Stephan Hartmann, Borut Trpin (2026), Philosophical Studies, 183(6), 1817–1843 — DOI-10.1007/s11098-026-02517-x
Key takeaways
- Measuring explanatory power as how well the whole set of explanantia coheres with the explanandum versus with its negation (with the CohOG+ coherence measure) gives a measure that provably satisfies Roche & Sober’s Two Likelihood Inequalities for single hypotheses, and that in two worked examples escapes their chain and negative-interaction problems by scoring a set differently from a conjunction. That is a possibility result, not a general theorem. The reply to Lange’s entailment problem is generic and, by the authors’ own account, partial.
Seeded from the abstract and a skim, before the reading in NOTE-168. What follows is what the work says about itself.
Recent critiques have questioned whether probabilistic measures of explanatory power can work at all. The authors respond with a family of coherence-based measures that avoids those problems. Instead of scoring only how well one hypothesis accounts for the evidence, the measures compare how well the whole explanatory package, which may contain several explanantia, coheres with the explanandum against how well it coheres with the explanandum’s negation. They show that this accommodates key scientific judgments while keeping the advantages of a probabilistic framework.
Standing in the record
Filed from the 2026-09-26 sweep of the papers-feed tracker for philosophy: 165 seconds of active reading over 2 sessions. It was filed Deferred, unread. NOTE-168 is the close reading of 2026-09-26, and it placed the work: Active — worth reading as a current, compact reply to Roche & Sober (2023) and Lange (2022) with checkable formulas, though its escape is representational (sets instead of conjunctions) and its positive case rests on two worked examples rather than a characterisation.
When seeded, before the reading in NOTE-168, the entry gave a priority for a deeper reading: high — the longest dwell in the batch (165 s over 2 sessions), recent, and technical. The skim gives the construction, but the §5.1 case analyses and proofs carry the result and were not read.
What the seeding asked a deeper reading to check (NOTE-168 is that reading):
- A formal, set-valued measure of explanatory power is relevant wherever the anthology weighs competing explanations of a training phenomenon that involve several interacting factors.
- Check §5.1’s handling of Roche & Sober’s temporal-chain and negative-interaction cases; that is where the measure earns its claims.
- Check the companion paper (Hartmann & Trpin, forthcoming) that introduces and defends CohOG+, since the result depends on that choice of coherence measure.
Access when seeded: link.springer.com returned a JS client challenge. I fetched the open-access (CC BY) publisher PDF via rd.springer.com (27 pp); it gives received 2025-04-23, accepted 2026-03-13, online 2026-04-04. Read the abstract, keywords, §1 in full (including the stated contributions and plan), the section heads, the §5 construction of the measure (eqs. 1–2 and the CohOG+ definition, Proposition 1), the §6 opening and §7. The proofs and the worked cases in §5.1 were not read.