Mechanisms sans Metaphysics

1. Mechanisms sans Metaphysics Stathis Psillos University of Athens Greece

2. Two approaches to causation One is less metaphysical, the other is more metaphysical • The first conception of causation: dependence or successionist. • Causation is a relation between two externally related events and is such that the cause is robustly related to the effect. On this conception, cause and effect are genuinely distinct events: their conditions of existence are not intertwined. Yet, they are related in such a way that their co-existence is not accidental. • Several ways to unpack this idea of robust (non-accidental) dependence (nomological, counterfactual, probabilistic). The key idea behind all of them: that causal relations need not be grounded in, or explained by, a distinct layer of facts or attributes that are supposed to enforce the causal connection.

3. The second conception of causation: productive or generative. • The cause produces or generates the effect. On this conception, cause and effect are really connected to each other in such a way that they are not genuinely distinct events. • N.B. On the second conception—strictly speaking—the causal relata are not events. The locus of causation is in substances and their properties. • The productive conception is meant to underpin the modal force of causation—and the accompanying claims of (non-logical) necessity. • Not all combinations of cause and effect are possible. • Given the cause, certain effect(s) has(/ve) to follow.

4. Models of Production Several ways to unpack the idea of a real connection, but the most prominent one (and, ultimately, the common factor in all ways) is that cause and effect are related by a mechanism, which is such that on being stimulated by the cause, it generates the effect. The presence of the mechanism constrains the possibilities: only a certain effect (or a family thereof) is possible, given the cause.

5. Varieties of Mechanism • Something gets transferred—Transference models Wesley Salmon’s process theory of causation I Wesley Salmon’s process theory of causation II • Persistence models--Something persists J. L. Mackie’s version Dowe’s version • Power-based production Generative mechanisms The MDC approach: powerful mechanisms The Glennan approach: powerless mechanisms

6. Something gets transferred—Transference models • But what exactly? Father Nicholas Malebranche Search After Truth (Researche de la Vérité), 1674-5 “There are some philosophers who assert that secondary causes act through their matter, figure, and motion (...) others assert that they do so through a substantial form; others through accidents or qualities, and some through matter and form; of these some through form and accidents, others through certain virtues or faculties different from the above. (...) Philosophers do not even agree about the action by which secondary causes produce their effects. Some of them claim that causality must not be produced, for it is what produces. Others would have them truly act through their action; but they find such great difficulty in explaining precisely what this action is, and there are so many different views on the matter that I cannot bring myself to relate to them” (1997, 659).

7. There is need for something X such that a) commitment to it is independently plausible and b) there is a relatively clear idea as to how it gets transferred (without draining the cause, or without losing its own identity etc.). • At the same time, the X that gets transferred should be such that causation is prevalent and widespread: it does not end up being an idiosyncratic feature of whatever entities happen to possess X; nor should it end up being constrained only to some levels of reality (e.g., only to fundamental reality).

8. Plausibility vs generality of X The tension. Not easily satisfiable together. Plausibility tends to come from science and its conceptions of the world—what gets transferred is a physical quantity (e.g., energy etc.). But this does not yield generality; or else, it does yield generality only if some form of reductionism is adopted: all causation is, ultimately, physical. More precisely: Three desiderata: • Plausibility (a plausible account of what gets transferred in causation) • Generality (the causal connection should be generalisable; it should obtain in all kinds of domain) • Autonomy (higher-level causation should have some kind of autonomy; it should not be parasitic on causation at the physical level) These three cannot be satisfied together by transference models.

9. Salmon’s process theory of causation I “Causal processes, causal interactions, and causal laws provide the mechanisms by which the world works; to understand why certain things happen, we need to see how they are produced by these mechanisms” (1984, 132). • Structure-transference (or more generally, mark-transmission) becomes the distinguishing characteristic of a mechanism. This can restore Autonomy, since there is no restriction as to what kind of thing a mark is—or on what kind of structure should be transferred. • But: The very idea of mark-transmission cannot differentiate causal processes (and hence mechanisms) from non-causal ones, since any process whatever can be such that some modification of some feature of it gets transmitted after a single local interaction. So: to restore Autonomy, Plausibility is sacrificed.

