Causal Mechanisms in the Social Sciences

Introduction

The central thesis of mechanistic explanation is that to account for a social phenomenon is not to correlate it with others but to exhibit the generating mechanism — the set of entities, activities, and relations whose operation produces the recurrent outcome. For MARTRO the implication is direct: an organisational diagnosis cannot halt at the statement that "communication is lacking" or "roles are not clear" but must show how that specific friction originates and reproduces itself. The difference is not rhetorical: it separates description from explanation.

In the small firm the mechanism is often more visible than it appears, though obscured by daily normality. An order passes from the commercial function to operations with incomplete information; a modification is communicated orally; the owner intervenes as an obligatory point of passage; the operator compensates with personal experience; rework presents itself as the ordinary labour of the work. Mechanistic explanation requires distinguishing symptom, generating cause, and compensation — and it is this passage that renders the diagnosis less decorative.

The junction with MARTRO is immediate: every finding should be reconstructible as a chain. Which initial condition renders the friction possible? Which recurrent behaviour feeds it? Which role compensates for it? Which observable effect does it produce? Which evidence sustains it, and which would weaken it? This form of reasoning does not claim experimental causality but increases the verifiability of the diagnosis and reduces dependence on the dominant narrative.

The paper further corroborates the priority of triangulation. A problem declared by the owner alone may reflect a leadership perception; if it emerges solely from operators, it may constitute a local experience; when it recurs across surveys, interviews, episodes, and process micro-evidence, it becomes a more solid structural candidate. Not because triangulation proves anything in a strong sense, but because it compels the commentary to remain anchored to heterogeneous sources, with the effect of a more sober and more defensible diagnosis.

The boundary is fundamental: MARTRO must not declare strong causality on the basis of surveys or interviews. The source authorises no scientistic shortcuts but a discipline of explanation — rendering findings inspectable, argued, and pragmatically falsifiable. The risk is the use of "mechanism" as a merely sophisticated term; the corrective consists in always requiring an example of process, an episode, or an artefact that shows its operation.