Wednesday, September 13, 2006

The Russellian Stance & Concluding Thoughts

(This is the final post in a series, for background see the previous posts.) There is a well-known philosophical argument regarding the limits of our knowledge of physical objects and the relevance of this for the problem of consciousness. This argument is most prominently associated with Bertrand Russell, although there are connections with the work of many other philosophers. The argument has two parts. The first part asserts that we only know physical objects by their dispositional, relational, or extrinsic nature. Our knowledge of physical phenomena leaves untouched their categorical, non-relational, or intrinsic nature. In his book, Daniel Stoljar refers to this as the categorical argument. The second part of the argument draws a connection between the (hidden) intrinsic aspect of physical phenomena and the seemingly intrinsic quality of the phenomenal properties of first-person experience. (For other discussions online, see the “Type-F monism” section of this Chalmers paper; an earlier treatment of the view -- “o-physicalism” – in this Stoljar paper; and, for a more extended treatment of related philosophical theories, see this SEP article by Leopold Stubenberg. Also, Chapter 2 of Gregg Rosenberg’s book employs a creative argument in this vein).

Stoljar briefly gives his own account of the categorical argument. He divides physical truths into three exhaustive categories: spatio-temporal truths, truths about secondary qualities, and truths about primary qualities (note that properties from our physical theories like mass and charge would fall into this grouping). He discusses ways these individually or in combination can be seen to fail to tell the whole story about physical objects and specifically only tell a dispositional story about physical things. While some of the details could be contentious, the categorical argument's conclusion appears plausible.

There are two ways to take this argument further. For the first, note how the terrain here overlaps the discussion of the “structure and dynamics” objection which was the topic of my earlier post (although in Stoljar’s book, it comes in a later chapter). One could argue that the categorical argument leads directly to this thesis (Interpretation 1):

1. All non-experiential truths are (and will always be) dispositional.
2. Experience concerns categorical truths.
3. Therefore, non-experiential truths cannot entail experiential truths.

Stoljar’s epistemic thesis would be false – there cannot be non-experiential truths of which we are ignorant which are relevant to experience.

I should note that for a composite system (like a human), it can be said that our experience depends on both dispositional and categorical truths rather than exclusively categorical truths; but the above argument would still hold.

Stoljar argues differently. He wants to extend the categorical argument into a "categorical ignorance hypothesis" akin to, but more specific than, his own view. The key to this is an assertion that non-experiential truths could encompass both dispositional and categorical physical truths. Just because we don’t know of any examples of non-experiential categorical truths doesn’t mean there aren’t any. Note that Stoljar wouldn’t think it relevant even if our ignorance of this latter putative type of truth was chronic. Given this assumption, we have Interpretation 2:

1. Known non-experiential truths are dispositional
2. We are ignorant of categorical truths
3. Categorical truths may be both non-experiential and relevant to experience.

The resulting “Russellian version” of Stoljar’s epistemic view does dissolve CA and KA, and thus solves the logical problem of consciousness. (Note that Stoljar does not positively recommend this version of the Russellian view, rather he is advancing the more general version of the epistemic argument and is using the Russellian version to show an example of how the more general version could be fleshed out).


Stoljar’s thesis is difficult to defeat, because it is often easier to sow doubt about the adequacy of our knowledge than it is to defend positive assertions (like the statement above saying “all non-experiential truths are dispositional”).

In responding to Stoljar, I think the best strategy is to look more closely at how the paired terms non-experiential/experiential, dispositional/categorical, and objective/subjective are used and how they relate to each other. To do this, I will employ what I intend to be a minimal description of the world which must obtain to make sense of how these terms apply to it.

I start with the objective/subjective pair, because I think it is easiest to agree upon what these terms mean. In my last post I argued that objective and subjective truths are exhaustive of all truths. In essence, I’m saying that this is a dichotomy which describes our world. This doesn’t mean the world is described by a fundamental dualism like that of Descartes: both kinds of truth could supervene on an underlying common ground (neutral monism). On the other hand, a pluralism is implied, of course: The world contains individuated natural systems, and one type of truth obtains when a system participates in an event (subjective truth), and another type of truth obtains when such event is indirectly encountered (objective truth). There are as many “subjective” points-of-view as there are natural systems.

If you grant me this much, then the most natural interpretation of experiential truths is to identify them with subjective truths, while non-experiential truths are identified with objective truths. As discussed in the last post, such an interpretation defeats Stoljar’s thesis.

With regard to dispositional versus categorical truths, the terms are a bit more obscure, but the analysis leads to a similar result. A system’s indirect non-participatory knowledge of truths about other systems will only be of dispositional truths. We can only know them by how they interact with further entities. Another way to see this is to use the synonyms “relational” or “extrinsic” as alternative pointers to the same concept. The most straightforward interpretation would likewise identify non-dispositional (i.e. categorical) truths with those known through direct participation. The idea of a "non-experiential categorical truth" appealed to by Stoljar above has no room to fit into this picture and is not a well motivated concept.

