In a previous post, I criticized George Sher’s view that merit-based desert is based on (the recognition of) existing conventions of merit. In these cases, the existing rules are already fashioned to reward merit in a justified way, so that justice (in the sense of rewarding desert) consists simply in acknowledging that a given person satisfies the criteria of merit, and acknowledging that in accepting the convention, we accept the further implication that the person deserves what the rules say they deserve.
In cases of this kind, claims of desert are parasitic both on conventional criteria of merit, and on obligations of veracity about, and fidelity to, those criteria. I probably should have been clearer than I was in my last post that the obligations of fidelity and veracity are themselves non-convention-based obligations to respect (certain existing) conventions. So Sher’s argument is not a matter of “convention all the way down,” and need not collapse into relativism; it involves an objective duty to respect conventions, not a conventional duty to respect them. Sher implicitly takes the convention-based argument to cover a fair number of cases. One disagreement I have with him concerns his optimism on that score. It’s not clear to me how well-grounded our conventions really are. (See my last post for that.)
In any case, in this post, I want to take on Sher’s Kantian argument intended to deal with cases where a convention-based conception of merit doesn’t apply. Convention-based conceptions of merit can, Sher reasonably acknowledges, get the appropriate criteria of merit wrong. Where they do, we need an account of merit-based desert that isn’t based on the requirements of mere convention. We need, in another words, a principle, not itself grounded in convention, that regulates existing conventions in such a way as to reward objective merit. For a variety of reasons, obligations of veracity and fidelity won’t do the job, so we need something else. The principle Sher defends is what might be called the principle of optimal selection.
The principle of optimal selection governs cases where a candidate is being selected from among a set of (presumably competing) candidates for a position. In these cases, merit-based desert (defeasibly) requires us, on Kantian grounds, to select the best of the available candidates for the position and, of course, to put him in that position. Though defeasible, Sher takes the principle to have a fair bit of normative force: in most cases, if you hire or select anyone but the best candidate for a position, you’re doing something wrong, or perhaps two things wrong. For one thing, you’re acting in a self-defeating way on instrumental grounds: if you’re aiming at the best outcome, as you should be doing, it’s irrational to choose a partner for your project that’s suboptimal relative to your goal. More to the point, you’re violating one of the requirements of justice, i.e., to give the deserving their rightful due, where the deserving happen to be those best qualified to perform a given task. (A person’s entitlement to their rightful due is to be distinguished from what they have a right to).
Sher’s argument for our obligation to choose the best candidate is, as earlier remarked, Kantian, or at least Kant-influenced, or maybe Kant-influenced by way of Dworkin with some Aristotelian herbs and spices thrown in.
When we hire by merit [i.e., in the sense of selecting the best candidate for a position], we abstract from all facts about the applicants except their ability to perform well at the relevant tasks. By thus concentrating on their ability to perform, we treat them as agents whose purposeful acts can make a difference in the world. Moreover, by concentrating on abilities that are internally connected to jobs and opportunities that exist to serve independent purposes, we affirm the applicants’ involvement in the wider life of the community. For both reasons, selecting by merit is a way of taking seriously the potential agency of both the successful and unsuccessful applicants. Conversely, when an application is selected on some other basis, there is a recognizable if elusive sense in which he and his rivals are not taken seriously. And this suggests that we may justify selection by merit by arguing that persons ought to be taken seriously in the relevant sense (Sher, Desert, pp. 121-22).
The basic Kantian idea here is that of respect for rational agency. The Dworkinian twist is that respect for agency is a matter of taking agency seriously. The Aristotelian element is the somewhat vaguely-expressed idea that agency warrants being taken seriously when it makes a value-oriented difference to the world–when it realizes the good–and in particular, when it makes a positive contribution to the common good of a community.
These ideas, treated as premises, are intended to yield the conclusion that we ought, in any context of candidate-selection, to apply the principle of optimal selection, where “optimal” is (we might suppose) agent-relative rather than agent-neutral. Though Sher does not explicitly say this, I take it that the obligation he recommends is to select the best of the available candidates for a position whether or not doing so leads to the best overall state of affairs, rather than to aim at the best overall, agent-neutral state of affairs and determine that the candidate that causally promotes that state of affairs is by definition the best candidate.
