Scientific polymathy: breadth, specialization, and borrowed competence

Meeting 12 | Tuesday, November 24, 2026 | 1:00-2:00 PM Pacific

Central question

Under what conditions does breadth across fields produce better science than depth in one, and does an agent that supplies competence a researcher does not hold change that answer or only relabel who is borrowing?

The paper pair

Pairing type: Historical retrospective / formal argument for specialization.

Paper A: The Polymath: A Cultural History from Leonardo da Vinci to Susan Sontag

  1. Yale University Press (2020). Publisher record | JSTOR

Read the chapters covering the age of specialization and Burke’s stages of retreat from the polymathic ideal. Ask what a prosopography of roughly five hundred recognized polymaths can and cannot establish: the sample is selected on success, so it does not tell us how often breadth failed.

Access: Publisher and JSTOR records; not open access. Use library reserve procedures for any chapter scan rather than posting a file here.

Paper B: The Division of Cognitive Labor

  1. The Journal of Philosophy (1990). Paper / publisher record

Read the model and the assumptions it needs. Note the central result: a community can be better off when individuals do not all work on the most promising problem, which makes specialization a collective strategy rather than a loss of ambition.

Access: Publisher/DOI page; full text may require institutional access.

Why these papers belong together

Burke narrates a decline; Kitcher argues that the division of cognitive labor is rational. The pairing asks whether “polymathy declined” and “specialization is optimal” are the same claim in two registers, or genuinely different claims that could come apart. Neither paper concerns agents, which is why they are useful here: they set the terms before the technology is introduced. An agent that supplies borrowed competence either adds a worker to Kitcher’s division or changes how much one worker can cover, and those two readings imply different things about what breadth is worth. (Burke 2020; Kitcher 1990)

Prepare before the meeting

Bring one project in which you used a method, dataset, instrument, or literature outside your own training. Note who or what supplied that competence, and what check, if any, would have caught an error in it.

Discussion questions

  1. Does a sample of recognized polymaths support any claim about how often breadth fails?
  2. In Kitcher’s terms, does an agent add a worker to the division of labor or extend the reach of one?
  3. What would we have to observe in this group to conclude that agent assistance produced real breadth rather than greater confidence outside our competence?

One-hour meeting

Time Activity
0-10 min Independent first judgments; surface disagreements.
10-25 min Compare the papers: claim, evidence, assumptions, and limits.
25-45 min Work through the case exercise below.
45-55 min Translate the discussion into agent requirements and tests.
55-60 min Record an output and one unresolved disagreement.

Case exercise

Take one published paper of your own and list every competence it required: instrument physics, field or shipboard logistics, numerical method, statistics, domain literature, software engineering, and writing. Mark each as held, borrowed from a coauthor, trusted without verification, or absent. Then mark which an agent could plausibly supply today, and for each of those state the specific check that would have caught a wrong answer. Separate the cases where such a check exists from the cases where it does not.

Agent-design or evaluation output

Polymathy Profile v1: a competence inventory using the held / borrowed / trusted / absent categories, a list of substitution candidates, and for each candidate the verification that must exist before delegation is acceptable. Record which verifications the group can actually perform and which are currently aspirational.

The hypothesis under test

The group is investigating a specific proposal: that agents for science let a researcher operate competently across more fields, making individual polymathy practical again in an era Burke describes as inhospitable to it. Nothing here assumes it is true, and the evidence could go either way.

It predicts that assisted researchers ask and answer questions that cross field boundaries they previously did not cross, and that the resulting work survives scrutiny by specialists in those fields. It would be disconfirmed if assistance instead produces work that specialists judge superficial or wrong, if the range of questions narrows toward what the tools handle well, or if apparent breadth turns out to be unverified trust in outputs nobody in the group can check. Note that “the group felt more capable” is not evidence for the hypothesis; it is consistent with both outcomes.

Optional extensions

Burke’s history, a formal model, a preprint measurement, and a bibliometric career study are four different kinds of evidence. None of them observed this group, and none of them tested an agent-assisted researcher.

Record after the meeting

Record the evidence for your main claim, what changed your mind, what remains unresolved, and one change to the agent or its evaluation. Keep confidential examples in private group notes rather than committing them to this public book.