Search This Blog

Monday 16 December 2013

A Discussion with Alexander Gillett About The Metaphysics of Information, Physicalism About Information, and Non Eliminative Ontic Structural Realism

Ontic Cafe 6 beans Strength: Academic Specialists and Masochistic Layphilosophers only.


My PhD thesis presents a case for physicalism about information: that information is and reduces to the physical and that there is no information without physical structures. An initial response to this position is often that it is simply a restatement of token physicalism. A lengthy debunking of this response is not within the scope of this post. However, and easy way to demonstrate it is not sound is to point out that numerous philosophers of information and information theorists have proposed that non-spatiotemporal non-causal structures are a sufficient condition for the existence of information. This means that there are non-phuysicalist conceptions of the nature of information itself. There are other such conceptions.
Another philosophical expression of physicalism about information, then, is that it is an anti-Platonist conception of information. This is at best a partial statement, but it serves to highlight the relevance of a physicalist position. Mine is also a non-subjectivist, non-pluralist (nominally) and non-statisticalist conception (the later being controversial.) I will not attempt to explain these positions extensively here. According to subjectivism about information, information is only realised in the context of being received and (usually) perceived by some kind of receiving agent (the agent need not be a cognitive agent - but can be simply an organismic consumer as is the case with the teleosemantic theories of Ruth Millikan and Nicholas Shea.)
The dissertation (Physicalism About Information with Applications) draws on resources from applied mathematical theories of information - algorithmic, quantitative and computational. I use premises and arguments put by quantum computing and information theorists like Rolf Landauer, David DiVincenzo and Daniel Loss. For the metaphysical conception, the argument relies to a significant extent upon the kind of non-eliminative ontic structural realism put forward by James Ladyman and Steven French. However, I do not commit myself to any particular OSR perspective. Teh leading philosopehr of information in the world today - Luciano Floridi - has demonstrated that part of the project of philosophers of information should be to investigate the nature of information. This is, ironically and perhaps unexpectedly at this time in the history of philosophy, a new metaphysical project. Those materialists familiar with the work of Rudolph Carnap and Willard van Ormand Quine will realise the relevance of this.

Recently I gave a presentation at the University of Melbourne Australian Postgraduate Philosophy conference on the subject of explanation in mathematics. The argument I presented, stated briefly, was that if mathematical abstracta are explanatory because they are informational, then there is a strong case to be made that, against the preponderance of philosophical intuition and thought on the matter, abstracta must be physical. This is not the argument that I am interested in making here, however.

To get to the point of this post, an attendee at the conference, Alexander Gillett, was interested in my use of ontic structural realism. Alex is a specialist in OSR, and had some interesting questions and perspectives to offer on the use of NOSR as a partial metaphysical basis for a physicalist conception and ontology of information. Alex kindly agreed to the publishing of relevant parts of our recent email communications on the matter. I offer them here for the interest of the Ontic Cafe audience.

First Exchange

Dear Bruce, Last month I attended the post graduate conference in Melbourne and I found your paper on OSR, information theory and a critique of Platonism (a.k.a. Mark Colyvan) fascinating. I have been researching structural realism for a while now, and having come from a different direction it was interesting to hear you discuss the information-theoretic versions of SR. I have only recently begun examining these approaches and trying to get to grips with them. I have read Floridi's "defence of ISR" - do you recommend any other papers by him or others?

After your talk I was asking you about an adequate definition of nature or the physical, and about Ladyman & Ross' work on real patterns, but unfortunately the time for questions ran out when you were about halfway through an answer on defining the physical. I was wondering whether you would be willing to discuss these issues now. Additionally, you made reference to a paper called "An Informational Physicalist Ontic Structural Realist Conception of Mathematical Abstracta" which sounds really interesting. Would you mind sending me a copy of this paper or any others that I might read?

You also talked about physical information sources or information-bearing structures as the basis of any possible explanatory account. In this regard, would I be right to say that information comes before meaning? Can I also ask, how do you consider information metaphysically? . . . Anyway, thanks for the talk, it was probably the most interesting thing I heard over the course of the conference and I'm sorry I didn't manage to catch up with you in person. Regards, Alex Gillett
Hi Alex,

I am pleased to hear from you Alexander. . . .

