1. Some preliminary suggestions
I propose that the scope of our discussion should be defined by the question
Q. Is physics a universal science?
This question is the title of Bialkowski's text (the chapter of his book) which along with Weinberg's "Reductionism Redux" will serve us as the starting point for our discussion,
This above is not to say that I propose that we should restrict our discussion to the debate between the YES and NO parties: the proponents of both Bialkowski's and Weinberg's view that whatever regularities can be observed in the world they must reduce one way or another to physical laws and those who -- like myself -- deny such reducibility.
I am sure that nobody is going to deny the foundational role of physics for scientific inquiry and the fact that various allegedly non-physical regularities can be adequately accounted for in physical terms. Consequently, one who do not believe that the laws of physics suffice to explain all the regularities may still strongly support the view that the idea of reducibility of various scientific disciplines (or parts of them) to physics is of considerable methodological significance and deserves to be carefully studied. The members of "no" party might surely be eager constructively participate in solving various questions to which the YES position gives raise.
Let me mention two such questions and briefly comment on them.
Q1. What does it mean for a certain law to be a law of physics?
------
SOME COMMENTS. Bialkowski undertakes this question when he discusses why from the very beginning thermodynamics deserved to be treated as a branch of physics. In his opinion the decisive reason was methodological unity of thermodynamics and other domains physics. To be sure, Bialkowski's position on this matter can be challenged. I do not believe that "methodological reducibility" (to use Bialkowski's term) of some area of investigations to physics suffices to treat that area of investigations as a part of physics.
-------
Q2. What does it mean for a certain domain of investigations (the set of laws discovered in that domain) to be reducible to another?
-------
SOME COMMENTS. In Bialkowski opinion the idea of reducibility is not homogeneous. On different occasions people using the term "reducibility" mean different things. According to him reducibility splits into the following four different notions:
(a) atomistic reducibility,
(b) logical reducibility,
(c) semantic reducibility,
(d) methodological reducibility.
The strongest is (a) the weakest is (d). I am not sure whether all those four notions were defined by Bialkowski in sufficiently precise way. On the other hand, I believe that except perhaps (d) -- methodological reducibility -- the ideas of atomistic, logical and semantic reducibility's are clear enough in order to allow us to use them on various occasions in a precise enough way. (The fact that a term is vague does not mean that using it must always result in ambiguities).
It is not clear for me how Bialkowski's and Weinberg's position are related to one another. In particular it is not clear for me whether Weinberg's idea of grand reductionism should be explicated in terms of atomistic or merely logical reducibility.
Another thing, which in my opinion is of considerable interest is the relation between the idea of reductionism and Fritz Rohrlich idea of "ontological levels". But surely these problems can be better presented by Fritz himself.
--------
2. Why I do not believe that physics is a universal science
In Bialkowski's opinion the main differences between natural and social sciences (let us treat humanities as part of the latter) are the following two. Firstly, any social investigations require taking into account a very large number of parameters in order to adequately account for the studied phenomenon. This is not the case of natural sciences in which one might be able to account for a very large class of phenomena using a very small number of variables. Secondly, people who are interested in social phenomena are very often interested in some individual events as such (cf. Bialkowski's example of special interest in Adam Mickiewicz the great Polish poet) not as representatives of some universal category.
To begin with let me note that these are not social science and humanities alone which are interested in singular phenomena. The ancient astronomers who studied the planetary system studied a singular system being a very peculiar episode of the history of the universe. For the same reason the large part of our knowledge about the Earthly flora and fauna may concern something extremely local and singular.
Also the difference that allegedly consists in the number of variables one has to control in order to arrive at some both orderly enough and adequate description of a phenomenon seem to be of secondary significance. There certainly are some fully natural phenomena which escape any adequate description just because the number of parameters on which they depend is enormous.
The substantial difference between natural and social phenomena consists neither in the alleged "singularity" not in "multiparameticity" of the latter. Social phenomena differ essentially from natural phenomena because there is no reasonable way (at least we do not know any) to study the latter without taking into account the fact that social events depend on FREE decisions of people. They depend on them often in an essential manner. In other words those phenomena are not merely CAUSAL but also INTENTIONAL.
