Mechanisms in Science

A mechanism is an account of how a cause brings about its effect — the productive process that underlies an observed correlation. Mechanistic reasoning contrasts with, and complements, statistical / regularity-based reasoning.

Two trends in science

Quantitative — large-scale studies, statistics, what-and-how-often.
Qualitative / mechanistic — how and why. Theory construction, process tracing.

Mechanisms hold more explanatory power than regularities: "married men live longer" is a statistical fact but does not license intervention without a mechanism (stress, diet, loneliness do the causal work, not marriage itself) (sources/2023-anjum-rocca-phi403-lecture-12-explanatory-power-of-mechanisms).

The Russo-Williamson thesis

For a causal claim, seek both statistical evidence and mechanistic evidence. The course notes the Semmelweis case (hand-washing preventing puerperal fever) as a cautionary tale — his statistical evidence was disregarded because no mechanism of disease transmission was known.

Mechanisms of stability

Not all causation produces change. Equilibrium (Moon's orbit) and homeostasis (body temperature, blood-oxygen levels) are instances of causation producing stability. Many theories of causation — those that require events or difference-making — mis-handle these cases.

Reductionism, emergence, holism (sources/2023-anjum-rocca-phi403-lecture-13-digging-deeper)

Reductionism — all higher-level phenomena should be explained by lower-level mechanisms. Causation only travels bottom-up or within a level.
Emergence — new properties/laws appear at higher levels; allows top-down causation. Weak emergence: epistemic; strong emergence: ontological.
Holism — only wholes are causally powerful; parts gain their causal role from context.
Demergence (Anjum & Mumford 2017) — higher-level processes produce new properties at the lower level (stress → physiological change; education → better diet).

Relevance to process science

Process mining's classical descriptive/predictive outputs are regularity-like; causal process discovery and agent process observability move toward mechanism-like representations, where the functional (not merely temporal) dependencies between activities are explicit. The distinction here mirrors the mechanism-vs-statistics debate.

concepts/causation · concepts/dispositionalism · concepts/regularity-theory-of-causation · concepts/evidence-hierarchy · concepts/causal-process-discovery