What counts as good evidence of mechanisms? The Support-Absence Framework

Abstract

Abstract Over the past two decades, evidence of mechanisms has achieved great relevance in methodology and philosophy of science. Several authors have argued that such evidence should play a more central role in scientific practice. Nonetheless, important concerns have been raised regarding the variable quality of evidence of mechanisms and the reliability of mechanism-based reasoning. In this paper, in response to those concerns, I address the assessment and evaluation of evidence of mechanisms. I first analyse and discuss previous contributions on the topic. Although valuable, these proposals face a number of significant limitations. Building on them, I then introduce and develop the Support-Absence Framework. This framework integrates support and absence-of-error criteria to assess evidence of mechanisms. To explore and illustrate the application of this framework, I consider process tracing and agent-based modelling.