I am afraid the idea of intentionality is indispensable in any realistic account for social phenomena. On the other hand, one may wonder if it is of any use when one deals with natural phenomena.
There are various metaphysical problems concerning the idea of intentionality but I believe we should avoid undertaking them. I propose to restrict the discussion to the problems which in the most simple way are defined by following simple example.
Consider the behaviour of a good chess player. The better is the player the more predictable are his successive moves. Yet, as any human behaviour, the chess player behaviour certainly is not deterministic. It is not probabilistic either at least in the sense that it cannot be adequately described in terms of objective probability in any systematic way. On the other hand since given any chess position we are able to narrow the class of moves which say Kasparov or Fisher would consider reasonable we are able to predict (though not in a unique way) the behaviour of those players. The question is how this is possible.
Well, the predictions can involve analyses of some past games of the player in question. But even if we ignore any individual characteristics of the players the very fact that they are chess masters allows us to considerably narrow the spectrum of their possible behaviours.
The predictability in question is based on the MATHEMATICAL fact that chess is a decidable game, so all the strategies available at a given "position" can be divided into winning and loosing. Roughly speaking, the better a player is the less likely is that she or he will use a loosing strategy.
The example I have produced illustrates, I believe, the following two things:
(1) There is no way to describe the behaviour of chess player in terms of laws of physics. Still,
(2) The behaviour of a master player can be fairly well accounted for in a fairly rigorous way. It can be described in terms of "tendency" of using "winning strategy". (Incidentally if one is sceptical about the possibility of defining winning strategies for chess games in an effective way, one may replace the above discussed example by a game for which those strategies can be defined quite easily).
Let us agree (as it has been pointed out by Weinberg) that there are some events which escape any scientific explanations. Not because they are extranatural, but because nobody is able to collect all the evidence necessary in order to adequately account for them.
If I have grasped Weinberg's idea of grand reductionism correctly, we might not be able to find out the physical causes of an event but nevertheless we might be sure that no event whatsoever violates physical laws. Let me note however that this view leaves open the question of whether physical laws are the only one needs to adequately account for phenomena or perhaps there might exist other laws one should take into account when one seeks to understand things going around us. I do suspect that those additional non-physical laws which we should not ignore are the laws of rational behaviour. These are the laws of the behaviour of a rational chess player which allow us to predict the chess player behaviour.
3. More about the laws of rational behaviour
One may wonder whether the rules of rational behaviour deserve to be called laws. One may also wonder whether the idea of such rules/laws is clear enough. Any more accurate analysis of the example I have discussed above should reveal that the idea of rational behaviour might not be as easily definable as one may expect. The rationality of the behaviour does not depend on the rules of the game only but it depends on the skill of the chess player, it experience, the time he has to his disposal etc. Yet, in the most rough sense these are the rules of the game which determine the chess player behaviour and allow one to classify his behaviour as rational or not.
There are various situation in our both individual and social life in which both "the rules of game" and the goals we want to achieve are fairly well defined. For instance various economical situations are of this kind. In all those cases there is a good chance that the idea of rational behaviour can be fairly rigorously defined. And in all those cases there is a chance that we might be able to define a fairly good mathematical model of human behaviour. Of course any such a model is a model of the behaviour of "rational" subjects. But even though the idea of rationality is not in general definable in terms of "rules of game" (not to mention that people do not always behave rationally), quite often human behaviour can be fairy good explained in terms of tendency towards selecting the winning strategy.
>From the formal point of view the rules of rational behaviour are some mathematical theorems about some "goal-oriented" mathematical systems, i.e. "games" of various kinds. In spite of what one may believe the formal nature of those rules does not necessarily differ intentional laws (as I am going to call rules/laws of rational behaviour) from the causal laws (as I am going to call regularities reducible to the laws of physics). Exactly as the former also the latter have to be applied to concrete situations in order to gain the status of laws that describe something specific -- a piece of (either physical or social) reality. Unless any their "factual interpretation" has been offered both casual and intentional laws are just theorems on some abstract mathematical system: Galileo transformations of coordinate systems, Hilbert spaces, abstract market, abstract chess game, etc.
