I just read John Earman and John Roberts’ interesting papers defending Humean Supervenience about laws of nature, in PPR 2005. They argue that facts about which generalizations are nomic supervene on non-nomic observable facts. Their argument is that without such supervenience, it would be impossible to find empirical justification for believing that a generalization is nomic (as opposed to something true by coincidence).
It seems to me that their argument relies on the implicit assumption that nomic facts (such as facts about causal relations) are unobservable. For if some nomic facts were observable, observations of such nomic facts may constitute an empirical justification for believing that a generalization is a law, even though Humean Supervenience would be violated.
Now, I know that Hume and many others deny that causal relations are observable. But at least some psychologists, and probably some philosophers, have argued that causal relations are observable. I find it extremely plausible that causal relations and other nomic facts are observable. For instance, I can try to push as hard as I can against the wall and observe, “I can’t break this wall by pushing it”. Isn’t this an observation of a nomic fact?
Does anyone know whether there is any consensus one way or the other about whether nomic facts, such as facts about causal relations, are observable? Does anyone else agree that causal facts and perhaps other nomic facts are observable?
Are Causal Facts Observable?
27 Comments
Comments are closed.
I can tell when someone is sunburned, and you can’t get sunburned unless your skin is burned by the sun. Similarly, I can tell when someone is talking, and you can’t talk unless your body is caused to move by your intentions to speak.
Well, I don’t know about the consensus, but my feeling, having argued for some time now argued that the causal relation is as observable as any other relation, that a lot of philosophers still think that it is unobservable…
What’s so Unobservable about Causation?”
I should clarify that by causal facts I didn’t mean facts that are described with terms that imply causal relations, but facts of the form A causes B. That someone is sunburn is not a causal fact in the present sense. That the sun burnt someone is a causal fact in the present sense.
I would guess that the standard humean argument against the observability of causal facts would go something like this:
causal relations are relations between events, and only events are observable, hence all that can be observed are successions of events, and hence the relations between those events can only be surmised.
I’m not certain I see, yet, what kinds of facts you have in mind. We stay at the beach for 4 hours, I observe you getting redder and redder, I see that you are sunburned. Suppose I now claim, “the sun caused that burn; I have observed this.” Why is this not a case of observing “a fact of the form A causes B”?
That’s it. If that is a genuine observation, it’s an observation of a causal fact in the sense I meant. (Contrast the simple claim that someone is sunburnt. It may be the sun that caused the sunburn, or something else, or the sunburn may just have appeared on her by coincidence right after exposure to the sun.)
Right. Now why should we believe that relations between events are unobservable? Or alternatively, why should we believe that (observable) events cannot be rich enough to encompass causal relations within their constituents? I don’t find these assumptions very plausible.
The idea (and mind you I am only playing devil’s advocate) is that events in general, and observable events in particular, are spatio-temporally un-extended. What you observe (to coin the phrase) is one damn event after the other. Relations between events are spatio-temporally extended. Hence they can only be inferred from the sequences of events that you observe.
The humean rejects that you see the sun causing the burn over four hours as one simple observation. He claims that you see the sun at time t1, and the burn at time t2, and you can only infer that the events at t1 caused the events at t2.
Thanks for playing devil’s advocate. I presume that what you meant to say is that according to Humeans, events are temporally unextended (rather than spatio-temporally unextended, which seems way too strong). But pretty much everything I know about brain processes goes against the view that a temporally unextended event could be observed by an actual brain. Neural processes take time, and they need temporally extended inputs in order to produce internal states that could possibly constitute observations (or the neural basis of observations). I’m not even sure that a temporally unextended event could be detected by anything, whether neural or not. So it seems to me that if events are temporally unextended, then events are unobservable. So if events are observable, they must be temporally extended.
I may be getting out of my league here, but the idea is that “events” which make up the “humean mosaic” are the properties that obtain at space time points (or the properties, if you prefer, of space-time-point-sized particulars).
Of course it is, as you point out, a fact of human psychology (and perhaps a necessary fact of any perceptual being) that we cannot observe such things. But we could treat it as an idealization.
Lets assume for the moment that I could convince you that IF human beings observed humean events (these idealized properties at s.t.points) THEN the unobservability of causal relations would follow.
