to high office, contributions to human rights, and so on. Relatively quiet activities such as the provision of medical services in under-served communities must be weighed against highly visible ones, such as appointment to the Supreme Court. As if this problem were not challenging enough, one then needs to identify the kinds of traits in the applicant pool that are predictive of mission-aligned success."
Will try to respond to the rest later (bit short of time).
Yes, the diversity in the Page model is in the space of "cognitive toolboxes" and I think he is quite clear about this. People do sometimes interpret the research as saying more than this, for example to justify demographic diversity, but the link between the two is tenuous.
Durlauf has a very good 2008 paper (cited and extensively discussed in our JEL paper) on these issues. Regarding the claim that ethnic diversity has major educational benefits, he says: "Given the frequency with which the claim is made, it is remarkable how poor the evidentiary support is for the proposition." I agree with this. The reason why the educational benefits of diversity are such a big part of our discourse is because that's the only door left open by Bakke. As Lee Bollinger put it in 2023 paper, this diversity-based justification was a “fragile reed” from which the constitutionality of affirmative action was hanging.
Yeah, and in principle I'm on board with the reparations and relational-egalitarian justifications for affirmative action... I don't have a problem with the idea that the moral justification for it could be different from the legal justification it technically rests on. I'm concerned though with the public choice implications of permitting it, in that I don't have a lot of trust that it'll be implemented in a reasonable way by the sort of administrators who are given charge of it in a university setting. Their track record has been pretty poor overall as evidenced by the treatment of Asians at Harvard.
But then once we bring in practicalities like that, it also must be said that laws against it are not easy to enforce in practice and can lead to more holistic processes that are harder to trust. It is really hard to say what to do about the whole issue IMO.
Don't make your judgement based on this comment alone. The Hong and Page paper contains an important insight, even if one can argue with their interpretation or believe that the result is oversold. There's a lot of research now (see Woolley and others) that reaches similar results. But this has nothing to do with racial or gender diversity, it is diversity in cognitive toolboxes. And Scott has been quite clear about that.
He's been sort of clear about it. The article has been cited a bazillion times to justify ethnic/gender diversity, and the phrasing in his book seemed aimed at teasing this preferred viral interpretation while maintaining deniability. (The book itself contained much food for thought, although I thought there was a tendency to smuggle in critical assumptions as mere asides or dependent clauses in some of the exposition.) I also read some very interesting old comments on the whole dispute over at Why Evolution is True. It is definitely a strange definition of "expertise" or "ability" to suppose that it doesn't include picking good starting points in search space.
As someone who's written about how game theory models can be constructed to rationalize any outcome--say, bank presidents setting their pants on fire--I am sensitive to Thompson's critique. She is quite right that the use of the term "diversity" in this context was arbitrary, and that other plausible modeling assumptions would give different results.
I think diversity of heuristics makes sense, but I agree that the research can be and has been interpreted to support policies that it does not speak to.
These issues are complex. A couple of points.
First, it seems like the bar of evidence that should be met to justify racial preferences should be high. In a situation where there was nothing to be gained or lost from differential treatment of racial groups, it wouldn't be neutral to discriminate, it would be actively wrong. I take from that that the burden of proof should rest pretty strongly with those in favor of discrimination.
There are some suggestive pieces of evidence about the interaction between diversity and merit, some of which you highlight here. But the evidence seems kind of spotty and disunified, and I'm not sure how it stacks up compared with the very reliable correlations one can find between standardized test scores and objective-looking measures of performance.
Second, it's somewhat unusual for college students to do group activities, and when they do the goal is normally in service of individual learning. I think there's a stronger case for diverse working groups in research or industry where the actual output is the thing that matters. Also, the Page model is about diversity of methods rather than demographic categories.
On the issue of whether to focus on merit vs value-added to more average students, there's a question about impacts here that needs more systematic investigation. In some fields it seems likely to me that the world benefits more from the very top tail-end of the distribution of skills... like training one Terry Tao or Einstein is much more impactful than training a thousand average mathematicians or physicists. In these fields, admissions lotteries could limit opportunities to develop skills for some of the students with the greatest potential, which would be a great loss. How many fields are like this? I'm not sure but it could be a lot of them.
