Source
Leonelli (2023), LIT-189 (Philosophy of Open Science, Cambridge Elements in the Philosophy of Science), read in full in NOTE-098. Page references are the Element’s printed pages.
What was actually shown
The sharing view (§2, Table 1, Figure 1). Mainstream institutional Open Science reads openness as sharing that is unlimited, digital, good, global and equal, with a value ordering transparency → quality (reproducibility) → inclusion. The EC 2016 definition is the exemplar. The author calls her reading “opinionated and … by no means comprehensive”.
Four cases (§3).
- GISAID: access conditions (a user agreement, credit, limits on linkage) protected low-resourced producers and drew wide participation, 240 countries and 15 million sequences by February 2023 (p. 21). Calls for fully open data rank transparency over inclusion.
- Software: in a Global Young Academy survey in Bangladesh, Ghana and Tanzania, researchers preferred expensive proprietary software, fearing reviewers would read free software as low quality (p. 23).
- Crop data: open linkage favours decontextualised genomic data over phenotypic and local knowledge, and can become “yet another form of bioprospecting” (p. 26).
- Reproducibility: taking controlled experiments as the universal model damages exploratory, field-based and qualitative research (pp. 26–28).
The mechanism (§§4–5). Systems of practice are local, entrenched in repertoires, more or less permeable, and “systems of demarcation and exclusion” (p. 35). Epistemic injustice is “typically, inversely correlated” with epistemic diversity (p. 39). The sharing view rests on an object-oriented epistemology in which outputs are self-standing commodities. But “the further a research object travels from its context of origin, the more difficult it becomes to assess” whether it fits a new use (p. 46). Hence sharing is “neither a necessary starting point nor a sufficient condition” for reliable open science (p. 8), and the ordering should start from inclusion (Figure 2).
What could have come out otherwise. In the cases, open access could have widened participation and preserved local expertise. As the author describes them, it did not. No case measures reliability.
What this does not say
- That sharing-first open science has in fact lowered reliability. The reading grades that link weak to moderate (C3): it is argued through cases and standpoint theory, not measured.
- That diversity and injustice are inversely correlated in general. That is an assertion with two examples (C4).
- That sharing is bad. It is not a sufficient condition and not a necessary starting point. Leonelli calls sharing as reuse “prima facie … a much better foundation” than sharing as access, while arguing both share the object-oriented epistemology.
- How to decide what is judicious. “How this vision of openness can be effectively realized remains itself an open question” (p. 66). There is no decision rule, only things to attend to.
- That it transfers to fully computational work. The base is biology and biomedicine. Computer simulations are the case where “both procedures and results are expected to be fully reproducible” (p. 27), and the book never discusses machine-learning model release.
- That every open-science initiative holds the sharing view. Many grassroots initiatives already embody the view she favours (pp. 2–3).
Connections
- THEORY-317 and THEORY-326 are the record’s accounts of epistemic injustice (Fricker and its revision). This account uses Fricker’s testimonial and hermeneutical kinds and adds a claim about their effect on reliability, which those accounts do not make.
- Worrall (THEORY-491) and Cartwright (THEORY-451) question the RCT’s standing from the confirmation-theoretic side; Leonelli questions it as a universal exemplar of reproducibility.
- Feest, LIT-161, is cited for the method-relativity of reproducibility, and Douglas, LIT-139, for the absence of a strict internal/external boundary for values.