Trust and Cognitive Reliability in Indian Gram Panchayats — Epoche B2
Trust in a village assembly improves the evidence and weakens the verdict The claim to be tested A familiar argument for local self-government runs on epistemic grounds. Villagers know their own fields, their own water, their own quarrels. A distant bureaucrat does not. So decisions taken in the village will more often be correct decisions. When the argument is made about India, it is usually made about the institutions created by the seventy-third amendment to the Constitution in 1992, which gave statutory form to the gram panchayat , the elected village council, and to the gram sabha , the assembly of all registered voters in the village. The panchayat governs; the sabha deliberates. It is the sabha that the epistemic argument is really about. The argument has a second half that is less often examined. Local deliberation is said to work through trust: through elders, mediators and long-standing figures whose word carries weight because of what they have seen and how they have behaved. This essay tests whether that second half supports the first. It does not. Deference to a trusted informant raises the quality of the evidence in the room and lowers the reliability of the judgement the room reaches, and the second effect is not a defect in practice but a consequence of the structure. An epistemic defence of village deliberation therefore has to be made about how evidence is gathered, not about how many people agree at the end. That claim can be denied, and the final section says what would refute it. The procedural opponent does not exist It is tempting to set this up against Jürgen Habermas [1] , on the grounds that his account of legitimacy in Between Facts and Norms is purely procedural and therefore has no room for the quality of outcomes. That is not what the book says. The discourse principle and the democratic principle are introduced in its third chapter, and the whole of chapter seven, on deliberative politics, is devoted to arguing that the procedure grounds a presumption — fallible, revisable, and no stronger than a presumption — that what emerges from it is reasonable. Habermas's proceduralism is already epistemic. He does not think fair discourse merely happens to be fair; he thinks it is the arrangement under which the better argument can prevail, and he expects that to show in the results. So the interesting question is not whether legitimacy has an epistemic dimension. Both sides agree it does. The question is what mechanism supplies it, and whether trust is that mechanism. Two mechanisms, with different requirements Deliberation could improve a group's judgement in two quite different ways, and they make incompatible demands. The first is pooling. Different people hold different fragments of evidence. Discussion puts the fragments in one place, and a person who then decides has more to go on than any of them had alone. Pooling requires that the contributions be different from one another; identical contributions add nothing. The second is aggregation. Here numbers do the work. Condorcet's jury theorem [2] , from the Essai of 1785, states the conditions. Suppose a group of $n$ members faces a question with a right answer and two options; suppose each member is independently more likely than not to get it right, with competence $p \gt 1/2$; then the probability that the majority is right rises with $n$ and approaches certainty. Christian List and Robert Goodin generalised the result in 2001 to plurality decisions over more than two options [3] , which is what real assemblies face. The theorem is the sharpest statement available of why counting heads might be a way of finding things out. Aggregation needs two premises: competence and independence. Independence is the demanding one. It requires that the members' judgements be probabilistically independent given the truth — that whatever leads one member astray does not, by the same stroke, lead the others astray. What deference does to each premise Deference to a trusted elder pushes the two premises in opposite directions, and this is the crux. It raises competence. If the elder genuinely remembers which household held which water turn before the canal was cut, then the villagers who accept his account are more likely to be right than they were before he spoke. Nothing here is romantic; it is an ordinary fact about testimony. It destroys independence, and it does so by the same act. Once the assembly has taken the elder's word, the members' judgements share a single source. Their errors are now correlated: if he is wrong, they are all wrong together. Take the limiting case, which settles the matter. Suppose every member simply repeats what the elder says. The majority is then right exactly when the elder is right, and the probability of a correct decision is his competence, whatever the assembly's size. Two hundred villagers deferring to one man are epistemically one man. The whole point of the jury theorem — that adding people adds reliability — has been cancelled, and cancelled by the very feature the epistemic argument was praising. Real assemblies fall between the two extremes, and the mathematics behaves as one would expect: as correlation rises, the accuracy gain from group size falls away. So the honest statement is not that trust is epistemically bad. It is that trust buys its improvement in one currency and pays for it in another, and nothing guarantees the exchange is favourable. What the Nyāya rule actually licenses Indian epistemology has a developed theory of exactly this transaction, and it is stricter than the practice it is often invoked to bless. The Nyāya school treats śabda [4] , verbal testimony, as an independent means of valid cognition. The definition at Gautama's Nyāya-sūtra 1.1.7 makes testimony the instruction of an āpta , a trustworthy speaker; Vātsyāyana's commentary on that aphorism explains an āpta as one who has direct acquaintance with the matter in question and who speaks with the intention