I am aware that I have merely outlined in the most preliminary and crude way the point I wanted to make. But the basic idea I advocate is, I believe clear. The intentional laws are as indispensable element of our knowledge of the world as the causal ones. There is no way "to reduce" science to causal laws of any kind whatsoever.
One of the open problems of methodology of science is whether the laws of real phenomena can be adequately divided into causal and intentional. Perhaps all of them, just as Leibnitzian monads, are both causal and intentional though perhaps at different degree. But this question leads us to the shaky area of metaphysical speculations.
a piece of Dennis Dieks's letter to Prof.Wojcicki from Thu, 19 Jun 1997)
[DD Jun 19]:
>From Prof. Rohrlich's letter on 23 Sep 1997:
(1) The two problems addressed by Bialkowski (is physics a universal science?) and by Weinberg (what is reduction?) are related. I believe that one must address the second one before being able to give a meaningful answer to the first one.
Reduction, in turn, requires a scientific theory, S, to be reduced to a deeper, more fundamental theory, T, so that in some sense T implies S. Thus, one must have, to begin with, two levels of theory, a deep one, the level of T, and a less deep one, the level of S. That, I believe is where one must start.
(2) Levels: The different fields of scientific research fall NATURALLY into levels which, to be sure, are not very precisely defined, at least at the beginning. This is true even within a given discipline: biology separates into many levels from zoology to cell biology, physics from cosmology to subatomic, etc. A good general criterion is size: the scale on which research is done. This can be elaborated further.
(3) The old and long debate on reduction would be greatly simplified, I believe, if one were to distinguish two concepts that are usually conflated:
(a) Theory reduction in the most rigorous sense possible: the deduction of one theory, S, from another, T. This necessarily requires bridge laws since the two theories contain terms, concepts, etc. that are incommensurable.
(b) Explanation of a scientific fact, rule, law, etc. in theory S by means of the deeper theory T. Example: why heat flows from hotter to cooler locations in a metal (a fact of thermodynamics) is explained by statistical mechanics.
I believe that making a clear distinction between theory reduction in the strict sense (a), and explanation, not of the whole theory but only of a specific fact or law, (b), would help greatly in settling disputes on reduction.
(4) Just to look ahead: physics is universal, I believe, in being able to explain in last analysis. But theory REDUCTION does not always carry through because the bridge laws can become a serious problem.
This, in broad outline, is my thinking. I am anxious to get your responses.
>From Prof.Wojcicki's letter on Fri, 03 Oct [RW Oct 03]:
In his paper published in BJPS 39 (1988) "Pluralistic Ontology and Theory of Reduction ..." Fritz rightly point out that any discussion concerning reduction requires a clear idea of a theory. I do agree. Moreover I think a very good starting point for clarifying that idea is just Fritz' paper.
Fritz distinguishes the following three concepts: empirical laws (inductively generalised observations), conjectures (conceptual schemes with poor empirical support) and theories. Theories are unsharply divided into accepted, mature and established. An established theory is a theory which cannot be discarded -- a mature theory becomes established at the time when we are able to define in the right way its "validity domain" - in other words we are able to tell what is the right area of applicability of such a theory.
If I interpret Fritz paper correctly I am afraid we never can be sure whether a mature theory is established or not. An example of established theory is Newton's mechanics, NM. Supposedly, that theory became well established when Einstein discovered relativistic effects, and hence we learn the right limits of applicability of NM. Note however that the discovery of QM also affected our views on what are those phenomena one is able to account for in a sufficiently accurate way by resorting to Newton's Laws. We can easily imagine that our knowledge on what are the phenomena to which NM can be adequately applied have not reach its ultimate stage, and besides quantum phenomena and relativistic phenomena there are others (at present unknown for us) to which NM cannot be applied.
