The Weaver's Wisdom: Contextual Values and the Social Structure of Objectivity — Epoche C2
The distinction the objection needs, and the one it uses Helen Longino's distinction between constitutive and contextual values is the pivot of the whole dispute about values in science, and the standard objection to her — that admitting values into justification opens the door to relativism — names the wrong one of the two. In Science as Social Knowledge (1990) constitutive values are those internal to the sciences: the ones that fix what counts as acceptable practice, among them truth, accuracy, predictive power, breadth of scope. Contextual values are the personal, social and political preferences of the environment in which the work is done. No one has ever objected to constitutive values; a demand for accuracy is not a threat to objectivity, it is a specification of it. The controversial claim, and the only one worth arguing about, is that contextual values are not confined to the choice of topic and the use of results but reach into the assessment of evidence itself. An earlier and shorter version of this essay had the two terms the wrong way round, presenting the admission of constitutive values as the risky move. The argument below is the one that survives the repair, and it is stronger for it, because the objection it must answer is now the real one. The essay's thesis is unchanged: that the entry of contextual values into scientific reasoning, properly located and properly disciplined, is an epistemic gain rather than a loss, and that the gain is realised only through a particular social arrangement. Its title's image can be stated once and then left to work. A strip-woven cloth is made in narrow bands on separate looms and sewn edge to edge; no band carries the design, which exists only in the assembly. That is the structural claim about objectivity in one sentence, and everything below is the argument for it. Why contextual values cannot be quarantined Longino's route to the conclusion runs through a specific form of underdetermination, and it is worth being exact about which form, because the loose appeal to Duhem and Quine that usually stands in for it will not carry the weight. Her claim is about the evidential relation itself. Let $e$ be some observed state of affairs and $h$ a hypothesis. That $e$ is evidence for $h$ is not a fact about $e$ and $h$ alone; it holds only relative to background assumptions $b$ that describe $e$ as the sort of thing that would obtain if $h$ were true. Vary $b$ and the same $e$ supports a different hypothesis, or none: $$P(h \mid e, b) \neq P(h \mid e, b')$$ and nothing in $e$ selects between $b$ and $b'$. This is not the observation that theories are tested in bundles. It is the observation that the evidential relation is three-place, and that the third place is occupied by material which is not itself under test on the occasion. Where $b$ carries a contextual value — a view about which human differences are natural, which causal directions are the plausible ones, which populations are the standard case — that value is doing justificatory work, and no methodological hygiene applied to the collection of $e$ will remove it. Longino and Ruth Doell's 1983 study of research on sex differences in brain and behaviour shows the mechanism in a real body of work, and it shows it with more discrimination than the polemical literature usually manages. They distinguish two research programmes by the length of the inferential path from data to hypothesis. In brain-organisation research the path is short: prenatal hormone exposure is measured, adult behavioural differences are measured, and the inference between them requires an assumption of one-way causal control running from hormones through neural development to behaviour. The same data are consistent with models in which hormone levels and behaviour influence one another, and it is the assumption, not the measurement, that selects. In the evolutionary reconstructions of early human social life the path is long, the data sparse, and the framework does almost all the organising. Longino and Doell's conclusion is deliberately restrained, and their restraint is what makes the case usable: they do not claim the observations were fabricated or the statistics abused. They claim that the value-laden material sat in $b$, where methodological rules do not look. That yields the argument for diversity, and it is an argument rather than an aspiration. Background assumptions are invisible to their holders in proportion to how widely they are held; an assumption that everyone in a field shares does not present itself as an assumption at all, but as the description of the subject-matter. It follows that the criticism which would expose it cannot reliably be produced from inside. It must come from someone who does not hold it. Objectivity, on this account, is not a virtue an individual can practise. It is a property of the traffic between people who differ, and Longino's four criteria — public avenues for criticism, shared standards a critic can appeal to, community uptake of criticism, and equality of intellectual authority among qualified practitioners — are the conditions under which that traffic does epistemic work. The fourth carries a qualification that is also the sore point: the word 'qualified' has to be doing real work, or the criterion is empty, and it must not be doing so much work that it certifies the existing hierarchy. Four advantages, and what each one rests on The four benefits claimed for value-laden inquiry are not the same kind of claim, and listing them as though they were is the mistake that invites the charge of wishful thinking. The ordering here is by how directly each bears on whether the community's practices are truth-conducive. The first three are ordered on that criterion; the fourth is not an epistemic advantage at all, and is marked as such. Enhanced critical scrutiny. This is the direct consequence of the argument above. Its ground is not that a variety of viewpoints is stimulating bu