It would then seem strange to argue that in virtue of our inability to resolve the world in such fine detail, we are able to observe causal relations. How could our ability to observe causal relations stem from the blurriness of our perceptual apparatus.
If you accept that argument, than the idealization of treating our perceptual atoms as humean events is harmless, and we have to deal with the argument I gave in the last post.
There are indeed some philosophers who think that causal facts can be observed. See the work of Rom Harre and Ed Madden, for example their joint work *Causal Powers: A Theory of Natural Necessity.* For psychologists, see the work of Michotte.
I am a scientist (biochemist) absent from the university 20 years and have recently pursued a study in philosophy. Science education is very restrictive and does not consider history. In my own mind I have reduced all things(including cause and effect) to the word witness(of both unique witness A and the act of his(its’) witness =A).A chain is thus described A.B.C.D …over time. This must be true even for temporally unextended events. A question can be posed-can a volume of space exist that is closed such that a witness process within it has no external record, and I do not think that is possible. Between the past and the present, in general, the existence of a closed space imply the alignment of all witness processes from an origin. Thus if the molecular interactions within the brain are chains of witness, thought produces a sum change of some type rather than saying A causes B. Like the layers of an onion, it is conceived from a differential..previous change.. and relates to itself that way…relates externally that way and cannot be detected if it did not occur..e.g. an external relation appropriate to the relations of another set of layers. If such relation did not occur,it did not exist; is the suggestion of whether such a relation could be detectable any different from the suggestion of being able to detect what does not exist. This is very different from a notion that “does cognition alone produces detectable or undetectable change”. The way one spends his time, where his thoughts rest can change the way he relates–self-relates. This fact might be instantiated to concepts of history.
https://www.marvinekirsh.com
https://philosophy.elte.hu/philphys_archive/2007/msg00015.html
I think a better way to give the arguments I was skirting around in the last post is this:
1. The world is a humean mosaic. It is composed of properties instantiated at space-time points (or space-time-point-sized particulars with properties). Call these humean events.
2. Causal relations are relations between Humean events.
3.Whatever it is we are capable of observing, (however fine our perceptual apparatus is) these observations are of patches of the humean mosaic.
4.Hence whatever is observed supervenes on the humean mosaic. An identical patch of humean events will produce the same observations in the same observer.
Now we can argue directly for the ER conclusion.
If two worlds were humeanly identical, but differed with regard to their nomic facts, they would be observationally identical. etc.
Of course: 4 is precisely what you want to deny. But what’s really doing all the work is 2.
Hi Gualtiero,
It may be of some use to you to know that the unobservability of causation was questioned by G.E.M. Ansocombe. I don’t recall what the article was, but I do recall the example of a dog and perhaps some biting. Maybe some one else knows what I’m talking about.
Thanks everyone for your comments and references.
Eric, thanks again. I like your new argument. It seems to rely on more metaphysical assumptions than the original one, and of course those assumptions need justification. But even if I were to grant premises 1 and 2, the argument goes through only under the further assumption that while observing patches of the Humean mosaic, we cannot observe the causal relations between the sub-patches. What justifies that?
Yep, Pete is right about Anscombe, the paper is called “Causality and Determination”, there is also Ducasse’s paper “On the Nature of Causation and the Observability of the Causal Relation,” though both seem to me to think that what we need to observe is THAT the relation is necessary, which I find implausible…Anscombe even goes so far as to suggest that since we do not see THAt the relation is necessary (but yet we do see it) then it must be the case that causation is not necessary, which seems to me even more implausible…
Hi folks + Gualtiero,
A brief question: What is the notion/method of individuation (of those causal “observable” relations) you are talking about – for instance here:
“… through only under the further assumption that while observing patches of the Humean mosaic, we cannot observe the causal relations between the sub-patches”
Gotta go, a
Yes, Robert is probably right about his miss Anscombie- interpretation. Those were the days when people were really enthusiastic about the de dicto-/de re- distinction (seeing x and seeing that X) in England.
By the way, nice to “meet” you, Robert. Greetings from sunny Helsinki,
Ansku
Does it though? I dont think so. Even if you can pick out the causal relations between the points in the patch, (the sub patches) as long as what you are observing supervenes on the events within the subpatch, then E&R’s conclusion is going to follow. There will never be humeanly identical worlds that yeild observationally different evidence. and that’s all they need.
