Methodological Pluralism

The view that no single scientific method suffices to establish scientific knowledge; several, potentially incommensurable, methods are needed. Originally J.S. Mill; systematised by Paul Feyerabend (Against Method) and Helen Longino.

Two readings

Descriptive — as a matter of fact, science does use plural methods.
Normative — science should use plural methods to avoid losing out on knowledge. This is the version the course defends.

Longino: any single approach to science is incomplete; there can be more than one correct scientific account of the world.

The course's application

Anjum & Rocca argue in sources/2023-anjum-rocca-phi403-lecture-20-plural-methods-one-causation that methodological pluralism follows from the nature of concepts/causation:

Causation has many symptoms — difference-making, regularity, manipulability, probability-raising, energy transference.
No symptom accompanies causation in every case.
Therefore each symptom requires its own method, and no single method is reliable for all cases.

This can be combined with causal monism (one causation, many symptoms) or causal pluralism (many causations). See concepts/causal-pluralism.

Evidence ranking with conflicting methods

When methods disagree, methodological pluralism requires an evidence-ranking scheme — not the rigid EBM hierarchy that automatically privileges RCTs, but a context-sensitive weighing of which method's symptom is most reliable for the case at hand.

Implications for process / data science

concepts/causal-pluralism · concepts/evidence-hierarchy · concepts/causation · concepts/rct-limitations