My prior post about metacognitive diversity pointed to the difficulty of reconciling invariance in procedural and analytic metacognition with the high variability in predictive practices. In this post, I attempt to address the puzzle itself. How can the coherence between the three sets of data be restored? My strategy will be to explore the procedural underpinnings of both B and C type of evidence, and show that the same kind of basic process explains them. Let us first consider the universality of Gettier and similar epistemic intuitions.
About epistemic intuitions
A standard way of explaining universality in intuitions consists in drawing on Chomsky’s own argument: Universality in syntactic intuitions is claimed to derive from the existence of a universal grammar that is hard-wired into the brain. A nativist theory about how children use and develop their concepts similarly claims that all humans rely on an intuitive mental lexicon when forming their spontaneous inferences in a variety of domains, such as folk biology, folk psychology (mindreading), numeracy, in addition to folk epistemology. My present goal is not to discuss the validity of a nativist conception of any of the core domains of knowledge.
What should be pointed out is that intuitions are not identical to spontaneous inferences. As discussed in the first post, intuitiveness is a metacognitive feeling, and hence, does not properly belong to the content of the first-order inference that feels intuitive. Intuitiveness is rather generated by the comparatively higher fluency of the processing sequence underlying the inference. Higher fluency, however, does not exclusively belong to innate informational processes. Various types of experimental manipulations of fluency such as implicit learning, repetition, experimental priming, have been shown to selectively increase intuitiveness of a judgment as well as one’s confidence in its validity. (Topolinski & Strack, 2009). Metacognitive feelings based on fluency, then, determine our preferences for using specific concepts, our dispositions to detect the semantic value of a definition, or the coherence between several claims.
Truth evaluation across cultures
How do people proceed to discriminate whether P is true? According to the eminent metacognitive scientist Norbert Schwarz, people ask themselves one or more of the following questions: (1) Do others believe it? (2) Is there plenty of evidence to support it? (3) Is it compatible with what I believe? (4) Does it tell a good story? (5) Does it come from a credible source? (Schwarz et al. 2016). In other words, people primarily assess consensuality, evidentiality, coherence, relevance, and source credibility. If P scores high on several of these dimensions, a thinker will tend to believe it is true; if P scores low, it will be held false. Note that the case of the “consensually wrong”, studied by Koriat (2008), does not contradict Schwarz’s view: both speak about what constitutes sensitivity to truth, not about truth. (Keep in mind this observation, we’ll need it later.)
How does this analysis affect the nature of epistemic intuitions? It appears that the epistemic judgments elicited by the experimental philosophers are generated by a heuristic combination of features 1 to 5. These features have been selected, not (I venture to claim) because we share an innate folk epistemology, but because we all converse, and learn converging standards of evidentiality and relevance when we do. In this process, we assess our own subjective confidence in what others tell us, using a polar scale of intensity and valence (from no confidence to full certainty.) According to the conversational normative patterns of our group, we will tend to weigh, for example, consensuality higher than evidentiality (or reciprocally). Even when trusting consensus and testimony, we will have had enough experience to know that they may occasionally fail to be reliable. For example, when you’re told that your wife is at the hospital, it’s likely to be true: why would people lie to you? But there might indeed be several Mrs Jones. Wait until you see her to be certain she is in this hospital, to report that you know where she is.
To summarize this point: we need only to assume that the same informational constraints apply to conversation across populations for explaining that the pattern of reporting/attributing knowledge should tend to converge at the limits of the evaluative scale (knowledge/ignorance). Hence they might converge when attributing a positive value to a true utterance and a negative value to a false utterance. Convergence does not need to extend to the desired level of informativeness. Human groups have different rules concerning who deserves to be informed, and who does not. How then, can this analysis solve the puzzle of the diversity in epistemic practices?
The Azande’s oracle and truth
Recall that the Poison Oracle in Azande claims to form predictions “beyond doubt.” Practitioners find the oracular practice to fulfill all the demands that laypeople from the West place when attributing reliability in a method aiming to truth: (1) it is consensual (2), it is based on evidence, (3) it is coherent with what people believe, (4) it is relevant (i.e. highly informative about an important matter) and (5) its source – “the Poison” is trustworthy. Note that these conditions always apply as a function of a specific social and epistemic pattern of practices. As proposed by Bloch (2008), divinatory practices are trusted in many traditional societies because, in contrast with testimony, their outcome can be perceived, and cannot be easily manipulated. Coming back to my former question in post 3: why do people consider a lethal outcome as caused by the response of the oracle rather than by poison alone? It is quite likely that there is no observable regularity between the quantity of substance ingested and the consequent death, because the effect of the poison depends both on the plants harvested – toxicity is likely to differ widely across samples – , and on the age and health condition of the victim. The Azande, furthermore, notoriously superimpose an intentional to a causal account for each significant event of their lives. Yes, the poison is lethal, but this does not suffice to explain this particular death at this particular moment. Hence, the poison is treated intentionally, in coherence with a local consensus. A final aspect of the epistemic process can only be mentioned here, although it deserves more space. Why does it seem important for the assumed reliability of the poison oracle that both the questioner and the practitioner should respect sexual abstinence several days ahead of time? Sexual abstinence might be a factor of trust; sexual abstinence might constitute both a salient and a fluent social rule for elevating confidence in social practices endowed with an epistemic authority. Westerners might fall prey to a similar pressure of a salient social rule when applying an equality rule in epistemic decision making, irrespective of the variable expertise of participants. (Mahmoodi et al., 2015). Both rules (sexual abstinence and equality) reflect consensual patterns of practices, which, being fluent, seem to be truth-conducive. But they both instantiate a case where people are consensually wrong. As shown by Koriat, however, consensuality about wrong rules, correlates with the participants being highly confident of being right to apply them.
In summary, I have argued that feelings of fluency play a major role in eliciting in participants a sense of the truth-conduciveness of the oracle. When a social predictive practice is based on locally accepted concepts, often consulted, difficult to manipulate, easy to understand, rich in affect, relevant to individual and social welfare, backed up by local social rules or taboos, its outcomes tend to be assessed as true. This heuristic allows for quite diverse practices to be trusted across cultures. This diversity, however, does not suggest that the concept of truth itself varies across practices. Nor does it exclude that ways of collecting evidence should not, when education and knowledge develop, be subject to more precise epistemic scrutiny.
The featured image of this post is a work by Ulysses Belz, entitled “Selfportrait contemplating the sky” (2006). It belongs to a series entitled “Metacognitive Painting”.