From different angles of science, there has been a growing interest in the abilities of groups to track the truth. The Condorcet Jury Theorem (1785) states that without communication, infinitely big groups will reach a correct majority opinion with certainty. Coughlan (2000), meanwhile formulated a model in which all agents communicate with each other, showing that majorities are only just as good as fully-communicating individuals. In reality, communication is usually between these two extremes: some agents communicate with some of the (...) others, but not with all others. We refer to this as partial communication. This thesis provides a Bayesian framework to study the influence of partial communication on individual as well as group accuracy, thereby generalising Condorcet’s as well as Coughlan’s setting. We obtain results for individual and group accuracy in three type of networks. Firstly, we study the extreme case where there is either no communication or everyone communicates with everyone. Secondly, we determine accuracy for regular networks, in which all agents communicate with equally many other agents. Thirdly, we derive a formula to express the expected accuracy in random networks, in which agents can communicate with various numbers of other agents. The formula enables us to determine the effect of various parameters on the individual and group accuracy in a random network with partial communication. Finally, we show that in random networks, despite correlation between agents, we can still obtain accurate majorities under some constraints. (shrink)

Cognitive Science, Volume 46, Issue 5, May 2022.

The paper introduces, compares and contrasts formal models of source reliability proposed in the epistemology literature, in particular the prominent models of Bovens and Hartmann and Olsson :127–143, 2011). All are Bayesian models seeking to provide normative guidance, yet they differ subtly in assumptions and resulting behavior. Models are evaluated both on conceptual grounds and through simulations, and the relationship between models is clarified. The simulations both show surprising similarities and highlight relevant differences between these models. Most importantly, however, our (...) evaluations reveal that important normative concerns arguably remain unresolved. The philosophical implications of this for testimony are discussed. (shrink)

Topics in Cognitive Science, Volume 14, Issue 3, Page 602-620, July 2022.

In recent years, the receipt and the perception of information has changed in ways which have fueled fears about the fates of our democracies. However, real information on these possibilities or the direction of these changes does not exist. Into this gap, Hahn and colleagues bring the power of Condorcet's (1785) Jury Theorem to show that changes in our information networks have affected voter inter‐dependence so that it is likely that voters are now collectively more ignorant even if individual voter (...) competence remains unchanged. (shrink)

Computer models and simulations have provided enormous benefits to researchers in the natural and social sciences, as well as many areas of philosophy. However, to date, there has been little attempt to use computer models in the development and evaluation of metaphysical theories. This is a shame, as there are good reasons for believing that metaphysics could benefit just as much from this practice as other disciplines. In this paper I assess the possibilities and limitations of using computer models in (...) metaphysics. I outline the way in which different kinds of model could be useful for different areas of metaphysics, and I illustrate in more detail how agent-based models specifically could be used to model two well-known theories of laws: David Lewis’s "Best System Account" and David Armstrong's "Nomic Necessitation" view. Some logically possible processes cannot be simulated on a standard computing device. I finish by assessing how much of a threat this is to the prospect of metaphysical modeling in general. (shrink)

The extent to which we generalize a novel property from a sample of familiar instances to novel instances depends on the sample composition. Previous property induction experiments have only used samples consisting of novel types (unique entities). Because real‐world evidence samples often contain redundant tokens (repetitions of the same entity), we studied the effects on property induction of adding types and tokens to an observed sample. In Experiments 1–3, we presented participants with a sample of birds or flowers known to (...) have a novel property and probed whether this property generalized to novel items varying in similarity to the initial sample. Increasing the number of novel types (e.g., new birds with the target property) in a sample produced tightening, promoting property generalization to highly similar stimuli but decreasing generalization to less similar stimuli. On the other hand, increasing the number of tokens (e.g., repeated presentations of the same bird with the target property) had little effect on generalization. Experiment 4 showed that repeated tokens are encoded and can benefit recognition, but appear to be given little weight when inferring property generalization. We modified an existing Bayesian model of induction (Navarro, Dry, & Lee, 2012) to account for both the information added by new types and the discounting of information conveyed by tokens. (shrink)