What follows is a direct reaction to the discussion I’ve had with Jochen below my first epiphenomenalism post. The discussion meandered into scary places, and Jochen ended up recommending me to keep it simple instead. I’m sure it is a good idea, so I’ll tackle what I think is Jochen’s main point and cut out all the discussion about metaphysics, ontology and epistemology. The latter subjects will need to be considered separately. In what follows, I’ll try summarising the crucial points in a way that is intended to make it superfluous to read the full original discussion.
As far as I can understand, Jochen’s main objection is (happy to be corrected – there is another secondary objection which I tackle in the conclusion):
[The anti-epiphenomenalist in an epiphenomenal world] would propose their theories, perhaps accounting for all the available evidence; but they still would be wrong: in fact, there is epiphenomenal experience in the world, that the theory of the anti-epiphenomenalist simply fails to account for.
What he’s getting at is a rejection of my own key point, where I propose that if something is an epiphenomenon, we can safely consider it non-existent. Jochen rejects this point via three counterexamples:
As I believe to have demonstrated, it’s possible to construct scenarios where your justification doesn’t hold up — the parallel universes, the lone astronomer who can’t detect the universe’s expansion, and the observers in front of their screens.
In a nutshell: there are situations where theorising about epiphenomena makes some sense, and importantly, not doing so may produce a theory which is factually wrong. Hence it’s possible to find some situations where, if one wants to be right, including epiphenomena is necessary. Moreover, Jochen also implies that the case of Phenomenal Experience (PE) might fall into this category. Thus, his attack has two arms:
- In some situations, we can’t safely regard epiphenomena as non existent (if we want to be right).
- The case of mental phenomena and PE might be one of those situations, so my overall claim is rejected.
I believe that trying to reject the first arm would be very interesting, precisely because it requires to tackle a wide range of claims that I’m already committed to, and because these claims are all, in Jochen’s words, “big f’n deals”. I agree they are big and controversial deals, and that’s why I would be very happy to discuss them – another time, perhaps. For this post, I’ll accept the suggestion to keep it simple, and limit myself to the second arm. The aim of this post is to show that even if we accept statement 1. (I don’t, but will accept it for the sake of the argument), it is still manifestly not the case that PE might be an epiphenomenon (i.e. statement 2. isn’t true). To demonstrate this, I will use the three cases made by Jochen, and then show that theorising about PE can’t fit in the pattern they exemplify.
To get started, I’ll use a generic schema of how epiphenomena can get included in a given theory. I will then show how the three cases offered can easily be matched with elements in the general scheme. Finally, I’ll show that PE can’t: it is a radically different scenario. Hence, statement 2. will be rejected, without having to thread in treacherous metaphysical waters.
Building a theory:
We have a phenomenon X, something that happens and is manifest in our world. We want to create a theory T which “explains” X. In typical scientific setting, a minimum requirement for T would be to set out the conditions that make X happen. At minimum, a successful T can be used to predict if, under known circumstances, X is present/exists/will happen. As it turns out, sometimes producing T also allows to hypothesise the existence of some epiphenomena E.
We now need to verify if this definition easily captures the cases proposed by Jochen.
Case 1: parallel universes.
The topic actually subdivides in at least 5 distinct hypotheses; for a high-level description, see here, this WikiPedia page offers some more detail and many pointers to the original literature. Depending on the different branches, our X can be evidence coming from cosmology (say the measurable fact that galaxies appear to move further and further apart), the spooky phenomena that emerge at tiny scales, summarised in what we call quantum theory, or both. In all cases, our resulting theory T can be either extended or interpreted in a way which entails the existence of Parallel Universes, where these are truly parallel, so they have no causal connection with our own (they count as strong epiphenomena to us). Thus, building a theory T, based on evidence X, produced the side effect of hypothesising the existence of strong epiphenomena. If one of these extended theories is right, Jochen argues that, for correctness’ sake, we need to regard epiphenomena as existent. This accounts for point 1., to move onto point 2. we need to figure out whether PE follows the same path (below).
Case 2: expanding universe.
The theory here is cosmological, and builds on accepted notions (our X) such as: the speed of light is constant and can’t be surpassed, the universe is expanding, etc. You put these things together and realise that parts of the universe will eventually get so far apart that they will lose all causal contact, becoming strong epiphenomena to one-another – in fact, for all we know, the universe might already be like this. Since the diverging parts will keep existing, point 1. is reached. In this case, X is cosmological, and once again we can (easily) use the theories T which account for X to predict/hypothesise the existence of stuff which is epiphenomenal to us.
Case 3: observers in front of their screens.