10. Salmon’s process theory of causation II Conserved Quantity theory: transference of CQ • Plausibility is restored, since conserved-quantities are well-understood. • But Autonomy is sacrificed again. Even if it were granted that this conception offered neat accounts of causality at the level of physical events or processes, it can be generalised as an account of causality simpliciter only if it is wedded to strong reductionistic views that all worldly phenomena (be they social or psychological or biological) are, ultimately, reducible to physical phenomena.

11. Persistence models--Something persists There is no transference any more, but is this genuine progress? J. L. Mackie’s version • The causal mechanism consists in the qualitative or structural continuity, or persistence, exhibited by certain processes, which can be deemed causal. There needn’t be some general feature (or structure) that persists in every causal process. But: this account borders on vacuity. Something always does persist (e.g. the shadow of a moving car). The notion of persistence is unhelpful unless there is a suitable characterisation of the properties, or the features, that persist. • Causal interaction is unaccounted for. When two processes interact, there may well be nothing that persists. The three desiderata: • Plausibility • Generality • Autonomy Persistence accounts satisfy Generality and Autonomy, but fail Plausibility.

12. Dowe’s version Possession of conserved quantities “The central idea is that it is the possession of a conserved quantity, rather than the ability to transmit a mark, that makes a process a causal process” (Dowe 2000, 89) But: anything can possess some conserved-quantity • Hitchcock’s counter-example: a moving shadow that is cast on a metal plate, which has a uniform charge density on its surface; the shadow will then possess a quantity of electric charge, which is a conserved quantity. • Reply: a necessary condition for a process being causal is that it exhibits identity through time. • But couldn’t the transference of a CQ ground the identity-over-time that a process possesses? How is identity-over-time to be grounded? Dowe takes it to be a primitive concept. It then seems that the following holds • transmission of a conserved quantity P = possession of a conserved quantity P + identity over time of the object that possesses P. So no genuine progress. The three desiderata: • Plausibility • Generality • Autonomy Dowe’s Possession account satisfies Plausibility and Generality, but sacrifices Autonomy.

13. Power-based production Nothing gets transferred from cause to effect. • Causes produce their effects because they have active powers to bring them about. Causation is productive precisely because the cause has the power to produce the effect. Leibniz: causes are ‘producers’ • Causal powers—in standard conceptions, anyway—constitute a deeper layer of reality, behind or beyond the constant succession between events of certain type. • Causal powers as regularity-enforces. • A causal power enforces (when it is exercised unimpeded) that the causal sequence is invariable; it ensures its future extendability. • But also: powers are meant to explain (at least partly) the deviation from regular behaviour—they can exist unmanifested/unexerted; they can be blocked from producing the standard effect/ masked etc. • Do powers explain anything in the end? They are meant to explain why there is regular behaviour, insofar as there is. But they also explain why there is not, insofar as there is not. • Powers explain why things tend to behave in certain ways. But tending to behave does not entail actual behaviour and is consistent with any actual behaviour. So powers do not explain actual behaviour.

14. Mechanisms to the Rescue Mechanisms add explanatory force to powers. They specify the ‘mode of producing’ (Leibniz): how powers produce their effects. e.g., alcohol has the power to intoxicate specifying the mechanism by which this occurs. Voltaire’s objection—in essence the problem is this. • To say that x has the power to produce y is not to say much unless the mechanism by which it does so is specified. Both poppy and lotus blossom essence are soporific, but by different mechanisms • Generative mechanism: Powers + mechanisms “The generative view sees materials and individual things as having causal powers which can be evoked in suitable circumstances.” Harré (1975, 121) “The causal powers of a thing or material are related to what causal mechanisms it contains. These determine how it will react to stimuli.” Harré (1975, 137).

