What distinguishes causal mechanisms from causal laws is the indeterminacy of the former, and the death knell this sounds for any attempt to make prediction a constitutive feature of social science explanation. More precisely, the lack of determinacy is understood as a serious epistemological obstacle to the elevation of mechanisms to the status of laws, whether this indeterminancy is linked to not knowing the identity of relevant triggering conditions, or to not knowing with sufficient precision the relative force of individual mechanisms acting simultaneously. Elster’s intervention thus decisively discredits one of the central pillars of the positivist paradigm by decoupling prediction, and thus a strict deductive-nomological form of reasoning, from social science explanation. While it may still be possible to offer predictions in social science, these predictions are understood to be constitutively precarious and, in any case, non-essential for purposes of explanation. It if for this reason that we feel justified in regarding his approach as conforming to a retroductive form of explanation in the social sciences (Glynos, Howarth, 2007 89).
… from the fact that X qua process is not reducible to the contextualized self-interpretations or intentions of subjects, it does not necessarily — or only — follow that X is independent of those contextualized self-interpretations or intentions. But this is precisely what is implied by Elster’s conception of causal mechanism. .. causal mechansims can be discussed entirely on their own, with no necessary internal connection to intentional mechanisms.
Of course, the subject-independent feature of causal mechanisms is very attractive from the perspective of a positivist programme seeking to import the causal law ideal and its correlative promise of (a certain conception of) objectivity into the social sciences. After all, one of the central ingredients of a natural science conception of causality is its subject-independence. The causal process is unaffected by what any of us think about it or do in relation to it. Take the law of gravitation for instance. At the very most, we as subjects can act in light of such causal laws, but we cannot modify, or be considered supports of, the laws themselves, whether intentionally or otherwise. The functioning of comparable processes (X) in the social sciences, however is parasitic upon human practices, in the sense that they are constitutively sustained and mediated by the discursive activity of subjects. If we insist on calling such a process a mechanism, then we must accept that, unlike laws, it has the property of fungibility, that is, it can suffer dissolution. At any point, the mechanism may find that it has lost its necessary support — intentional or otherwise — in the relevant subjects. Thus, mechanisms are not ‘proto’ laws that may one day be transofrmed into ‘proper’ causal laws. This is because the functioning of causal laws does not require the passage through the subject: the content of causal laws is not parasitic upon the subjects’ self-interpretations. This is why we prefer the term ‘logic’ to ‘mechanism’. The term logic better avoids the connotations of subject-independence that talk of causal laws and mechanisms suggest. At the same time, it allows us to maintain the central insight which promoted the turn to mechanisms in the first place, namely, that not all is reducible to the contextualized self-interpretations of subjects: logics are thus meant to capture the subject-dependent aspect of social processes, as well as aspects which are not reducible to the empirical context (Glynos, Howarth, 2007: 97).
While Elster’s theory of causal mechanisms responds to certain limitations of the causal law paradigm, he nevertheless accepts the search for laws as an ideal. And on of the reasons for this is the atomistic ontological grounding of his account, in which the world consists of discrete events, facts, and mechanisms. 103