Yes I put information before meaning in a sense, but I take information to exist physically and as such to be intrinsically semantic on a particular causal basis.

I have a paper in review called “Information is Semantic but Has no Alethic Value”. [I will upload the pre-pulbishing version of this paper to Ontic Cafe in the next instalment]

I have a couple of papers that I can send you. I have to submit one of them to a journal this week first (Physicalist Ontic Structural Realism About Information in Applied Mathematical Theories of Information). The other paper I will send you tomorrow.

Best Regards,

Bruce.
Hi Bruce, . . . I went to a conference the other week and it left me even more convinced that I need to engage with information theory more. Just a quick question: since I've come at information theory from a structural realist POV, what is it about Floridi's approach that you disagree with? Do you think you're approach is more amenable to some metaphysical version of SR? Regards, Alex
So, brief bio out the way, I'd like to say thanks for the paper, and if you are willing to send anything else I'd be delighted to read it.

Just a few comments vis-a-vis structural realism (SR). Both Ladyman & Ross (2007) and French & Ladyman (2003) - your paper seems to mix references here perhaps? - are more readily seen as promoting an eliminative metaphysical version of SR. They oppose the notion that structure is fundamentally causal, arguing that since fundamental physics is currently agnostic about causality, metaphysics should not posit it as a fundamental feature of reality (see Chapter 5 of Every Thing Must Go, 2007). Additionally, they propose an eliminative account of objects. Some have pointed out that this is inconsistent and they have subsequently relented and allowed in thin-objects - which are purely relational. Michael Esfeld & Vincent Lam have argued for this position more consistently (along with Fred Muller & Simon Saunders) - what they call Moderate Ontic Structural Realism (MSR) - on the basis of empirical evidence from fundamental physics and philosophical arguments surrounding the methodological issues relating to how to cash out objects in an ontology. Additionally, they also argue that reality is fundamentally causal or dynamic. As such, perhaps their approach is more conducive to your own?

In addition to these issues, Ladyman & Ross (2007) propose an ontological reading of Daniel Dennett's 'Real Patterns' as the basis of their ontology, and they couch this in information-theoretic language (see chapter 4). Their reason for doing so is to avoid explicit reference to causal process and to replace this with 'information-bearing'. What interested me about your work, was given your background, you could probably appraise this approach far better than I. I was wondering whether you had had a chance to engage with this material? The reason I ask is that you are the first person beyond Floridi to be explicitly discussing SR and information theory - and you disagree with the former. This is what interested me greatly.

Anyway, thank you for sending the paper and I hope to hear from you soon;
Hi Alex,

Thanks for your response.

I think it is the 2003 paper (it is possible that I have mixed references) where when challenged by Cao to describe the kind of structure that they were talking about.

I have read the material that you are referring to. (Not all of it recently.)

My position is hyper-physicalist. People find it annoying and think I am making a mistake. But I have not heard a good reason to drop it yet. It is difficult to argue for because it requires informational structure to be real structure, and this involves the structural realist mathematician to be a physical structural realist. It also requires that there is a distinction between statistical measures of various kinds and what is being measured. In other words there is only one kind of real structure that is a necessary condition for the existence of information. It is non-eliminative physicalist causal structural realism. Yes – the dumb ass bumping together kind. This is highly unintuitive – but then so is the statistical conception of the measure of information (a MEASURE.)

I cannot do justice to it in one email, but I will send you some papers soon.

Interestingly, they start that paper (‘THE DISSOLUTION OF OBJECTS: BETWEEN PLATONISM AND PHENOMENALISM’) with:


One of the motivations for Ladyman’s ‘ontic’ form of SR is that it offers the realist some hope that she may be able to get away with carrying less metaphysical baggage than the ‘standard’ realist without having to fall into the clutches of the constructive empiricist (2003, 73.)


