The variety-of-evidence thesis: a Bayesian exploration of its surprising failures

Abstract

Abstract. Diversity of evidence is widely claimed to be crucial for evidence amalgamation to have distinctive epistemic merits. Bayesian epistemologists capture this idea in the variety-of-evidence thesis: ceteris paribus, the strength of confirmation of a hypothesis by an evidential set increases with the diversity of the evidential elements in that set. Yet, formal exploration of this thesis has shown that it fails to be generally true. This article demonstrates that the thesis fails in even more circumstances than recent results would lead us to expect. Most importantly, it can fail whatever the chance that the evidential sources are unreliable. Our results hold for two types of degrees of variety: reliability independence and testable aspect independence. We conclude that the variety-of-evidence thesis can, at best, be interpreted as an exception-prone rule of thumb.

Evidence amalgamation in the sciences: an introduction

“…Scientists must of course be party to these development, but philosophers are perhaps especially positioned to contribute in a different way, by abstracting general considerations about topics from specific cases-e.g., concerning the confirmatory role that varieties of evidence may and may not play (Claveau and Grenier 2018), and the justification of procedures for meta-analysis (Vieland and Chang 2018)-then applying those lessons to new cases in different disciplines. Indeed, what has emerged as a common theme in many contributions to this special issue is how one can and should transfer knowledge from one domain or problem to another: extrapolation from one population to another, the external validity of an experiment, the robustness of a measurement technique, extending static to dynamic causal structure, etc.…”

“…Claveau and Grenier (2018) extend the Bovens and Hartmann model by formalizing the notion of unreliability in a way which is closer to scientific practice. Furthermore, they consider consequence variables and reliability variables that are dependent to a degree, showing that the VET fails in many of their models.…”

“…(See also Reiss's sketch of his pragmatist-contextualist theory of evidence for causal inference in Reiss 2018. ) 3 Contributions 3.1 Claveau and Grenier on the surprising failure of the variety of evidence thesis Claveau and Grenier (2018) follow up on Claveau (2013) in offering more nuanced understandings of varied evidence in the sense of Bovens and Hartmann (2003) and Sect. 2.1.…”

“…In their contribution, Claveau and Grenier (2018) extend this analysis by first reintroducing the testable-consequence variables into the models and then dropping the assumptions that these variables are either fully dependent or fully independent. These, so far, most general models, are depicted in Fig.…”

“…In a natural sense, the set in which the VET fails is much larger in Claveau and Grenier (2018) than it is in Claveau (2013). Consequently, the conclusions drawn for the thesis are less favorable in Claveau and Grenier (2018) than in Claveau (2013). Heesen et al (2018) are interested in the work of Du Bois (1996/1899) concerning the question, "What do Negroes earn?"…”