15. According to Harré (1970, 85), the ascription of a power to a particular has the following form: X has the power to A = if X is subject to stimuli or conditions of an appropriate kind, then X will do A, in virtue of its intrinsic nature. • Power-ascriptions are analysed in terms of a) a specific conditional (which says what X will or can do under certain circumstances and in the presence of a certain stimulus); and b) an unspecific categorical claim about the nature of X. “To ascribe a power to a thing or material is to say something specific about what it will or cando, but to say something unspecific about what it is. That is, to ascribe a power to a thing asserts only that it can do what it does in virtue of its nature, whatever that is” (Harré & Madden 1975, 87) • The claim about the nature of X is unspecific, because the exact specification of the nature or constitution of X in virtue of which it has the power to A is left open. (Discovering this is supposed to be a matter of empirical investigation.) Harré stresses that the above way to analyse ascriptions of powers commits to an “unspecific categorical referring to the nature of constitution of the thing or material concerned” (1970, 90).

16. This implies that ascription of powers is explanatorily incomplete unless something specific is (or can be) said about the nature of the thing that has the power. Otherwise, power-ascription merely states what needs to be explained, viz., that causes produce their effects. This is where generative mechanisms come in. Harré (1970a, 276): “In the ascription of powers the categorical component is like a promissory note, we only believe that it is not logically impossible for it to be cashed. In order for it to be proper to ascribe a power we must believe only that it is in principle possible to ascertain the nature of the subject”. Specifying the generative mechanism is cashing the promissory note. As he put it: “Giving a mechanism … is .. partly to describe the nature and constitution of the things involved which makes clear to us what mechanisms have been brought into operation” (1970a, 124). • So the key idea is this. Causes produce their effects because they have the power to do so, where this power is grounded in the mechanism that connects the cause and the effect and the mechanism is grounded in the nature of the thing that does the causing.

17. A Liberal Conception of Mechanism Movement towards a broad and liberal conception of mechanism. • Quasi-mechanisms • Harré: does not “intend anything specifically mechanical by the word ‘mechanisms’. Clockwork is a mechanism, Faraday’s strained space is a mechanism, electron quantum jumps [are]. . . a mechanism, and so on”. • Generative mechanisms are taken to be the bearers of causal connections. It is in virtue of them that the causes are supposed to produce the effects. A generative mechanism is virtually any relatively stable arrangement of entities such that, by engaging in certain interactions, a function is performed, or an effect is brought about.

18. More specifically, a mechanism is a complex system that consists of some parts (its building blocks) and a certain organisation of these parts, which determines how the parts interact with each other to produce a certain output. The parts of the mechanism should be stable and robust, that is their properties must remain stable, in the absence of interventions. The organisation should also be stable, that is the complex system as a whole should have stable dispositions, which produce the behaviour of the mechanism. • Thanks to the organisation of the parts, a mechanism is more than the sum of its parts: each of the parts contribute to the overall behaviour of the mechanism more than it would have achieved if it acted on its own. The priority of the parts over the whole—and particular, the view that the behaviour of the whole is determined by the behaviour of its parts—is the distinctive feature of this broad quasi-mechanical account of mechanism. The focus is not on the mechanical properties of the parts, nor on mechanical principles that govern the behaviour of the parts and determine the behaviour of the whole. Instead, the focus is on the causal relations there are between the parts and the whole.

19. The three desiderata: • Plausibility • Generality • Autonomy It seems we have found a way to satisfy all three of them. But have we? Let’s look at the details.

20. Monistic vs Dualistic Conceptions of Mechanism Glennan (2002, S344): “A mechanism for a behaviour is a complex system that produces that behaviour by the interaction of a number of parts, where the interactions between parts can be characterized by direct, invariant, change-relating generalizations”. • ‘direct, invariant, change-relating generalizations’: it connects mechanisms with Woodward’s replacement of strict laws with generalisations (or relations) that remain invariant under actual and counterfactual interventions. This view might be called monistic since it does not aim to account for the productivity of the mechanism (the feature of the mechanism in virtue of which the cause produces the effect) by positing a distinct kind of entity. But there is a dualist conception of mechanism Machamer-Darden-Craver (2000, 3) “Mechanisms are entities and activities organised such that they are productive of regular changes from start or set-up to finish or termination conditions” • MDC: a mechanism consists of two distinct kinds of building blocks—entities (organised in a stable way into a spatio-temporal pattern) andactivities. the concept of activity: to account for the interaction between the parts of the mechanism and its overall causal efficacy. • MDC: an appeal to causal laws, or to invariant generalisations, fails to capture the productivity of a mechanism. This productivity, they say, “requires the productive nature of activities”. • Activities are the ontic correlates of (transitive) verbs and are necessary for the grounding of the productivity of mechanisms. Mechanisms are supposed to be “active”: “they do things”.