And in it go on to say that:

Cao persists in lumbering us with two seemingly contradictory identifications that we thought we had rejected in our paper. The first concerns the identification between physical structures and mathematical ones, which Cao then takes to imply that the ontic structural realist must be a Platonist.


I see why Cao does this. Because it looks like the only way that L & F have to go. As I will mention below, I think that Cao’s misinterpretation points to the right view for the metaphysics of information.

Then:

Now, we did say that the distinction between the mathematical and the physical may become blurred, particularly if the mark of the latter has to do with ‘sub-stance’ or individual objects or the like. Nevertheless, blurring does not imply identity. The mathematical can be trivially distinguished from the physical in that there is more of it; there is more mathematics than we know what to (physically) do with, which is what Redhead expressed with his notion of ‘surplus structure’. What makes a structure ‘physical’? Well, crudely, that it can be related – via partial isomorphisms in our framework – to the (physical) ‘phenomena’. This is how ’physical content’ enters. Less trivially, the mathematical can be distinguished from the physical in that the latter is also causal, (2003, 75)


Something very interesting, and controversial, happens with my physicalist conception of informational and real structures, and the concept of mathematical abstracta that goes with it. The distinction between the structural realism of the mathematician and that of the metaphysician in the philosophy of science IS eliminated. Only the latter physicalist conception is retained for structures that can be considered informational or information bearing.

There are lots of ways to approach this unpopular position. First however, here is what Ladyman and French then said to Cao:

Cao understands us as advocating ‘the dissolution of physical entities into mathematical structures’. But, first of all, by ‘dissolution’ we mean metaphysical reconceptualisation. And secondly, as we tried to em-phasise, to describe something using mathematics does not imply that it itself is mathematical – the structures are what they are and we describe them in mathematico-physical terms. Let us put it as clearly as we can: we are not mathematical Platonists with regard to structures (2003, 75.)


So mathematical structure does reduce to physical structure for Ladyman and French. It must. I am saying that this also means that there is no information without physical structure, and that in fact all information must reduce to physical structure.

Shannon’s work creates big problems here, but only because it gets misinterpreted. Shannon does say that the model of a source (stochastic process) is also an information source, and he uses a statistical ‘measure’. But it is the measure that is statistical – not the information. The Boltzmann conception of entropy makes this worse not better. Even though entropy is a probabilistic measure in Boltzmann’s theory, the entropy or disorder is in fact physical. There is no real entropy without the physical configuration of the physical particle system. Saying that information is statistical is nonsensical according to Shannon’s account, especially if information is entropy. The measure of entropy is a measure of physical entropy, and even if you do not apply it to a real physical system, you don’t have real entropy without the physical system. Likewise with information.

What about relations between physical point, entities, and structures? What about mathematical patterns that are not physical structures? Aren’t they informational? Kolmogorov did not think they were even real. Neither do I. I therefore think that they are not informational (neither did Kolmogorov – his data objects were all strings of physical symbols of one kind or another. Tables of random numbers were tables of physical symbols for Kolmogorov.)

What about the fact that mathematical patterns seem to carry information in the sense that they can tell us things about a system even if the system is not realized? This seems to suggest (per Colyvan and Lyon) that there must be real Platonist informational structures. Ladyman and French do not want to be Platonist, but they misinterpret Shannon on the nature of information. I have suggested a way in which possibility spaces and probability spaces can in fact be physical structures (controversial but it is a solid argument) – but this is not what Ladyman and French do.

The best way to bring this out is with an error made by Dretske (and I think that it simply is an error.) Dretske demonstrates (1981, 26-31, 38) that there is a real distinction between causality and information. However, the conclusion that does not follow is that there can be information transmission and generation without causality. This is demonstrably false (I provide an argument in my thesis.) it also follows from this that if Ladyman and French pursue a statistical conception of information like Dennett, they have not eliminated causality at all.

No causal physical structure – no real information.

Kind Regards, Bruce.
Hi Bruce,

Thanks for the emails. I've got to stay in all day waiting for a delivery, so the article on information theory and biology shall be both a great time killer and brain-food.