Granted, in most cases, selecting the best candidate will also ipso facto yield the best overall outcome, but where there is a trade-off between choosing the best candidate, and producing the best overall state of affairs, Sher’s principle prioritizes the former over the latter–selecting the best person for the job over realizing the best state of affairs all-in.*
This passage later in the chapter gives a clear illustration of the Kantian flavor of Sher’s claim:
Properly understood, the requirement that we select among applicants on the basis of their qualifications is a consequence of the more general requirement that we treat all persons as rational agents. Because we most readily associate practical reason with deliberation and choice, the general requirement is most often taken to demand that we do not coerce others’ decisions. But practical reason encompasses not only decisions, but also the acts and intended effects to which these decisions lead. Moreover, it ranges over both acts intended to serve agents’ own interests and acts with other aims. Thus, any requirement that we treat persons as rational agents must demand respect not only for their choices, but also their ability to do things which advance their own and others’ ends. Yet…when we select among applicants for reasons other than their ability to perform the tasks that define positions, we treat them as passive recipients of largesse or links in causal chains rather than as active contributors to anyone’s ends. By thus disengaging their practical wills from the aims that have generated the positions, we violate the requirement that they be treated as rational agents (Sher, Desert, p. 126).
Sher’s argument has undeniable power and plausibility, which I don’t mean to dispute or disparage. But I also think it involves some loose ends worth querying, so read the following criticisms accordingly.
Many of my criticisms here mirror the claims I made in the previous post about Sher’s appeal to conventional standards of merit. Even if we accept Sher’s claim that failure to hire the best qualified candidate violates the Kantian strictures he invokes, the concept of “the best qualified candidate” remains a black box. Sher seems to take for granted that the criteria for optimal selection are relatively transparent when it comes to, say, university admissions or hiring in a professional context. In the first case, it means being the best of a set of candidates to learn university-level academic material. In the second case, it means being the best of a set of candidates to perform the tasks relevant to a job or office. But these formulations strike me as vague or indeterminate to the point of meaninglessness. Even if they end up being true as stated, they lack practical application unless supplemented by more in the way of analysis than Sher gives us.
For one thing, contrary to Sher, “merit” is not one thing, but many different things, realizable in many different ways. Given this, the contrast between “hiring by merit” and “hiring by considerations extraneous to merit” is not as sharp or clear as Sher takes it to be. Many indications of merit at t2 are inferred by way of some proxy observed at t1. But it’s often unclear what is supposed to be a proxy for what in such inferences. Sher writes as though currently-existing ability to perform a task, narrowly conceived, was the only qualification for deserving to be hired for a job, as though jobs were themselves to be conceived as sets of tasks mapping onto candidates’ currently-existing abilities. But this is an extremely artificial conception of hiring and job performance. Hiring officers often (justifiably) gamble on the supposition that someone is qualified for a job in the hypothetical sense that if given the opportunity, the candidate might well turn out to be a better choice than another candidate with greater ability. Sher’s account seems to imply that such gambles violate the principle of desert, but it’s unclear why they would. Does actual ability always trump potential ability, even in cases where the potential promises better performance in the long-run than actual? Sher seems prematurely committed to saying “yes.”
Consider a favorite example, especially in polemical contexts–college admissions. It should be obvious to anyone who has taught at a university that universities engage in many different activities, and therefore have many different functions–from teaching calculus to hosting football games to hosting networking events to enabling political activism to serving the corporate community. Sher gives no account of which of these activities is essential to the university, hence essential to the criteria for applying the principle of optimal selection to candidates for admissions.
Given this omission, it’s not clear what it means to evaluate a high school student who has never been to university in terms of her existing ability, at t, for learning what there is to learn at the university. There’s a huge variety of things to learn and master, even at the smallest university; ex hypothesi, the high school student has so far learned very few or none of them. In many cases, she’s had no opportunity to learn anything like what she might encounter at the university.
Sher writes as though merit in this context were something transparent at a given time–as though a high school student at t1 “had” the very merit that she might be expected to display in college at t2. But in reality, it’s not at all clear what this means. “[S]electing by merit,” he writes, “is a way of taking seriously the potential agency of both the successful and unsuccessful applicants” (Sher, Desert, p. 121). The reference to “potential agency” in this claim is unclear. Does it refer to actualized agency at t1 considered as a proxy for the actualization of a qualitatively identical capacity at t2? Or does it refer to unactualized agency at t1 which might, if all goes well, come to be actualized at t2? In other words, does it refer to the prep school math prodigy at t1 who will go on to be a college math prodigy at t2? Or does it refer to the abused child/math incompetent on Pine Ridge Indian Reservation at t1 who might (with proper nurturing) become a college math prodigy at t2? It’s not obvious why the first exhibits greater merit than the second for college admissions.