Concluding Thoughts

Stoljar’s book was valuable to me not only for his interesting thesis, but for how useful it was as a touchstone for reflecting on the debates in (analytic) philosophy of mind which I’ve been reading about for around 15 years now. Within this philosophical domain, I have been most persuaded by the views I’ve discussed in these posts: a version of the Russellian stance, bolstered by arguments primarily associated with Chalmers and Nagel.

But, of course, consensus is far off. I haven’t discussed the parts of Stoljar’s book where he argues against competing physicalist perspectives on the problem of consciousness (arguments one agrees with always generate less scrutiny!). But many philosophers still subscribe to these arguments.

It seems that a significant move toward consensus will probably require input from outside philosophy of mind. Even in my “minimal description of the world” outlined above I was essentially “cheating” by bringing in ontological assertions about the existence of natural systems and how they relate to each other. But I think some sort of “triangulation” is necessary to make progress. Consciousness needs to fit into an improved metaphysical portrait. As an example, this is what appealed to me about Gregg Rosenberg’s thesis (see posts here), which combined insights from an analysis of causation and the composition of natural individuals to find the right “place for consciousness” in the world. It is also the reason I continue to be very interested in the philosophical interpretation of foundations of physics.

It seems right that a deeper understanding of consciousness will go hand in hand with a deeper understanding of nature.


Daniel Stoljar said...

Hi Steve,

First of all: thanks again for taking the time to write this series of posts! It’s great to have some feedback, particularly of such an insightful and thoughtful kind.

I had quite a few thoughts on what you said but let me restrict attention to the point about objectivity you made in your second post. You wrote:

“The difference is that in the original case the two types of truths (objective and subjective) are presumed by definition to be exhaustive of all truths. In the cases of Stoljar’s examples, the additional facts which led us to see that {if A then B} could be necessary even when not synthesizable invoked other types of facts beyond the types represented by A and B. In the original counterexample #8 and #9 I need to know about abstract and concrete objects and what kind of objects can have pain. These facts were the “new” facts beyond just knowing everything there was to know about the antecedent “numberish” facts. But in the original objection, there are no new types of facts beyond the objective facts that could help me erase the appearance of contingency, because objective and subjective facts exhaust all facts.”

This seems to be your main objection because you mention it again in your third post, saying that it shows that the epistemic view (i.e. the view I defend) is false.

But to be quite honest I don’t at all see the force of this criticism. In the original case we have N which summarizes all the non-experiential truths that obtain in the actual world, and E which summarizes all the experiential truths. Now understanding N does not entail understanding E, i.e., because you can understand N from more than one point of view, and yet to understand E requires one particular point of view. But it does not follow that N does not entail E, or that it appears that N does not entail E. You point out that N and E together exhaust all the truths there are. That’s correct, but I don’t see why it affects the underlying point. It is still true that the non-synthesizability of ‘If N then E’ does not entail any failure of necessity, nor does it entail an appearance of such a failure.

Maybe you are thinking that if a conditional of the form
‘If A then B’ is necessary but not synthesizable, this will only be so if A and B together do not summarize all the facts. If that is so, then from the fact that some conditionals are necessary but not synthesizable, we could not infer that ‘If N then E’ is—because here N and E together summarize all the facts. But why assume that a conditional is only necessary but not synthesizable if its constituent propositions summarize all the facts? I don’t see any reason to accept that. Here is an example. Imagine a world whose non-experiential truths concern nothing but extension. Now, it seems to be true that in a world of pure extension, nothing is in pain, just as a Cartesian Dualist might have thought, i.e., a world of extension would need to be enriched with something else if is going to support the existence of pain. Now consider a statement DN that summarizes all the non-experiential facts at such a world; and consider too the statement ‘nothing is in pain’, which we might call ‘DS’. The conditional ‘If DN then DS’ is necessary, but it is not synthesizable: understanding DN requires understanding what extension is, whereas understanding DS requires understanding what pain is. On the other hand, DN summarizes all the non-experiential facts of the world we are imagining. So here we have an example of a conditional that meets your condition and yet is necessary but non-synthesizable.

Does that help?


Steve said...

First, let me thank you very much for the response.

Yes, I probably jumped too fast in thinking there was something particularly important about cases where N and E summarize all facts merely from the fact that your examples appeared to differ from the original case in this respect. Your new example is helpful.

One issue which springs to mind in the new example is that we needed access to truths external to the (model) world to see the necessity of the conditional. But in our objectivity case we say: Understanding N does not entail understanding E, but ‘If N then E’ could nonetheless be necessary and could appear to be necessary (if one had all the facts). This “appear to be necessary” component is still troublesome to me. Assume that someone like Mary already has all the N facts (available from within the world). Can ‘If N then E’ be not synthesizable but appear necessary from a point of view within the world? Perhaps this doesn’t matter to the argument. I’ll think about this some more.

