How Science Works · Philosophy of Science · An Argument I’m Still Inside

What Makes Science Good? Reaching for a Standard Nobody Ever Finishes

DRAFT — review before publishing to students

Every other reading on this site tries to explain something I’m reasonably confident about. This one is different, and I want to be honest about that up front. It’s about a question I have been genuinely stuck on while writing the book this site supports: what makes a piece of science good? Not “correct” — we have decent machinery for correct, and other readings here cover it. I mean good. Worth doing. Worth funding. Worth a career. That question turns out to sit in the middle of a live argument among professional philosophers, one I have read my way into and not fully out of. I’m writing it up anyway, because watching someone work on an unfinished question is a more accurate picture of intellectual life than watching them recite a finished one — and because if you go into science, you will eventually have to answer this for yourself, with no philosopher in the room.

The question your science class skips

A lab course teaches you to get an answer right. It rarely asks which questions deserved the effort in the first place. But somebody is always deciding that. A review panel decides which proposals get funded. A journal decides what’s worth publishing. A graduate student decides what to spend five years on. Every one of those is a judgment about value, not just accuracy, and accuracy alone can’t settle it — you can run a flawlessly rigorous study on something that mattered to no one. So what standard are those people actually using?

The obvious answer, and why I stopped believing it

The intuitive answer is: whatever produces the most total benefit. Add up the good, pick the biggest number. I believed some version of that for a long time. I don’t anymore, and not because I went soft on the idea that some choices really are better than others. I stopped because I don’t think anyone can actually get their bearings on what the greatest good is, in full, in advance. Try to specify it completely for even one funding decision — whose benefit, over what timescale, weighed how against whose cost — and you will still be specifying when the deadline passes.

Two ways to say “better”

The economist and philosopher Amartya Sen, in his 2009 book The Idea of Justice, names this problem more precisely than I could. He distinguishes two ways of judging. A transcendental approach specifies the complete, ideal arrangement first, then measures every real option against it. A comparative approach skips that step: you don’t need a finished picture of perfection to know that one real, available option is better than another real, available option. You rank the two in front of you, and you act on the ranking.

Sen argues the second is what we actually need, and you can feel the appeal immediately. You don’t need a completed theory of justice to know that arresting people arbitrarily is worse than not doing that. Applied to science: you don’t need a finished theory of the public good to judge that funding a tuberculosis vaccine beats funding a study nobody will read.

Here is where it stops being tidy

I was ready to leave it there. Then I read the philosophers who disagree with Sen, and I think they are largely right — which is the part of this reading I most want you to see, because it’s what actual intellectual work looks like from the inside.

Three objections, escalating. First, that Sen draws the line too cleanly: he makes his main target, the philosopher John Rawls, sound more rigidly ideal than Rawls actually was. Second, that the two categories aren’t really rivals, because any workable moral theory contains both — some comparative machinery for acting in the real world, and some ideal element that keeps the comparisons honest. Third, and sharpest: a comparative approach cannot stand alone at all. The moment you rank two options, you have already assumed some direction that makes one of them better. You didn’t escape the ideal. You just stopped saying it out loud.

That third objection is the one that got me, because it’s not a quibble — it’s a demonstration that the tidy version of Sen’s argument, the version I found appealing, doesn’t survive contact with its own logic.

What a standard can be, if it can’t be finished

The way out isn’t to pick a side. It’s to notice that an ideal can do real work without ever being completed. Philosophers after Kant call this a regulative ideal: a standard whose job is to orient the work, not to be reached and checked off. Two thinkers give it useful shape.

Charles Sanders Peirce described truth itself this way — not what any one scientist currently believes, but whatever a community of investigators would eventually converge on if testing and disagreement ran their full course. Nobody holds that. Everybody in the field is oriented by it. Notice that this is exactly how scientific consensus already works, which is the subject of another reading on this site: no single person carries the whole answer, and the reaching is the process rather than a substitute for it. John Dewey took the same move into ethics and called the good growth — not a number to hit, but a direction you can tell whether you’re moving in.

That gives a standard you can actually use, and it satisfies the objection instead of dodging it. Rank the real options in front of you, because that’s how anything gets done. Stay oriented toward a good you will never finish specifying, because that’s what makes the ranking mean anything. Take away the ranking and you have a philosophy that never acts. Take away the ideal and you have motion with no direction.