Is there any theory established in the final way or is there no such theory seem to be of secondary significance for our debate. What matters is that the problem of reduction can reasonably be discussed with respect to mature theories. It matters also that reduction cannot be adequately viewed as the problem of interrelations between mathematical formalisms (mathematical-logical structure) of the theories in question. So I do agree with Fritz' when he writes (p. 104) "The characteristic components of [mature theory] which are important for theory of reduction are its mathematical-logical structure, M; its domain of validity D; its language and conceptual content, its epistemic component E, and its ontic component O". Also I agree that both formal and informal languages of the theories are both relevant to the idea of reducibility.
Can the above be summarised as follows? Suppose two mature theories T and T' are given and suppose that by enlarging T by a set of appropriately selected definitions we are able to imbed T' into T then the fact of existence of such purely formal reducibility of T' to T does not proves that T' is reducible to T in ``factual" sense of the word. In order to get factual reducibility we must have something more that formal reducibility.
Before I continue, I would like to know whether Fritz (and perhaps other participants of our panel) accept the way in which I tried to present Fritz' views from his BJPS paper. Also I would like to know whether Fritz' is ready to accept my conjecture that the idea of an established theory is in fact redundant for our discussion. The reducibility problem can be restricted to mature theories.
>From Prof.Rohrlich's letter on 6 Oct 1997 [FR Oct 6]:
In response to Ryszard's 3 Oct 97
First, I must explain my late reply: I spent the weekend in Pittsburgh at the Pitt-Konstanz meeting on the limits of science.
(1) Yes, I agree that the maturity of a theory is enough for both theories, T and T', where T' is to be reduced to T, BUT the validity limits of T' relative to T must be known. One can never be certain whether all validity limits of a given theory are known, but it is reasonable to call a theory 'established' when all validity limits to all known lower (more fundamental) theories are known, that set being non- empty.
(2) The term 'embedding' must be explicated. Mathematical reduction (a rigorous procedure) is by itself NOT sufficient. The main problem is the relationship of central terms of T and T', C(T) and C(T'). These terms are used in the symbols of the equations. Some of these can be (and usually are) INCOMMENSURABLE. They must be related by AD HOC assumptions often called 'bridge laws'. Just because these are ad hoc, a complete reduction is logically incomplete. That is my key point: reduction in any reasonable sense fails.
(3) Example to (2): 'temperature' is a central term in thermodynamics (THD) that can be made quantitative in THD. In order to relate it to the terms of mechanics (molecular velocity, etc.) of statistical mechanics (SM) one needs a bridge law. That bridge law differs in different branches of SM: ideal gases, condensed matter, etc.
>From Prof.Zytkow's letter on Wed, 8 Oct [JZ Oct 8]:
Let me make several comments:
1. One of my readings has been very disappointing: the Weinberg paper. Instead of arguments and evidence I found plenty of unsupported declarations of beliefs and other emotional states. Name calling replaces analytical thinking. The paper is no match to philosophical work of master physicists-philosophers such as Plank.
In his article reductionism becomes a piece of scientific creed: "But phenomena like mind and life do emerge. The rules they obey are not independent truths, but follow from scientific principles at a deeper level; ... the nervous systems .. have evolved to what they are ENTIRELY because of the principles of macroscopic physics and chemistry, which in turn are what they are ENTIRELY because of principles of standard model of elementary particles" [my emphasis]. Bialkowski's paper is so much more true to the practice of theoretical physics and to what we know and what we do not.
2. My position is very close to a number of Ryszard's statements. The following I share fully: "Social phenomena differ essentially from natural phenomena because there is no reasonable way (at least we do not know any) to study the latter without taking into account the fact that social events depend on FREE decisions of people. They depend on them often in an essential manner. In other words those phenomena are not merely CAUSAL but also INTENTIONAL."
The key concepts are "free decisions" and "intentional." Making those terms sufficiently precise and making a convincing argument that free will and intentional events exist is a perrenial philosophical problem. The problem can be viewed as the key to the discussion of physical reductionism. I doubt whether we can avoid free decisions and intentionality. If the discussion is going to be published, let us try to make those terms clear and the existential arguments thoroughly examined.
