3.2 Beyond the MDC Account of Mechanism
3.2.3 Regularity
3.2.3.2 One-Off Mechanisms
bution of indeterminacy” or “Known interfering factors” (Andersen 2011). Neither of these conditions supports one-off causal chains. Any mechanism for which there is a known sta- tistical distribution of indeterminacy must be one that has succeeded multiple times, so one-off mechanisms are not covered by that clause. One example she gives of a case where known interfering factors are responsible for a mechanism not bringing about its end state
reliably is the attempt to produce Higgs bosons in CERN’s Large Hadron Collider. In that case, the mechanism that should produce the bosons is part of a well-supported scientific theory, but producing Higgs bosons at all, and what is more, producing them in sufficient quantity for reliable measurement, is a process hindered by multiple interfering factors of a financial, practical, and political nature. Again, this is not a one-off causal chain. In fact, Andersen explicitly excludes one-off causal chains from being mechanisms; she states, “there exist causal chains that only occur once and thus are not mechanisms” (Andersen 2011). I agree that there might be causal chains that shouldn’t count as mechanisms, including some that occur only once, but I think Bogen is right that occurring only once is not what makes an instance of causation fail to be mechanistic.
DesAutels agrees that jettisoning regularity completely from the account of mechanisms is a mistake. He argues that without regularity, mechanism becomes a term so wide in scope as to cease to be very useful as a scientific concept. Getting rid of regularity means that “any singular causal chain seems to be allowed to count as a mechanism” (DesAutels 2011). As an aside, I do not find DesAutels’s rather brief arguments for why irregular mech- anisms could not be used for intervening and predicting convincing. He suggests that if mechanisms can be singular causal chains, any single result of an intervention would be attributed to a unique mechanism, so could not ground predictions of further occurrences, and likewise interventions could not be used to test general claims. This seems too strong. Even with a regularist account of mechanism, doing a single intervention isn’t usually a good test of a general claim, and single experimental results are only very rarely used as the foundation for predictions. Scientists repeat their experiments when possible, and gather statistics about how frequent different outcomes are. These statistics are what ground pre- dictions and test claims. Scientists can still do this if they are not regularists. They can still group together singular causal chains that came to the same result and talk about types of mechanisms and their frequencies of occurrence when making predictions or testing general claims. Allowing singular causal chains as mechanisms does not prevent one from sometimes grouping such chains together, nor force one to consider each individual chain as something sui generis.
as a scientific concept. I think it’s worth a closer look at one-off causal chains, to decide which ones we might want to include as mechanisms and what the criteria for their inclusion should be. Andersen’s two conditions that can allow infrequent occurrences to nevertheless be counted as examples of mechanisms both involve the possession of background knowledge about a type of occurrence. One is knowing statistics about how frequently (or infrequently) it happens; the other is knowing what sometimes prevents it from happening. Although we can’t have this sort of knowledge about a one-off causal chain, we might have this sort of knowledge about the types of causal chains to which the one-off chain belongs or about the causal chains that are its parts. We may not be sure of having produced a Higgs boson in a high energy collision, but we know about these sorts of collisions more generally, and about the production of other sorts of particles thereby. We also have physical theories that include a role for Higgs bosons.
Part of Bogen’s point was that whether something was caused mechanistically by the coordinated activities of its components does not depend on how often it happens, and he allows for the possibility of mechanisms that operate only once. He does not go into detail, but I imagine the sorts of one-off causal chains he might have had in mind as mechanisms are events like the birth of the universe, or the speciation of Homo sapiens, which cosmologists and paleontologists certainly seek to explain, and quite possibly in terms of the coordinated activities of the relevant components. These are also cases where there is a background of scientific theory that the examples can be fit into. The speciation of Homo sapiens is an example of speciation more generally, and we know what some of the general mechanisms for that are. The Big Bang theory of the birth of the universe is likewise connected to theories of star formation and singularities. What makes these examples of one-off chains acceptable mechanisms is their relationship to more general mechanisms.
Some one-off causal chains should perhaps not count as mechanisms. Luka Magnotta murdering his lover, or me falling off my bike and skinning my knee are causal chains that I’m less sure should be called mechanisms. In these less obvious cases, I tend to think it depends on the details and the context of scientific knowledge. Presumably forensic psychologists would be interested in the Luka Magnotta case, so it may well be the subject of scientific explanation, but I would only call the causal chain leading to those events a mechanism if
it relates either to other cases of sadistic murder, or to general psychological principles. If Luka Magnotta ended up murdering his lover based on a series of random events like rolls of a die (however hard that is to imagine), and not because of precipitating circumstances in his psychological life, this might still be of interest to gossips and bloggers, but perhaps not to forensic psychologists. The case might be of scientific interest if it tells us something about the effects of living with abusive grandparents, for example, or if the habit of creating fake online personas turned out to be a reliable warning sign of murderous intent. If in some way Magnotta follows patterns characteristic of other sadistic killers, or if his actions can be explained in terms of principles of abnormal psychology, the causal chain leading to his actions might make for a mechanism. I do not want to suggest that the test for what should count as a mechanism is whether it happens to be of interest to scientists, because this might change on a whim. The rough idea, which I will make more precise shortly, is that the test for what should count as a mechanism is whether the process fits with other scientific theories. The interests of scientists are just an imperfect proxy for this.