There's also the question of where to draw the line. I think all can agree that eg the bottom quartile of US high school students wouldn't benefit much from top-flight university education in the present day. A much better argument can be made for lottery-type admissions at the primary and secondary ed level. But once kids are college-age, my impression is that a lot of general cognitive skills are already "baked in" and hard to train up further. Maybe a "top five percent" system instead of top ten would be ideal.
The points you're making are the right kind of considerations to think about, though. I agree that the "pure merit" planks of this platform are too controversial to belong on a university's constitution.
Many good points here Dave, thanks for taking the time.
Regarding the issue of group activities, I don't think colleges are concerned primarily with how students do while enrolled, they are much more concerned with what happens later (your examples of Terence Tao and Einstein illustrate). Here's a bit from my paper in JEL (link in post):
"While it is important to measure and quantify performance in relation to institutional mission, this is especially difficult to do for accomplishments made long after graduation—the development of vaccines, the invention of products, the founding of companies, the winning of awards and honors, the leadership of foundations, election to high office, contributions to human rights, and so on. Relatively quiet activities such as the provision of medical services in under-served communities must be weighed against highly visible ones, such as appointment to the Supreme Court. As if this problem were not challenging enough, one then needs to identify the kinds of traits in the applicant pool that are predictive of mission-aligned success."
Will try to respond to the rest later (bit short of time).
A few more quick comments.
Yes, the diversity in the Page model is in the space of "cognitive toolboxes" and I think he is quite clear about this. People do sometimes interpret the research as saying more than this, for example to justify demographic diversity, but the link between the two is tenuous.
Durlauf has a very good 2008 paper (cited and extensively discussed in our JEL paper) on these issues. Regarding the claim that ethnic diversity has major educational benefits, he says: "Given the frequency with which the claim is made, it is remarkable how poor the evidentiary support is for the proposition." I agree with this. The reason why the educational benefits of diversity are such a big part of our discourse is because that's the only door left open by Bakke. As Lee Bollinger put it in 2023 paper, this diversity-based justification was a “fragile reed” from which the constitutionality of affirmative action was hanging.
Thanks again for your comment.
Yeah, and in principle I'm on board with the reparations and relational-egalitarian justifications for affirmative action... I don't have a problem with the idea that the moral justification for it could be different from the legal justification it technically rests on. I'm concerned though with the public choice implications of permitting it, in that I don't have a lot of trust that it'll be implemented in a reasonable way by the sort of administrators who are given charge of it in a university setting. Their track record has been pretty poor overall as evidenced by the treatment of Asians at Harvard.
But then once we bring in practicalities like that, it also must be said that laws against it are not easy to enforce in practice and can lead to more holistic processes that are harder to trust. It is really hard to say what to do about the whole issue IMO.
Just as a side note, quoting Scott Page about diversity being more important than merit may not be the best idea:
https://www.ams.org/journals/notices/201409/rnoti-p1024.pdf
Don't make your judgement based on this comment alone. The Hong and Page paper contains an important insight, even if one can argue with their interpretation or believe that the result is oversold. There's a lot of research now (see Woolley and others) that reaches similar results. But this has nothing to do with racial or gender diversity, it is diversity in cognitive toolboxes. And Scott has been quite clear about that.
He's been sort of clear about it. The article has been cited a bazillion times to justify ethnic/gender diversity, and the phrasing in his book seemed aimed at teasing this preferred viral interpretation while maintaining deniability. (The book itself contained much food for thought, although I thought there was a tendency to smuggle in critical assumptions as mere asides or dependent clauses in some of the exposition.) I also read some very interesting old comments on the whole dispute over at Why Evolution is True. It is definitely a strange definition of "expertise" or "ability" to suppose that it doesn't include picking good starting points in search space.
As someone who's written about how game theory models can be constructed to rationalize any outcome--say, bank presidents setting their pants on fire--I am sensitive to Thompson's critique. She is quite right that the use of the term "diversity" in this context was arbitrary, and that other plausible modeling assumptions would give different results.
I think diversity of heuristics makes sense, but I agree that the research can be and has been interpreted to support policies that it does not speak to.