3. Fritz is right that reduction cannot be done by strict logical inference. Since both theories T and T' are strictly speaking inconsistent, we cannot use logical inference in a meaningful way. But what about approximate inference. I used the following in the previous round of the discussion, more than a year ago:
--------------------------------------------
This is an attempt to define the notions of APPROXIMATE consequence (implication), inference, and truth, in analogy to STRICT logical (mathematical) notions.
1) What does it mean to say that T2 is approximately implied by T1 plus structural assumptions?
First, about "logically implied": T2 is implied by T1 plus structural assumptions S: Each statement in T2 can be inferred from (T1 + S) but not from T1 or S alone.
In strict analogy we can say: T2 is approximately implied by T1 plus structural assumptions S: Each statement in T2 can be approximately inferred from (T1 + S) (but no statement in S belongs to T1 and to T2).
Of course, this calls for a definition of approximate inference.
2) What does one mean by approximate inference and/or approximate truth?
First about logical inference. Any schema of logical inference is truth-preserving:
When the premises are true the conclusion is also true
In strict analogy, we can require that approximate inference is approximate-truth-preserving:
When the premises are approximately true the conclusion is also approximately true
or
When the premises are true within the range of empirical error, the conclusion is also true within the same range
Now we need to define approximate truth or truth within the range of empirical error.
First the strict (logical, Tarski's) definition of truth, as a relation between sentences and a domain in mathematics:
Sentence P is true in domain D if and only if (iff) the relationship described by P holds in D
The domains (worlds) in mathematics are different from the domains in the natural world. Mathematical domains are built on well-defined primitive entities, such as empty set or natural numbers. Each object is defined by a precise operation on the primitive entities.
In contrast, the primitive entities in empirical domains are not a matter of definition but are subject to empirical investigation. Often it is not clear what they are. It is not the matter of a definition, but it belongs to empirical investigation, how are complex objects built from more primitive. In practice, each time we apply a theory to an empirical situation ES and each time we make empirical measurements on objects in ES, we approximate ES by objects which are on one hand defined in the theory and on the other hand determined operationally.
For such an object x we can compare theoretical statements in the category of facts (A(x)=a , that is the value of attribute A for object x is a), with the results of empirical measurements:
If the measured value of A for object x is b then A(x)=a is true within tolerance (error) E iff |b-a| < E
-----------------------
Structural assumptions used in reduction are expressed as mathematically unconventional relationships, such as v<<c, or m~=0 (m is approximately equal 0).
The intended reading of those conditions and their use in approximate inference schemes goes beyond the provided definitions and requires further elaboration
4. Structural assumptions S are different from laws. S are violated in many situations while satisfied in others. They are necessary for reduction, yet are overlooked in many accounts, for instance in Weinberg's. To me they are equally important to laws. Entries in both the categories laws and structural conditions are necessary and I do not understand why would laws be more necessary.
Structural conditions can be extended to robots, which can be helpful in our discussion as we can explore both the similarities and differences between robots and humans.
Robot is driven by a particular structure, which is its program. It is also driven by the hardware structure of the processor, memory, etc. In addition, I have no reasons to doubt that the laws of physics are satisfied. But the laws of physics are insufficient to explain the functioning of a robot.
5. Not only theories but also structural assumptions S can be explained by theories T' and structural assumptions S' on a lower level. Here the explanation can focus either on the possibility or on ubiquity of conditions S. Are situations described by S possible? How common are they and why?. We are interested in practically existing structures that occur by reasons which can be further explained. They are explained by laws and structures of lower level. Bialkowski makes a number of interesting arguments on the need for structural assumptions in his interesting analysis of deuteron and tritium/helium 3.