I really enjoyed the responses to my questions - very much appreciate the frankness. In particular, what interests me is what I shall refer to as the "Blurring problem". I have recently had a paper which gives an overview of OSR and this issue of dissolution of the distinction between the mathematical and the physical accepted to the Polish Journal of Philosophy. In this I explore the various (inadequate) responses to this very problem. Now, although in the paper you cite, Ladyman agrees with French that causation can be used to "trivially" articulate this blurred boundary; mostly Ladyman has rejected the notion of causality as a distinguishing factor in solving this problem (see Every Thing Must Go and a recent interview on rationallyspeaking.org maintains the agnosticism).

Furthermore, I think their attempted response to Cao just begs the question via appealing to a tautology: the physical = the physical.

Additionally, Ladyman's relationship with platonism is also confusing - in some places he rejects it out right, and in others he advocates a naturalised version. And in the interview above he seems affable to pythagoreanism or platonism (a la Max Tegmark's extreme OSR mathematical universe hypothesis) although not committed to either.

For all of these reasons it seems to me that your position is more amenable to Esfeld and Lam's perhaps. For them physical structure is causal and thus distinguished from the mathematical in this sense, and they reject platonism in this sense. (I have attached papers by Esfeld & Lam for your perusal if you are interested)

I think such a view is strengthened by your comments regarding the difference between a measure and the thing that is measured. This seems to be similar issue to the blurring problem, where I think an account of scientific representation is missing. Ladyman & Ross's recent paper, "The World in Data" (2013), combines these two issues. In addition to stating that physical structure is modal, Ladyman & Ross conclude this paper by stating, following CS Peirce, (and presumably extrapolating from their statistical RP approach at representation) that...


The fundamental empirical structure of the world is not mathematical but statistical. And there is no such thing as purely formal statistics. The 'principles' of statistics are simply whatever dynamics emerge from our collective exploration of, and discovery of patterns in, data. [...And this is premised on...] the grandest discovery of the twentieth century, that fundamental physics, and therefore reality itself, are irreducibly statistical. What is the world? It is the endless weave of patterns to be extracted from noise, at an endless proliferation of mutually constraining scales, that we will go on uncovering for as long as we have the collective and institutional courage that comes from love of objective knowledge, the great moral core of the Enlightenment.


On your account I take it this is rather like mistaking the finger pointing at the moon for the moon itself? However, it could also be interpreted in a more reasonable fashion when read in the light of there comments at the close of chapter 3 ETMG (pg189). In trying to unify fundamental physics with the special sciences they state that there are two options: either [a] reality is composed of infostuff; or [b] the world is not made of anything but that information is the main concept for understanding the objective modal structure of reality. They choose view two, in which case they see some form of information-theoretic notation as the basal and collective account of scientific representation. This would again affirm their refusal that dissolution = identity, but point to an obscure fact that scientific representation is so accurate as to entail partial isomorphism - an astonishing fact colloquially referred to as the Wigner Puzzle by Colyvan. Textual support for this interpretation comes from their quoting of Zeilenger: "it is impossible to distinguish operationally in any way [between] reality and information" (ibid). So although they may be wrong about the statistical aspect - is there a consilience with your view here, or how does it differ? Do you maintain a more robust distinction between informational representation and physical structure as source? I'm guessing this is your view given your last comment, but how do you cash out this distinction in a philosophically robust manner?

I'm not in the best place to judge these views, and I am more interested in exploring the terrain of SR and its connectivity to information theory and an account of scientific representation than throwing my hat in with any side in this debate. As such, I'm throwing out these comments and questions as a devil's advocate and interested to see how you respond.

What I'm most intrigued by is your proposal that possibility and probability spaces are physical. Does this amend the indispensabilitist arguments for mathematical realism in favour of a form of nominalism? How does it relate to other issues with the applicability of mathematics? Could you perhaps spell out the argument in a little more detail?

Looking forward to your responses, Regards,

Alex

No comments:

Post a Comment

POLLS:

Does information exist apart from perceiving agents (without some thinking organism to perceive it)?

Is Philosophy Relevant to Science?