Where this shows up in a real career

This is not a seminar exercise. When a review panel scores a proposal on whether it benefits society beyond the lab, it is making a comparative judgment under exactly this kind of unfinished ideal — nobody hands reviewers a completed theory of the public good, and the work still has to get scored by Friday. When a scientist chooses a research question, the honest version of that choice is rarely “this maximizes total benefit.” It’s “between the two projects I could actually start, this one gets closer to something I think matters.” And when I decided years ago to leave a selective program for a public school with no entrance exam, I couldn’t defend that against any abstract greatest good. I ranked two real options and asked which one built more actual capacity in more actual students. I still ask a smaller version every year: did this build something real in the person in front of me, from wherever they started? I can’t tell you what the greatest good is. I can tell you whether something grew.

What I’m still not sure about

Three things, stated plainly so you can see the shape of an open question. I don’t know whether the requirement that comparison needs an ideal is a matter of strict logic or only of practice — the philosophers I’ve read disagree about that, and it matters. I don’t know whether “growth,” Dewey’s word, is precise enough to settle hard cases or whether it just relocates the vagueness somewhere less obvious. And I’m aware that assembling three thinkers who never wrote about each other is a move a professional philosopher might reasonably object to, however carefully I do it. If any of that bothers you, good. That reaction is the same instinct that makes someone good at science, pointed at a different target.

Transcendental approach: Sen’s term for judging real options by first specifying a complete, ideal arrangement and then measuring everything against it.
Comparative approach: Judging by ranking two real, available options against each other and acting on the ranking, without completing a picture of the ideal.
Regulative ideal: Kant’s term for a standard whose function is to orient and guide inquiry rather than to be reached; it does real work while remaining permanently unfinished.
Convergence theory of truth: Peirce’s account of truth as whatever a community of investigators would ultimately agree on if inquiry and disagreement were allowed to run indefinitely.
Growth (as a moral standard): Dewey’s reframing of the good as a direction of improvement you can detect you’re moving along, rather than a fixed endpoint to be reached.
Normative question: A question about what ought to be valued or done, as distinct from an empirical question about what is the case; “what makes this science good” is normative, “is this result correct” is empirical.

Think about it

Explain the difference between a transcendental and a comparative way of judging which of two research projects is better. Give your own example of each.
The strongest objection in this reading says that ranking two options already assumes an ideal. Explain that objection in your own words, and then explain why it doesn’t force us to abandon comparative judgment entirely.
Using the idea of a regulative ideal, explain how a standard can be genuinely useful even though no one will ever fully reach it. Then explain how scientific consensus works the same way.
A reviewer has to score a proposal on whether it benefits society, but nobody has given them a finished definition of “benefits society.” Describe how they could still make a defensible judgment, and what would make that judgment indefensible.
This reading ends by naming three things the author is still unsure about. Pick one, and argue either that it’s a serious problem for the argument or that it isn’t. Use evidence and reasoning from the reading.
Think of a real decision you’ve made where you couldn’t define the ideal outcome in advance. What did you rank instead, and what direction were you orienting toward without saying so?

Sources: Sen, Amartya. The Idea of Justice. Harvard University Press/Belknap Press, 2009 — the transcendental-versus-comparative distinction. Peirce, Charles Sanders. “How to Make Our Ideas Clear.” Popular Science Monthly, 1878 — truth as the convergent limit of a community’s inquiry. Dewey, John. Reconstruction in Philosophy, 1920 — the good reframed as growth. Kant, Immanuel. Critique of Pure Reason, 1781 — the origin of the regulative ideal. On the criticism of Sen’s dichotomy, in escalating strength: Valentini, Laura. “A Paradigm Shift in Theorizing About Justice? A Critique of Sen.” Economics & Philosophy 27(3), 2011; Biondo, Francesco. “Comparative vs. Transcendental Approaches to Justice: A Misleading Dichotomy in Sen’s The Idea of Justice.” Ratio Juris 25(4), 2012; Ege, Ragip, Herrade Igersheim, and Charlotte Le Chapelain. “Transcendental vs. comparative approaches to justice: a reappraisal of Sen’s dichotomy.” The European Journal of the History of Economic Thought 23(4), 2016. On Sen’s already-recognized pragmatist affinities: Zimmermann, Bénédicte. “Pragmatism and the Capability Approach.” European Journal of Social Theory, 2006. See also this site’s companion pieces, “How Science Actually Gets Funded: Peer Review, NSF Criteria, and What ‘Consensus’ Really Means” and “Finding Your Big Idea: A Real Method for Navigating the Hypothesis Space.” This reading is the companion to Ch. 12 of Everybody’s Science, where the same argument appears as the book’s “Reaching Standard” breakout box. Personal reflection: J. Reid Schwebach, on working through this question while drafting that chapter — including the decision to leave a selective program for a public school. DRAFT — verify current Virginia Science SOL alignment (if any) with the current Curriculum Framework before publishing.