>From Prof.Rohrlich's Letter on Thu, 9 Oct 1997 [FR Oct 9]:
This is in response to Jan's remarks below. I fully agree that reduction is only approximate because the higher level theory T' has an error characterized by its validity limit (for example, in Newtonian mechanics: (v/c) squared.) But, in addition to that, there is the problem of terms in T' which don't exist in T (for example, if T' is thermodynamics, temperature is not defined by the mechanical description of a gas in T = stat. mech.) one must express temperature in terms of mechanical terms, i.e. one needs also bridge laws.
I find it very helpful to give examples, especially because sometimes it turns out that two people use different words for the same thing. That would become clear by examples. Suppose I am given S(t0) = the present position of Mars, and I am to calculate S(t1) = the position of Mars one year from now. I can use either Newtonian or Einsteinian gravitation theory, (NGT) or (EGT). To determine that, I have to be given the accuracy with which the answer is requested. If there is only one theory available, this is of course no problem, but otherwise, one needs an ee-question.
To clarify terms: I speak of a THEORY (NGT, say) and of specific MODELS within that theory (e.g. the sun and the earth ignoring all the other planets) the theory has a well-known validity limit (relative to EGT for example), but the model is in general more limited than that (e.g. by the error of ignoring the influence of the other planets). This gives me an opportunity to discuss reduction, our main topic. Mathematically, there is no problem to derive the equations of NGT from those of EGT. But this is not the whole story! (It is for Weinberg, but not for a philosopher of science). One notes that NGT involves a flat spacetime in which the individual masses (planets, stars, etc.) are embedded and act by forces of action-at-a-distance variety on each other. That perception is entire different anf foreign to EGT where there are NO forces, the action propagates with the speed of light, and there is a curved spacetime. This CONCEPTUAL jump from EGT to NGT is completely ignored by Weinberg. He ignores the fact that the symbols in the equations of EGT and NGT have very different meaning (interpretation). And it is this fact that prevents reduction from being coherent. The bridge laws that provide the crutch that leads from one set of concepts to the other set (often incommensurable), must be added AD HOC. Thus, even within physics, reduction does not carry through in this sense, not to speak of reducing chemistry to physics, etc.
I would like to make the following remarks, which may serve to outline my position on the issues under discussion.
1. I am sympathetic to what seems to me the basic intuition behind Weinberg's paper: that physics is complete in the sense that there are no such things as an "elan vital" or a "Lebenskraft". Moreover, I think that this position is part and parcel of modern science and can hardly be called controversial. But I think that the way Weinberg formulates his views is rather awkward, and perhaps even incoherent. Weinberg distinguishes between grand reductionism and petty reductionism; the former pertains to the general validity of fundamental laws, whereas the latter refers to the relations between the constituents of matter on different levels of complexity, aggregation, etc. However, I do not think that it is coherent to speak of laws as if they were entities existing in their own right. Laws are about the behaviour of physical systems; it makes no sense to discuss laws without specifying the systems they are about. I therefore do not see how it is possible to maintain that the fundamental laws of physics apply on all levels without entering into an analysis of how different levels of complexity interrelate as to their material constituents. For example, it has no concrete meaning to say that I am subject to the equations of quantum field theory if it is not explained how I can be regarded as a system quantum field theory is about; that is to say how I am built up from elementary systems. But this leads us directly into Weinberg's petty reductionism. I therefore think that the distinction between grand and petty reductionism is not suitable for Weinberg's purposes. Still, I am in complete agreement with Weinberg that what he calls petty reductionism is in many cases a programme that is practically impossible to work out in a detailed way. As he points out, even a seemingly simple case as the relation between a proton and its constituents on a deeper level (quarks) turns out to be terribly complicated. However, given that I do not accept Weinberg's retreat to grand reductionism, I think that the conclusion must be that it is unavoidable that different levels of reality are treated with their own laws and own concepts, without clear and general reduction relations between the levels. Nevertheless, sometimes it is possible, for specific cases, to demonstrate that a system can be treated on different levels and that the results agree. This lends credibility to the basic idea, mentioned above, that fundamental physics applies to all systems in reality. But as just argued, this idea should not be elaborated into the direction of traditional reductionism, i.e. it should not be taken to mean that general and precise relations can be established between the concepts and laws on different levels of reality. It is more appropriate to express the fundamental idea by means of the concept of supervenience: all properties of systems, described on whatever level, supervene on the fundamental physical description. By this I mean that if two systems have exactly the same state as described on the most fundamental level, they perforce have the same properties on whatever other level of description. Two chess players who share the same fundamental physical properties, and the same conditions, will be in the same mood and have the same intentions. This, it seems to me, is what modern physicalism says and what Weinberg defends. The formulation via supervenience has the advantage that it is not necessary to maintain that there are unique or simple relations between the various levels of description. In the case of the example: it can remain open how the thoughts and intentions of the chess player are connected to the processes in his brain, etc.