Sher also ignores the possibility that what there is to learn at a university might change over time, or might be retro-fitted for the specific purpose of attracting a specific demographic to the university. Put another way, there just seems a radical mismatch between the abstract claims of Sher’s argument, and the actual activities engaged in by the average university’s Office of Enrollment Management. This is not a conclusive argument against Sher, but it does exact a price: given the way he puts things, it’s not clear how his account applies in real life, and not clear that Sher sees the gap between the very abstract prescriptions he offers and the ground-level realities where they’ll ultimately apply.
This is a general problem that ranges well beyond university admissions. We inhabit a milieu obsessed with the need to quantify merit, an obsession that arises in part from philosophers’ failure or inability to provide qualitative measures of merit that aren’t quantifiable, but still have application in real-world contexts. It’s all well and good to insist that Kantian respect for agency requires us to hire or admit “the best” candidate from the available alternatives. But even apart from the multiple realizability of forms of “bestness,” we face the problem that no particular conception of optimality is easy to articulate, measure, or operationalize. Yet a conception of optimality that can’t be articulated, measured, or operationalized is useless to us in practice, even if every claim we make about it happens to be truth-apt or even true.
In practice, defenders of “meritocracy” have done their best to reduce such complexities to least common denominator conceptions of merit, ideally expressed in quantitative terms. Thus, academic merit is, in some quarters, thought to be equivalent to an SAT score, or a GPA, or some combination of SAT score and GPA, etc. Conversely, universities’ attempts to do away with such scores are often interpreted ipso facto as an attack on merit as such.**
Contrary to prevailing wisdom, it’s not clear why high school GPA is a good proxy for success in a college major that had no counterpart at the high school. Clearly, we need some argument demonstrating a connection between the one thing and the other. Again, unless we make some distinction between essential and non-essential functions of a university, it’s not clear why high SAT scores have any bearing on “college performance” at all. What does a high SAT score have to do with a student’s ability to excel on the college soccer team? Yet excelling on the soccer team may be the reason why she selected a given college in the first place. Sheer appeals to “bestness” and to Kantian respect for persons cannot resolve these difficulties. We have to get further down in the institutional-level weeds to do so. But Sher doesn’t.
Defenders of “meritocracy” have adopted the same strategy–often with ludicrous results–in professional contexts, as well. Consider a few examples that I happen to know from the inside.
Universities measure academic merit in faculty in part through algorithms regarding publication, and in part through teaching evaluations, themselves broken down into peer observations and student evaluations. The problems with all three methods are widely acknowledged. The people who write algorithms don’t really know anything about academic merit. The people who know a lot about academic merit don’t know how to write algorithms. Peer observation of teaching is highly biased, and easily abused. Student evaluations are worse than a joke. And it’s common knowledge that academic job searches are exercises in horse trading that bring out the worst in just about everyone involved. To watch one up close, or worse still to participate in one, is an exercise in demoralization. It’s fundamentally unclear how we get from this state of affairs to a procedure that tracks objective merit.
When I worked in a hospital OR, we had a whiteboard devoted to recording the day’s “Orthopedic Surgery Efficiency Metrics,” which consisted, as far as I could make out, of a series of essentially pointless and arbitrary figures intended to represent our “merit” as OR workers. Many of these numbers were falsified; almost all of them were nonsensical. And most were forgotten by most of us the next day.
Nonetheless, the data were, with a reverence befitting a holy rite, fed each day into a spreadsheet which was (apparently) studied with utmost care by the administrators tasked with inferring merit and demerit from numerical values. The results were the same no matter what data was fed into the spreadsheet: We weren’t working hard enough or fast enough to meet the OR’s expected revenue goals. We also weren’t working carefully enough to meet the OR’s expected safety goals. And we weren’t making efficient use of the resources at our disposal, either. Yet we weren’t doing badly enough to merit termination, at least not yet. No data was collected on the trade-off between working quickly and working carefully. Nor was data collected on the adequacy of the resources at our disposal. Some things are just givens.