21. Activities and Kant What are activities? • Activities as the exercising of causal powers. Eric Watkins on Kant • Kant had a distinctively non-Humean conception of causality: Causation as an exercise of the causal powers of substances Substances are active • “the activity of a substance is the temporally indeterminate exercise of its causal power in accordance with its nature and circumstances so as to bring about a determinate effect” (2005, 393) Activity marks the difference between event-based causation and agency-based causation (power-based causation) • “(…) the exercise of a casual power is temporally indeterminate activity that is constrained by a rule provided (1) by the nature of the thing whose causal power is exercised, (2) by the relations the thing stands to other things, and (3) by their natures” (2005, 396). This, as Watkins notes, is akin to the idea of generative mechanism. Harré & Madden (1975, 131) “There is an ontological tie that binds sequential events together, but it is not event-like. It is the persisting generative mechanism consisting of powerful particulars and natural agents which produces the sequence of events and states and endures throughout.”

22. Activities add power to mechanisms Perhaps it’s best to see the difference between the two accounts of mechanism in terms of the distinction between powerful vs powerless mechanisms. The MDC approach: powerful mechanisms But there is a problem: • What explains/grounds the powerfulness of the generative mechanism as a whole? The powers of its components. • But what explains/grounds the powers of the components? Further generative mechanisms • Etc. Are we in for regress? • Need to appeal to bare powers (Parmenidean individuals)

23. Harré is alive to this problem. “A regress of explanation is closed, in a proper way, by adverting to entities, individuals, and materials, which are characterised solely by their powers; that is, specifically by what they can do, which we do know, and only unspecifically by what are, which for the entities which close regresses of explanation we do not know. To ascribe a power, to characterise by powers, is to open the question of the nature of things, without being obliged to answer it. The ultimates can be bare powers, indeed for rational discourse, they must be bare powers, since to say that some thing or material is of an ultimate kind is to refuse, at least for the moment, to offer any account of its intrinsic nature” (1970a, 283) Pure or ungrounded powers—let’s not fuss much about it now. Let’s face up to some consequences of accepting them.

24. Ascription of powers amounts to a promissory note that needs to cashed by means of generative mechanisms. Here is a (sketch of an) argument • (1) Powers require generative mechanisms in virtue of which they are exercised Better: (1’) If powers are to be explanatory, they must be connected to mechanisms by means of which they bring about their effect (2) Ultimate powers, being pure, are not connected to mechanisms Therefore, ultimate powers are not explanatory. • If (1) is denied, then powers are promissory notes • If (2) is denied, then a regress is in the offing. So the friends of powerful mechanisms have to live with either • The undermining conclusion • Or the claim that some powers are promissory notes • Or a rather painful regress.

25. What if the friends of powerful mechanisms simply accept that there are (must be) ungrounded powers? • Let’s go back to the three desiderata: Plausibility Generality Autonomy • MDC-mechanisms have either to flout Autonomy (since the powerfulness of generative mechanism has to, ultimately, depend on ungrounded powers at the fundamental level—the only plausible candidate for ungrounded powers) or to flout Plausibility (since they might have to be committed to ungrounded powers at higher levels) An alternative: forget about the metaphysics of mechanisms and stick to methodology. Refrain from raising and answering the metaphysical question of the grounding of the mechanisms.

26. The Glennan approach: powerless mechanisms • The later Glennan: The interaction of the parts of the mechanism is characterised in terms of invariant, change-relating generalizations. No appeal to powers and/or activities Instead, interventionist counterfactuals explain (or ground) the laws that govern the interaction of the parts of the mechanism. • Mechanisms (e.g., the thermostat) are supposed to explain why certain counterfactuals hold, e.g., if the temperature had risen, the furnace would have turned off. And similarly, the breakdown of a mechanism would explain why certain counterfactuals fail to hold. • Mechanical laws are those whose holding is explained by the presence of mechanisms.