Thanks again,
- Steve

Gregg Rosenberg said...

Hi Steve, thanks for writing these entries. I rarely have time to read philosophy anymore, so I haven't read Stoljar's book. But I have some questions about how you've reported some of Stoljar's arguments. Let me know if I've misunderstood.

On his deduction of a subjective truth from objective truths: It seems non-parallel to the kind of inference he needs. If I understand correctly,

a) John is a number, therefore he is not in pain.

is true because,

b) Numbers do not feel pain

is true.

This is an objecive truth about subjectivity itself, as a category. I don't tghink anyone has ever claimed that there are no objective truths about subjectivity. (e.g., I take it that "Subjective states exist" is an objective truth about subjectivity, and not even the most hardened Nagel-fan would deny that statement).

So no harm deducing objective truths about subjectivity using such premises.

The crux of the Nagel argument is that truths about the characters of individual conscious states are subjective. Does Stoljar produce a counter-example to that?

Even that I would expect to be false in certain ways. For example, if there are laws connecting brain states to subjective states, one could deduce objective truths about subjective states using these laws. What would be the harm of that? I'm not sure what Stoljar is trying to show from what you wrote. I'm sorry to be obtuse.

I also find the argument from lots of other things being explained reductively, to the conclustion that consciousness probably will be too, to be based on a misuse of induction.

Induction is famously logically invalid. There is no formal basis for justifying it like there is for deduction.

To use it reliably it has to be underwritten by metaphysical assumptions, viz. that one is dealing with a universal or a natural kind, and therefore by finding the truths common to many instances one is likely finding the truth common to all.

But there is no reason to think -- and lots of reason not to think -- that consciousness and the reduced elements of nature like economics, life, etc are each instances of a universal or a natural kind, such that one can inductively infer from the reduction of some that they are all reducible.

So I think it is illogical to apply the rule of induction here, as there are no formal criteria that make the rule valid and the metaphysical assumptions required don't seem to hold. Does Stoljar say anything more about this other to make a naive appeal to induction?

Steve said...

Hi Gregg.
I take his argument as more subtle -- it's not a direct argument for why one should think objective truths entail subjective truths. His strategy is to defeat the concievability argument by an appeal to our ignorance. It seems as if we can conceive that the non-experiential truths fail to ential the experiential truths (to use the terminology he mostly employs), but this is only because we are ignorant of some type of "experience-relevant non-experiential truth".

Casting non-experiential and experiential truths in terms of truths derived from objective and subjective points of view (with Nagel in mind) puts stress on this "epistemic view" of the problem, since even Stoljar concedes that a complete "understanding" of the objective truths would not lead of necessity to an "understanding" of the subjective truths.

And yet (and this is the difficult part), he argues that this lack of "understanding" entailment does not mean that the actual entailment of subjective truths by objective truths must fail (if only we could learn some additional relevant objective truths). The example above is specifically geared to convince one that this kind of thing could be the case. (If I only understand 'number-ish' truths, I would not understand 'pain' truths, but nonetheless being a number does entail not being in pain)

I still can't agree at this point, because I argue that in the actual case we are considering there is no room in the real world for any further relevant objective truths; I conclude that if these additional putative truths exist, they must only be available from a God's eye view beyond our world (and it doesn't seem appropriate to implicitly be appealing to these given the terms of the debate).

The arguments here get pretty involved and it may be I'm not able to really do it justice here. The book is interesting because Stoljar is in agreement that most arguments for physicalism fail (a posteriori necessity, phenomenal concepts, etc.), but this argument from ignorance, he thinks, could still defeat the conceivability argument and lead one to conclude that physicalism is true.

Steve said...

Gregg - Let me add something which addresses the spirit of your comment a bit more. The thrust of the book is about using the ignorance hypothesis to defeat the conceivability argument, and that's what I've focused on here. But, there is also a presumption that the conceivability argument's success is the only thing standing in the way of concluding physicalism is true.

In an early chapter (ch.2), Stoljar spends some time directly defending the thesis that experiential truths are entailed by non-experiential truths. Here the preferred route is an argument from 'manifest supervenience' (from 'manifest vs. scientific images' as discussed by Sellars). Science explains pre-scientific manifest or common-sense truths. It is therefore plausible to suppose all manifest truths, including experiential truths in themselves, supervene on scientific (non-manifest)truths. Non-manifest truths seem clearly to be non-experiential truths. So the plausibility of this manifest supervenience hypothesis supports the entailment of experiential truths (E) by non-experiential truths (NE).

Clearly this is not a deductive argument. We are in the realm of deciding the relative plausibility of competing theses. Stoljar is saying (if my reading is right), that defeating the conceivability argument removes the key support for the plausibility of the {NE does not entail E} thesis, leaving the other, plausible, {NE entails E} thesis standing.