Biology & Philosophy Humanities and social sciences
A general account of scaffolding explanations: cognitive science, evolutionary biology, and beyond
Abstract
This paper presents a general account of scaffolding explanations that fits with previously recognized uses in cognitive science and evolutionary biology that can also be extended to other areas, such as AI and biochemistry. Scaffolding explanations are causal explanations that explain how a system realizes, maintains, or acquires a capacity/property by interacting with an arrangement of entities and activities external to the system. I show the potential of this general account by using it to interpret empirical work on the capacity of large language models for long-range reasoning. The general account also shows the ways in which scaffolding explanations differ from other basic types of causal explanations, such as etiological (e.g., causal history), constitutive (e.g., mechanistic), and developmental explanations. Scaffolding explanations have different explananda than etiological explanations, and unlike constitutive and developmental explanations, they are able to attribute the realized capacity/property to the system that interacts with the scaffold, instead of the constitutive whole which includes both the system and the scaffold. This means that scaffolding explanations are a unique type of causal explanation. However, there are cases where competing empirically adequate and internally coherent constitutive and/or developmental explanations can be offered. Using research on the evolution of multicellularity as a case study, I argue that, in at least some cases, there are pragmatic reasons for preferring the scaffolding explanation over these alternatives. This analysis shows that there are cases where pragmatics will play an essential role in philosophical work on scientific explanation.