Oddly, despite our many (supposed) failings, one external accreditation agency gave our hospital an “A” rating for efficiency and safety, making us one of the best hospitals in the state. As far as Medicare was concerned, however, we were on a kind of censure list for our failings of efficiency and safety, making us one of the worst hospitals in the state. All in all, the operationalization of merit was a mystery, which led to a lot of skepticism and cynicism about its existence and objectivity.
I currently work in a field, hospital revenue-cycle management, where “data-driven performance ratings” of billers, collectors, coders, payment posters, and the like are a near obsession, essentially on par with the obsession over revenue-maximization that gives the field its raison d’etre. Indeed, a fair bit of what I do each day is to tweak computerized workflow for executives and managers in Patient Financial Services, Hospital Billing, and the like, in order to find the optimal arrangement that gets the “best” out of our employees (themselves presumed to be the “best” of the candidates selected from the searches that hired them), and yields the greatest net amount of revenue for the hospital, itself conceived as the “best” of all possible outcomes.
But aside from being easily operationalizable and translatable into computerized workflow diagrams, it’s unclear (at least to me, after about six months on the job) why the numbers on which corporate managers and supervisors fixate have anything at all to do with merit or desert. I suppose they do if you retro-fit the concepts of merit and desert to fit the operationalizations designed to measure them. But that seems like cheating, doesn’t it?
At this point, a reasonable person might come to question the idea there really is any objective conception of merit available for practical purposes. Perhaps, to express a Hobbesian thought, the best-qualified candidate is not the candidate with some fixed, determinate set of traits and capacities that make him the “best fit” for the job, but the candidate with a capacity for infinite personal malleability. The best qualified candidate, in other words, is the candidate most willing and able to satisfy the expectations of his managers, whatever those expectations may be, however whimsical, malleable, or ephemeral. The meritorious OR janitor is not the one who cleans ORs well by some objective standard of that task, but the one who best pleases his manager by producing the right metrics–however they’re produced. The best academic is the one who does the best job at pleasing the Dean. And so on, the theme of Plato’s Sophist writ large.
Another possibility is that perhaps we’ve come to mis-conceptualize merit altogether, so that some version of Aristotle’s proviso applies:
Our discussion will be adequate if we make things perspicuous enough to accord with the subject-matter; for we would not seek the same degree of exactness in all sorts of arguments alike, any more than in products of different crafts. Now, fine and just things, which political science examines, differ and vary so much as to seem to rest on convention only, not on nature….Each of our claims, then, ought to be accepted in the same way [as claiming to hold good usually]. For the educated person seeks exactness in each area to the extent that the nature of the subject allows; for apparently it is just as mistaken to demand demonstrations from a rhetorician as to accept [merely] persuasive arguments from a mathematician. (Nicomachean Ethics I.3, 1094b20-1095a1).
In other words, maybe our quantitative obsession has to be replaced with something more appropriate to the subject matter. That said, it’s easy to see why a numerical conception of merit has polemical value: numbers are easier to deploy in polemical arguments than qualitative considerations, however conducive or ill-conducive they are to tracking the truth about merit.
At one point, Sher entertains what might seem an obvious objection: is he prescribing a principle of optimal selection or of merely competent selection? In other words, do we have an obligation to hire/admit the best candidate for a position, or simply to hire/admit any competent one? Given the absence of any worked-out account of optimality, it’s not really clear what practical consequence the answer has, but in any case, Sher insists that we have a higher order obligation to seek the best, not the merely adequate. His illustration of the point is both revealing and peculiar:
In general, however, purposive activity aims not merely at achieving satisfactory results, but at achieving the best results that prevailing conditions allow. A weekend gardener may be resigned to his small plot, his poor soil, and his lack of time to pull weeds; and he may therefore be happy to produce tomatoes above a certain number and quality. Nevertheless, his aim is still to grow as many tomatoes, as his resources permit (Sher, Desert, pp. 122-23, emphasis in original).
The assumption here seems to be that the pursuit of the best in this instance requires the gardener to produce the most and best tomatoes from the resources available to him. I don’t see why. Couldn’t the principle of seeking the best just as well require the gardener to make the best use of his time as require him to maximize his tomato output?
Suppose that the gardener plants tomatoes primarily for (broadly) therapeutic reasons. He has, let’s say, suffered some trauma, and is told, in Wordsworthian vein, that it would be a good idea to go out into his garden, plant some tomatoes, and watch them grow. If so, all that matters is that he’s out there gardening and getting the best therapeutic value out of the gardening, not that he’s growing the most or best tomatoes.
