This thesis introduces mechanism domains as a cognitive representation underlying causal reasoning. It has been observed that people believe they understand causal workings in more detail than they actually do. Despite the fact that our knowledge of causal systems is often skeletal in nature, we oftentimes remain capable of successfully navigating the complex systems for which we lack detailed, mechanistic knowledge. To explain this phenomenon, it is proposed that people employ the domain-matching heuristic. Over time, we learn and refine the associations by which different kinds of causes lead to different kinds of effects, in accord with how similar we believe their supposed mechanisms are. By matching the mechanism domains between specific causes and effects, we are able to dramatically reduce our search space during various types of causal reasoning, as in these cases our attention is focused on likely relations. A clustering study reveals three domains of causal mechanisms in the space of artifacts: the mechanical, the chemical, and the electromagnetic. In a series of experiments, these domains are evaluated in the context of causal attribution, prediction, believability judgment, and subjective understanding. The findings show that mechanism domains offer a unified and economical representation that can predict human performance across a variety of tasks. The domain-matching heuristic offers a way of thinking about causality that does not require theoretical knowledge; instead, it relies upon a set of abstract representations that reflect the true nature of underlying mechanisms.
Goldin, Gideon,
"Mechanism Domains in Causal Reasoning"
(2014).
Cognitive Sciences Theses and Dissertations.
Brown Digital Repository. Brown University Library.
https://doi.org/10.7301/Z0HQ3X8V