27. Two worries/problems • First: what is the relation between laws, mechanisms and counterfactuals? • It seems that mechanisms ground certain laws (the mechanical laws) by grounding certain interventionist counterfactuals. But it is also that certain interventionist counterfactuals render some (mechanically explicable) generalisations laws. For if they did not hold, we could not talk about a mechanism proper (since we could not explain/ground the interactions between the parts of the mechanism) and hence we could not talk about a law proper. • Call it a loop and not a circle! It is not clear where it can be broken so that the relation between mechanism and interventionist counterfactuals can get going.

28. Second: Perhaps, the loop is broken at the level of fundamental laws of physics, of which Glennan says that they are not mechanically explicable. “all laws are either mechanically explicable or fundamental, tertium non datur”. • But then, if fundamental laws are not mechanically explicable, and if they too support counterfactuals (as they certainly do), it is not necessary for the truth of a counterfactual that there is a mechanical explanation of it. So (the truth of) a counterfactual conditional need not have a mechanical explanation.

29. But is the presence of mechanism sufficient for the truth of a counterfactual conditional? • Glennan (1996, 66) says: Although the mechanism responsible for connecting two events may supervene upon other lower-level mechanisms, and ultimately on mechanically inexplicable laws of physics, it is not these laws which make the causal claim true; rather it is the structure of the higher level mechanism and the properties of its parts. • This simply evades the problem. For how can it be that Y supervenes on X, but facts about X are not part of the truth-makers for claims about Y? Even if they do not fix the truth-makers of claims about Y, they certainly contribute to them.

30. Let’s go back to the three desiderata: Plausibility Generality Autonomy • Glennan-mechanisms flout Autonomy (at least in a strong metaphysical sense). Here again, an alternative would be to leave metaphysics behind and look to methodology. Even if mechanisms do not have metaphysical autonomy vis-à-vis fundamental non-mechanically explicable laws (and counterfactuals), they have methodological independence.

31. A moral Both approaches to mechanism (powerful vs non-powerful) agree that it’s not (cannot be) mechanisms all the way down. • But MDC—brute powers • Glennan: non-mechanical laws Given plausible assumptions, there is a residual problem: how is the fundamentally non-mechanical ground level relate to the mechanical higher levels? This question creates certain metaphysical anxieties. • Are there ungrounded powers? Should a theory of mechanism be committed to them? • How do fundamental laws contribute to the truth of claims about mechanisms? These might be alleviated if we leave the metaphysics of mechanisms behind and think of mechanism in methodological terms.

32. Insofar as mechanisms are taken to be the locus of causation, there is little, if any, hope that there is anything that can play this role; let alone anything along the lines of the diverse things that are identified as mechanisms by philosophers and scientists. • Insofar as mechanisms are taken to be stable explanatory structures (whose exact content and scope may well vary with our best conception of the world) which enhance our understanding of how some effects are brought about or are the realisers of certain functions, mechanisms can play a useful role in the toolkit of explanation.

33. Addendum • Glennan (1996, 56): “a relation between two events (other than fundamental physical events) is causal when and only when these events are connected in the appropriate way by a mechanism”. • A quick problem: it follows that the parts of mechanism (or the parts of the parts of the parts of …of the part of the mechanism) are not connected causally with each other since there is no mechanism that connects them—unless the whole idea of mechanism is trivialised (mechanisms all the way down). • In any case, Glennan denies that it is mechanisms all the way down. Fundamental laws are not mechanically explicable—there is no mechanism that underlies them.

34. Only when? • Putative counterexamples—causation by disconnection; causation by absences. • But also: the sinking of the Titanic and the collision with the iceberg. Causation yes! But no mechanism (at least no non-trivial mechanism). • No causation at the fundamental level—where there are no mechanisms?

35. When? • Notice: ‘events are connected in the appropriate way by a mechanism. If this is omitted, anything admits of a mechanical explanation in terms of M.   E.g. Shadow and height of pole. There is a mechanism that connects the length of the shadow and the height, but the shadow does not cause the height. • Presumably, the events are not connected in the appropriate way. But what makes a way appropriate? •  Whatever it is (call it X), it is only mechanism + X  causation.