Likewise, if it’s best for my health to start running four miles a day, it doesn’t follow that what’s required of me is that I resolve to run the fastest four mile run that I can. I might simply run the most enjoyable four mile I want. Again: if it would be best for me to go back to playing the guitar after several years’ hiatus, perhaps that means that I should make the best (most enjoyable, most fulfilling) use of the time I spend playing the guitar. I need not try to become the best guitar player out there, or even the best player I can become. Nothing about the principle of seeking the best entails that I should sacrifice pure enjoyment to the quest to remedy all of the flaws in my guitar playing–assuming that the best is what we’re obliged to seek.
The video above nicely illustrates the point. In it, Anthrax guitarist Scott Ian comments on the rhythm guitar playing of the late Malcolm Young, rhythm guitarist for AC/DC. At one point in the video, Ian comments on the simplicity of Young’s guitar playing as compared with the sort of rhythm playing that Ian himself does in Anthrax. Yet he says that he gets pleasure out of strumming the simple A-chords associated with Malcolm Young’s playing style–something he says he could do “all day.” Nothing about strumming those A-chords would make Ian a better guitarist, much less the best guitarist he could be. But clearly something about strumming them seems eminently worth doing. This example, and others like it, point to a flaw common both to Sher’s account and to conventional wisdom, at least in the United States: a failure to acknowledge the multiple realizability of “merit” and “bestness,” and a problematic insistence on equating both with productive output as measured by “data points.”
One final set of criticisms applies to the Kantian account Sher invokes to undergird his argument as a whole. A basic problem is that it’s not clear to me how Sher’s intended conclusion follows from any identifiable set of premises. How exactly do we get from the obligation to respect rational agency to the obligation to select the best candidate? Suffice it to say for now that I find Sher’s argument somewhat vague and obscure.
But however we understand it, Sher’s Kantian account doesn’t explain why merit is to be rewarded by external goods. It’s one thing to say that if you are the one with the best ability to do a job, you should be the one selected to do it. It’s another thing, and a puzzling thing, why if you’re the one with the best ability to do a job, you should be paid certain quantities of money and benefits for doing it. Since these rewards are not internally connected to desert, it is possible to argue, as Thomas Nagel does, that there is no strong connection between being the best qualified for a job, and being well-remunerated for having those qualifications.*** Why isn’t getting the job the appropriate reward for having the ability to do it, whatever the implications for compensation?
I don’t see that Sher has an adequate answer to this challenge, but I’ll address it at greater length in my discussion of chapter 8, where it seems to have more bite than it does here. If being paid for merit is problematic, being paid for virtue is even more so. To be continued.
*This is an exception, presumably a minor one, to the claim made earlier in the text that failure to hire the best candidate gets two things wrong. There may well be cases in which it gets just one thing wrong.
**See, for instance, Jason Riley’s “Meritocracy Is Worth Defending,” Wall Street Journal, August 17, 2021, summarizing the arguments of Adrian Wooldridge, The Aristocracy of Talent. For further discussion, see Kwame Anthony Appiah, “The Myth of Meritocracy,” The Guardian (Oct. 19, 2018), and Scott Alexander, “Targeting Meritocracy,” Slate Star Codex (July 24, 2017). Thomas Edsall’s “Conservatives Are Happier than Liberals,” New York Times (Oct. 20, 2021), though on an apparently unrelated topic, throws some of these issues in an interesting light. In a sense, the Ur text of the approach to meritocracy described in the text of the post is Richard Hernnstein and Charles Murray’s The Bell Curve: Intelligence and Class Structure in American Life.
Sher’s references to academic grades as an uncontroversial measure of merit strike me as cavalier. For a thorough, well-argued critique of college-level grading, see Jason Brennan and Philip Magness, Cracks in the Ivory Tower: The Moral Mess of Higher Education, ch. 5. Brennan and Magness’s discussion of student evaluations in chapter 4 is likewise on target.
***Thomas Nagel, “Equal Treatment and Compensatory Discrimination,” Philosophy & Public Affairs 2:4 (Summer 1973), discussed in Sher’s Desert, pp. 123-25.
(I edited this post for clarity initial posting, but not in ways that changed the substance of anything in the original version.)