MetaScience, Scientific Inquiry, and Agents

Scientific advance through plural inquiry

A thirteen-meeting reading club studying scientific advance through plural epistemic modes, and designing agents evaluated against documented scientific work.
Author

Denolle Lab, University of Washington

Published

September 24, 2026

About this Group Meeting

Drafted curriculum v0.4 · convenor authorized revision · meetings 1 and 10 recorded

MetaScience, Scientific Inquiry, and Agents · Denolle Lab · thirteen one-hour meetings, taken one at a time

Why

  • Scientific advance is the main outcome: a warranted gain in knowledge, understanding, or research capability.
  • Study plural epistemic modes in geosciences, informed by the history and philosophy of science and empirical studies of scientific process and impact (Fortunato et al. 2018).
  • Measure novelty, correctness/warrant, and impact to understand the newness, support, and consequences of a gain; do not substitute any one for advance.
  • Design agents around human-defined modes and evaluate them against documented past work, then new research.
  • Test when interdisciplinary integration contributes to advance. Individual competence and collective breadth are related, distinct hypotheses.

How we meet

  • One anchor, read in selected sections by everyone; one companion, presented in depth by a rotating reader. Others read its abstract or overview.
  • Optional readings are resources, not extra assignments. No dates are fixed.
  • A flexible hour: 5 minutes framing, 10 companion, 25 discussion, 15 case exercise, 5 shared record.
  • One brief episode entry and three sentences: what changed; what evidence warrants it; what agent action or evaluation follows.
  • Pages stay concise. Detailed worksheets are reusable resources; the research studies are optional participation.
  • Distinguish a primary result, philosophical argument, historical reconstruction, ethnography, empirical association, and benchmark.

Working status

Material Status
Scope and implementation of v0.4 Authorized by the convenor, September 10, 2026.
Readings and eleven initial mode specifications Implemented draft for the reading club; no claim of group discussion or empirical validation.
Meeting notes Meetings 1 (September 15, 2026) and 10 (September 22, 2026) recorded from the group’s notes.
Agent procedures and historical evaluation Specifications; exported instructions are not an evaluated agent.
Corpus and prospective studies Designs; no results or selected private projects asserted.
  • Drafted: proposed text. Discussed: taken up in a recorded meeting. Adopted: an explicitly recorded decision.

Meetings

Meeting Theme Anchor / companion
1 Is there a scientific method? Platt (1964) / Cleland (2001)
2 Exploratory experimentation is not theory-free Steinle (1997) / Karaca (2013)
3 From coastal traces to earthquake histories Atwater (1987) / Nelson et al. (1996)
4 Plate tectonics: concept formation and a discriminating test Wilson (1965) / Sykes (1967)
5 New sensing capability versus trustworthy measurement Lindsey et al. (2019) / Lindsey et al. (2020)
6 Explore freely; distinguish exploration from confirmation Tukey (1962) / Nosek et al. (2018)
7 Field and marine research under real constraints Powell (2007) / Becker et al. (2019)
8 Serendipity, anomaly, and scientific interestingness Moore (2025) / Yaqub (2018)
9 Hypothesis testing when several models fit Oreskes et al. (1994) / Beven and Freer (2001)
10 Scientific novelty: new to whom, compared with what? Uzzi et al. (2013) / Fontana et al. (2020)
11 Interdisciplinary integration: breadth, competence, and scientific advance Nersessian (2022) / Shi and Evans (2023)
12 Assessing scientific advance: novelty, correctness, and impact Wu et al. (2019) / Petersen et al. (2025)
13 What architecture follows from practice, and what would show it worked? Boiko et al. (2023) / Chen et al. (2025)

Standing questions

  1. What was known and possible at the beginning, and what scientifically meaningful gain occurred?
  2. Which observation, instrument, representation, inference, or question change helped produce it?
  3. What was new relative to work available then, and what evidence warranted the claim?
  4. Which later checks and uses support its assessed correctness and impact?
  5. What is documented about the process, and what remains unknown or retrospective?
  6. What could an agent do on comparable evidence, and what would show a gain beyond a capable baseline?

What we build together

  • Working definitions and linked human-defined mode specifications, revised on cases.
  • Episode and claim records, with novelty, warrant, advance, and follow-up assessments.
  • A historical evaluation: reconstruct decisions, reproduce/reassess results, and test extensions.
  • Agent design derived from mode specifications; comparisons measure scientific gain rather than mode vocabulary alone.
  • A bounded reference sample for evaluation; a larger corpus study if the group pursues population questions.

Access and version

  • Bibliography: anchor/companion selections, required corrections, optional literature, and access notes. No journal PDFs are redistributed.
  • Contributing: signed notes, reading suggestions, and public/private boundaries.
  • v0.4 implements the converged plan. The earlier audits remain in review/; implementation does not establish the proposed agents’ effectiveness.