And I think (though I wont spell it out here) you can argue from 1 and 2 to 3. (you dont need to suppose it)
In other words, the force of 1 and 2 is to commit you, essentially, to the idea that the events are all there are in the world to have observational content. the claim about your inability to observe anything else is not a psychological claim, its a metaphysical one.
Hi! Nice to cyber-meet you as well. Sunny? Sounds nice. Here in NYC it is cold and rainy; some Spring!! Am I observing Global Climate Change causing this very bad weather?
…one small matter, though: it’s ‘Richard’ not ‘Robert’ 🙂
It seems to me that one can accommodate (a) the reasons for thinking that causal relations are sometimes observable with (b) the Humean observation that information about causal relations is not directly given to us in our perceptual data, by taking seriously the von Helmholtzean idea that perception is a kind of unconscious inference. Sure, we INFER the existence of causal relations. But then we infer the existence of objects, too. This does not impugn the idea that objects, or causal relations, are observable.
Richard,
In Finnish: Voi, olen pahoillani… (“Oh, I am sorry”) Anna anteeksi. (“Please, do forgive me”)
What was I thinking…??? (“Robert”, oh my god – how embarrassing is this?)
Ouch,
a
Somewhat tangentially, I think the von Helmholtz idea is a bit of a red herring. We can infer causality without perception itself being an inferential process. Recently, Zenon Pylyshyn in two books (2003, 2007) argues that vision is an encapsulated process and recovers signals more akin to what Sperber and Wilson call a ‘coding’ as opposed to an ‘inferential’ process. However, the fact that vision is not inferential in character does not entail that other parts of cognition aren’t, as it’s a putatively modular system. Serious cognitive explanations both about how we determine when a correlation is a cause (ex. Patricia Cheng’s 1997 “Causal Power” theory) and how we use causal judgments in explaining other phenomena in the world (ex. Glymour’s 2003 book “The Mind’s Arrows”) are explicitly inferential, and the idea that vision is not does not impugn on this fact.
I’m not sure if the recent work about causation and inference touches on whether causes are observable or not. I always read them as investigating the Humean problem of “how we get our idea of cause,” which is noncommittal on metaphysical status in general, and observability in particular.
-Jeremy
Just a reference: Susanna Siegel has a great paper on whether causal relations can be visually experienced (she says they sometimes can) that seems highly relevant to this discussion:
https://www.people.fas.harvard.edu/~ssiegel/papers/mscmay06.html
There are at least two questions which need to be disentangled here. One is whether a theory of causation can be given which shows how causal facts supervene on non-causal facts (ideally, on non-modal facts). Another is whether causal facts can sometimes be known non-inferentially. The questions seem to me to be independent of one another, so I am confused about how the answer to the second could shed light on the first without some auxiliary assumptions.
I guess in Hume’s case it is his empiricism which bridges the gap: the idea of necessary connection is incoherent since not capable of being obtained through impressions, so the truth conditions for causal ascriptions must be the regularities that are the source of whatever ideas of compulsion we do have. Or something like that.
What is playing the role connecting these issues in the Earman and Roberts paper? A similarly empiricist notion of observation?
I think this is right. There is a huge difference between an analysis of the causal relation and determining whether or not we can observe the causal relation. I think that in Hume’s case what has happened is that he thinks that in order to see the causal relation, he has to see THAT it is NECESSARY and certainly he does not get that from experience (even Kant agrees with him on this point). I pointed out earlier how Anscombe makes the same kind of mistake, she thinks that since we do not see THAT the causal relation is necessary, that must mean that causation is not a necessary relation. But this is to confuse the two questions just distinguished. Whether or not the causal relation is necessary or not is irrelevant to determining the answer to he question ‘can I see the causal relation?’. It seems to me perfectly obvious that we do in fact see the causal relation every where we look. It is that thing that we observe that we thengo on and try to give an analysis of.
Oh, I just noticed this…don’t worry about it! After all, what’s in a name? :=)
(Though I think that ‘Robert’ is supposed to mean something like ‘strong protector’ and ‘Richard’ ‘strong ruler’…which does tend to make one wonder if the Millian idea that names denote but do not connote is just plain silly…)