Nancy Cartwright (2021), Synthese, 199(5–6), 13095–13119 — DOI-10.1007/s11229-021-03368-1
Key takeaways
- Demanding “rigour” (as exemplified by RCTs) works against the evidential diversity that singular causal claims need; reliable predictions about whether C will cause E in a specific setting require evidencing every step of the causal process, which takes a hotchpotch of mostly non-rigorous methods.
Seeded from the abstract and a skim, before the reading in NOTE-161. What follows is what the work says about itself.
Cartwright argues for evidential diversity and the mixed methods it requires, focusing on singular causal claims about what a cause will do, or did, in a specific setting. She offers a template that catalogues the kinds of evidence such claims need, generated by asking what must happen for a causal process to carry through from cause to effect. Unlike the usual mixed-methods argument, which triangulates on one overall claim, her template targets the many subsidiary claims that must hold for the overall claim to be true. The methods needed seldom meet today’s standards of rigour, so she asks what “rigour” actually means and argues that emphasizing it can be counterproductive. RCTs, the paradigm of rigour, are used to show what rigorous methods can and cannot deliver.
Standing in the record
Filed from the 2026-09-26 sweep of the papers-feed tracker for philosophy: 10 seconds of active reading over 1 session. It was filed Deferred, unread. NOTE-161 is the close reading of 2026-09-26, and it placed the work: Active — worth reading as a compact, citable statement of the “rigour is overvalued, evidence the causal process” position. The pToC template is concrete, but the paper openly offers no method for weighing the evidence it catalogues.
When seeded, before the reading in NOTE-161, the entry gave a priority for a deeper reading: medium — 10 s dwell, but a clear statement of an influential position in the evidence cluster (LIT-213–LIT-118). The skim captures the argument, and the template is what remains to extract.
What the seeding asked a deeper reading to check (NOTE-161 is that reading):
- Transfers to ML evaluation: a benchmark gain, like an RCT average, does not warrant the claim that a method will work in your setting. That needs evidence about the steps connecting intervention and outcome in the deployment context.
- Check the full pToC template (§4 and its figures) for an explicit checklist of evidence types that could be adapted.
- Read alongside Worrall (LIT-175) and Reiss (LIT-118) on evidence hierarchies and pragmatist evidence.
Access when seeded: link.springer.com returned a JS challenge. I fetched the open-access (CC BY) publisher PDF via rd.springer.com (25 pp) and the PMC copy (PMC8727433). The PDF gives received 2020-12-31, accepted 2021-08-16, online 2021-11-01. Read the abstract, keywords, §1–§4 heads and openings, footnote 1, and the closing “In sum” section (§5).