This case is trickier, because it’s synthetic, and the only aim that we know of is “finding the right description”, but we don’t know why, or what our initial evidence is. The basic setting is: there is a computer doing its thing (computing something), the computer has a monitor, which in some way displays/reports the computation. Something/someone observes the screen, but can’t in any way influence the computation or what is shown on the screen. We want to fully describe the whole scene. What is left unspecified is what our X is (the data we have), which in turn would allow to better understand what our description (T) needs to look like. I can see two different scenarios:
Case 3.1: we know about the observer, which isn’t epiphenomenal to us (the theory builders). X in this case is the evidence needed to describe the whole scene above. T becomes what we need to learn in order to describe the computation, what is shown on the screen, and what the observer sees/detects – T may go as far as describing the effects of what is seen/detected. This case is interesting for reasons we’ll see below, but doesn’t include epiphenomena (so doesn’t directly apply): the observer can’t influence the observed, but ex-hypothesis the observer can be detected by us (otherwise the scenario is inherently mysterious: it doesn’t explain how it’s possible that we want to describe the seeing/detecting when we have no reason to believe it’s happening).
Case 3.2: we don’t know about the observer, which is epiphenomenal to us (the theory builders) as well as to the computation. This case better resembles a standard scientific situation, where by definition we don’t know what’s out there and want to understand as much as possible. X in these setting is some evidence about the computer, the screen and perhaps the computation, but we have no direct evidence about the observer. Building T would allow us to figure out that there is a computer with a screen attached, that the computer is computing something and that this computation produces some visible effects on the screen. We eventually may also figure out that there is no input device able to change what is happening inside the computer and on the monitor. Having such a T, would make it somewhat trivial to extend it and hypothesise that, since there is a screen attached showing some stuff, it’s likely that the whole thing was installed because something/someone is going to look at the screen. We wouldn’t be able to say what this something/someone is, but we could hypothesise its existence. If indeed an observer exists, we would be right, so including an epiphenomenon would be justified. Again, this supports statement 1. above, we still need to verify if this kind of scenario applies to PE.
In summary: the directly applicable cases (1, 2 and 3.2) start with some detectable, not-epiphenomenal evidence X, produce a theory T of X and as a consequence, by inference or some other non-empirical means, hypothesise the existence of some epiphenomenon E (parallel universes, parts of this universe which are too far away to ever be reached/detected/interacted-with, the observers). In all these cases, if the epiphenomenon in question does exist: not including E in T would be a mistake, reality would fail to be fully described by T.
The next step therefore is to show if the same kind of situation applies to PE: if PE is indeed epiphenomenal, we would be wrong in rejecting its existence. Now, this is not what I wanted to demonstrate: what I want to show is that PE is not epiphenomenal (not in the strong sense). However, the approach above can be used to make my case, by showing that the kind of situation described in Jochen’s examples cannot be applied to PE. To show that this is indeed the case, all we need to do is apply the pattern shown by Jochen’s scenarios to the case of philosophy of mind: we have some evidence (X), based on this evidence we build a theory T, and T itself then allows us to hypothesise the existence of some epiphenomena, in this case PE. Trouble is, we can’t do this: in our case, the evidence is or includes PE, we start our quest by noticing that pain is painful, that pleasure is pleasurable and that seeing something red is somewhat different from seeing something green.
[Note. I’m not questioning the mystery of PE: it really is puzzling! For example, why is it impossible to explain to a colour blind person the qualitative difference between our perceptions of a green and a red chair? Or, conversely, what’s going on with that dress?]
Anyway, in this context, we have the puzzling evidence of the existence of something, we call this something PE, and want to build a theory T, which would help explaining how PE gets to exist, and therefore would make it possible to understand when PE is present and when it isn’t. One (incomplete) theory that is being proposed is that PE is generated by mechanisms in the physical world (brains, to be precise), but PE itself happens to be a strong epiphenomenon. Clearly, we are already diverging from the paradigm we explored above: the theory is incomplete, doesn’t specify the mechanisms, and at the same time, proposes that our evidence is an epiphenomenon. How can that be? I see two ways to try making this T somewhat coherent:
Escape 1: the PE that is part of our evidence X is something different from the Epiphenomena we are proposing. In this situation, the evidence we have can’t be epiphenomenal (because it’s evidence!), and following the pattern explored by Jochen, we produce a theory which proposes the existence of something over an beyond what’s needed to explain the evidence. Fine, but this epiphenomenon isn’t X, nor is contained in X, so it’s not the PE we wanted to explain. Discussing whether proposing the existence of some other element which is epiphenomenal helps explaining our original X is beyond my aim here (I don’t see how it could be), but under this interpretation, which is supported by Jochen’s examples, we don’t reach the desired conclusion. In this case, the X that we wanted to explain (EP) is not the epiphenomenon. Going one step further, saying that our original X (still EP) is actually the epiphenomenon proposed by T is just a plain mistake and produces an unanswerable question: how did we get to detect the “what is it like” quality of experience if this quality can’t have any effect? Answer: we didn’t, because PE is detectable. In short, under this route PE is not an epiphenomenon, suggesting the opposite is just an insult to basic logic.