Similarly, explaining how I skinned my knee might be a subject of scientific interest to emergency medicine specialists insofar as it relates to other cases of skin abrasions. They might fairly describe skidding along gravel and asphalt as a mechanism explaining a par- ticular pattern of skin injury. Civil engineers might also find this causal chain of scientific interest insofar as it relates to the safety of intersections, and they might, as far as I know, talk about heavy vehicle traffic creating potholes and loose gravel as a mechanism explaining bicycle accidents. But simply as a one-off causal chain, if it were unrelated to more general mechanisms of skin abrasions or traffic accidents, it seems a bit much to glorify as a mecha- nism this episode of skidding around a gravelly corner, falling off my bike, and skinning my knee. The intuitive idea is that a context of scientific theory matters to these decisions of what counts as a mechanism, but how exactly the connections are made between these sorts of examples and scientific theories remains to be clarified.
So far I have used examples to motivate the point that some but not all one-off causal chains should count as mechanisms, and claimed that the important distinction between mechanistic and non-mechanistic causal chains is whether they relate to more general sci- entific principles. Now I will break down the problem further. A mechanistic causal chain
might be unique in either of two ways: because its steps have never before been executed in that particular order, or because it involves a unique step that has never occurred before. 2
Take the first case, where it is previously unknown that a given effect can be produced by performing the steps of a mechanism in a different order, or mixed in a novel way with steps from other mechanisms. If the new combination were nothing mysterious, in that there is productive continuity, and each step reliably produces the right type of intermediate effect, then it seems uncontroversial that this would be considered a variation on an old mechanism, or a newly discovered mechanism.
If the steps consist only of accidental, random events that just happen to combine in such a way as to produce an effect that none of these steps are usually involved in producing, it’s less clear that this should be called a mechanism, although there is no problem with saying that the random series of accidents caused the effect. That it’s a random collection of accidents that bears little relation to known patterns of behavior for those sorts of entities seems to me a stronger reason for denying that it is a mechanism than that it has only happened once. If the same weird constellation of accidents were to happen a second time, I’d still be inclined to deny that it is a mechanism. There may be reasons for preferring mechanism to have a wider scope that includes such random causal chains, butBogen(2005) and Machamer (2004) haven’t said what they are.
The case where a hitherto unknown step occurs is more interesting. It might happen that a step bearing some similarity to a known effect occurs, but where this version of the effect was previously unknown. Early discoveries of light’s wavelike properties provide an example. Grimaldi’s posthumous 1665 publication described for the first time the diffraction patterns produced when light passes through a narrow slit. He concluded light must be a fine fluid in a state of constant vibration. Within a couple of decades Hooke and Huygens proposed precursors to the wave theory of light. Hooke delivered a paper to the Royal Society in 1672 critiquing Newton’s ‘New Theory about Light and Colors,’ and suggesting that light is propagated by waves, like the vibration of a string. Huygens developed a wave theory of light that explained reflection and refraction results, which he first presented in 1678 and
2To even get started with this discussion, we need to allow for steps to be identified in a somewhat
published in 1690. Much later, in 1845 Faraday showed that light’s wavelike properties are related to electromagnetism, thenMaxwell(1873) developed a full mathematical description of the phenomenon, and finally Hertz confirmed the theory experimentally in 1887. Because waves were already a known phenomenon, and even in the 17th century there was some theoretical apparatus for explaining their behavior, this newly discovered behavior of light could be fit into the wave schema. Similarly, Eddington’s observation during a solar eclipse in 1919 that light from stars was deflected as it passed near the sun was, at the time, just a single observation of this phenomenon occurring. Nevertheless, because it fit with available theory, it immediately was taken as a reliable effect.
If a hitherto unknown causal chain occurs, but there is no analog for it to be compared to, no known schema that it fits, no type of phenomenon it can be generalized under, then matters are rather different. The new phenomenon would be considered a mystery to be solved. If trying to characterize it in terms of existing theory fails, the next step would be to try to get it to happen again. If the mysterious effect were found to be repeatable, it would be considered a gap waiting to be filled with a mechanism. The Vikings may have discovered circa 700 AC that Iceland spar polarizes light, and can thereby be used as a navigational device (Alcoz 2012). The sunstone is described in several sagas as a stone that can show the position of the sun on overcast days. The Vikings presumably had no established science of waves to help them understand the phenomenon, so polarizing light with sunstones would not have been understood in a wider context of scientific knowledge in that context. My Icelandic friends assure me that the sunstone is depicted in their sagas as an everyday tool rather than a mysterious object. In this sort of case, that a causal chain is repeatable and reliable is grounds for believing that it must be produced by a mechanism, even if the mechanism is completely unknown.
If neither connecting the event to other phenomena, nor repeating it worked, people would probably be inclined to call it a miracle. Others might try to deny it happened by calling it a measurement error or a malfunction of equipment. In such a case, it would seem strange to call it a mechanism that only happened once, although barring miracles, the effect presumably was caused. Repeated occurrences can ground belief that there must be a hidden mechanism at work, but failing to repeat is not grounds for denying that there is a
mechanism. In that case, a connection to general scientific principles trumps repeatability, and one-off instances of causation are treated as mechanisms.
What is essential about regularity for mechanisms is not how often a causal chain occurs. Repeated occurrence is a good reason for believing there to be a mechanism at work, but is not constitutive of a mechanism. The deciding factor in whether we should call something a mechanism or not, is whether there is a background of theoretical knowledge into which the causal chain can be fit. Mechanisms should be repeatable in principle, or should instantiate more general schemas if mechanisms are to be a useful scientific concept.
3.2.3.3 Causation and Mechanism Although I think Bogen is right when he claims