To me, this “escape” counts as a self-inflicted bait and switch: we start wanting to explain EP (our starting X, the fact that we do have experiences – whatever we do while conscious, there is a “something it is like” doing it) and end up trying to explain something else (an epiphenomenon E which might also exist) while being mistakenly committed to the idea that EP = E. EP is our bait, what we want to explain, but instead (the switch) we end up trying to explain E, which isn’t what we wanted.
Escape 2: the PE that is part of our evidence X is somehow a-causal. We just know that PE exists, but since it is an epiphenomenon, there is nothing that causes this knowledge. I suspect this is the interpretation that is usually taken seriously, but I must confess that I don’t see how to make it work. Jochen offered plenty of words to support this view (1, 2), but to me, it is spectacularly unsatisfying. The reason is simple: under this interpretation, our theory T may (if completed) explain what is necessary to produce PE. However, at the same time, it would tell us nothing about how we get to become aware of our own phenomenal experience. Because we have now added in our hypothesis the claim that our evidence is epiphenomenal and that we get to know about it in some a-causal way, we have implicitly accepted that we can’t explain how we got our evidence in the first place (if there is no cause, there is nothing to explain!). This would be a dodgy strategy in any case, but is especially unsatisfying in the case of PE, because one could say that the mystery of PE is precisely what we are accepting as unexplainable. After all, we get to know that experience exists by having experiences: the act of experiencing is one and the same as creating the knowledge that experiences exist. How this unity subsist is, under many interpretations of the mind-body problem what makes the problem problematic. Thus, we get to another bait and switch situation (just like in escape 1): we wanted to explain how experience is experienced, but instead, by starting with the assumption that experience is just experienced (there is no reductive way to slice the phenomenon up in sub cause/effect components), we explain something else (at best), while assuming the unwarranted hypothesis that there is nothing that can be explained about how experiencing happens. This escape looks a bit like the scenario 3.1 above, where we simply avoid questioning how we got to know that the observer exists. However, unlike that case, when it comes to PE, how we get to know about it is very much part of what we want to explain, so case 3.1 remains a dead-end, it is inapplicable to PE.
I feel a little self-indulgent, because all of the above adds precisely nothings to what I’ve already written. Moreover, it isn’t even original, for example, Brian D. Earp explicitly says that Chalmers implements his own bait-and-switch strategy (Earp BD 2012, page 17), which, under my approach, correspond to my discussion of scenario 3.2 and/or Escape 2. Thus, if I am proposing something new, it’s merely a new way of making an old claim, with the icing being my belief that we are dealing with a self-inflicted (not deliberate) case of bait and switch. Moreover, the main assumption I’m making is that we are (also) looking for what causes our knowledge of PE, we want to figure out what mechanism produces experiences, or what makes “experiencing” possible. If this wasn’t the case, I would be wrong, but clearly, plenty of people want to figure this out, for example, just a couple of weeks ago Eric Schwitzgebel proposed a particular definition of:
phenomenal consciousness, a.k.a. “qualia” or “what-it’s-like-ness”
In his proposal it is manifestly clear that the intent is produce a definition which isn’t an intrinsic source of trouble (i.e. avoid self-inflicted bait and switch) and the same time, that the starting definition of our explanandum directly stems from non-epiphenomenal evidence (“it’s not neutral, and it depends on there being some obvious feature/property that is shared [between different kinds of experiences]”), furthermore, Ned Block (who, famously distinguishes between access consciousness, and phenomenal consciousness, the latter being what I call PE in here) is taking him very seriously. Thus, even if I abhor making arguments by appeal to authority, in this case I admit that I must: the citations above demonstrate that my main point, the one assumption that is indispensable for the work I’m trying to do (that our knowledge of something, whatever that something is, needs to have a cause – and symmetrically, that explaining something requires to identify what causes that something) may not be uncontroversial, but at the same time refuting it by saying that this assumption is clearly false as it “leads to infinite regress” (as I think Jochen does) doesn’t hit the target. Searching for explanations means looking for what causes our explananda, assuming our explananda has no causes (instead it “just is”), precludes us from ever finding what causes it. Since I can’t see a way of making PE epiphenomenal without also accepting that our knowledge of PE isn’t caused by PE, it directly follows that PE isn’t epiphenomenal – that’s because PE is, by definition, at least a necessary condition for our experiencing, while our experiencing causes us to want to understand how it all happens.
Brian D. Earp (2012). I can’t get no (epistemic) satisfaction: Why the hard problem of consciousness entails a hard problem of explanation. Dialogues in Philosophy, Mental and Neuro Sciences, 5 (1)