2. I concur with Weinberg that physicalism, thus conceived, is an important element of the scientific tradition of the West. I don't think this physicalism is very controversial in scientific circles: there surely is a fair amount of evidence that physical approaches are often possible and successful within the boundaries of other disciplines whereas clear evidence of the opposite seems lacking. However, it must be admitted that the evidence is not conclusive. Still, it seems to me that the onus of proof is on the shoulders of those who contest the physicalistic worldview. Now, I am not sure whether I understand Prof. Wojcicki's argument relating to the rational decisions of a chess player. If the argument is meant to illustrate that we describe the behaviour of chess players not in physical terms but by means of concepts having to do with strategy, moods, genius, laws of rationality, etc., and that the relation with the underlying physics is far from clear and moreover not very interesting, then I could not agree more. But in that case I don't see any problem for physicalism (see my explanation of physicalism above) and Weinberg's outlook. However, if the argument is meant to show that physicalism is false, I am not sure exactly what the argument is. Is it that laws of rationality cannot supervene on the physical description? Why would that be the case? Has this to do with the claim, also made by Wojcicki, that human behaviour is indeterministic? Usually, the notion of determinism is regarded as being relative to a theoretical description. Under what description is human behaviour indeterministic; under a psychological, sociological or physical description? And what is the relevance of the determinism issue for the problem of supervenience or reductionism? So I need some clarification on this point.
Dennis Dieks
Professor of the Foundations of Science, Utrecht University
Dear panelists,
I was not able to contribute to our panel discussion for a few days for I have moved from Poland to the US where I am going to stay till the middle of January. Note please that my e-mail address may change. The recent (wojcickir@aol.com) expires on November 1.
Welcome Dennis Dick in our group. I believe his contribution to our discussion might be very useful. It certainly should allow us to clarify several points. My own comments on his contributions are the following.
COMMENT 1. Dennis introduces the idea of a physical system and insist that laws do exist only as concerning specific physical systems. I think this position needs some qualifiacation.
We have to -- this is not a new idea in our discussion, see e.g. some points made by Jan -- discern abstract systems (i.e. certain theoretical constructs in terms of which we seek to understand the real world) from concrete physical entities. Needless to say that physical theories taken literarly concern abstract systems; physical laws describe some ideal situations which might correspond, often in a far going way, with events one encounters in the real world. In order to use some physical laws L as in order to account for a specific physical phenomenon P (meant to be something real) the user has to be able to adequately judge whether can P be reasonably treated as one of those abstract syetems to which L are applicable. The idea of "reasonableness" which appears in the above context is pragmatical. Whether or not treating P as an abstract system whose behavior is described by L depends on the expectations of the users. This is the user who sets the limits of error acceptable in the needed description of P, and this is the user who decides how reliable should be those conclusions concerning P's behavior the user wants to get by applying L to P.
In fact the situation is even more complicated. As a rule we do not apply physical laws to a physical phenomenon P directly but rather via theoretical models formed in order to handle certain specific questions Q concerning P. If instead Q one seeks answering to another set of questions Q' it may happen that the models good for handling Q are not good for handling Q'. Models are forerme by combining laws with some additional hypotheses whose selection depends on the problems the model is expected to solve.
In the context of our discussion it seems of little interest to speak about physical laws as concerning abstract systems. But then we should resist temptation to introduce the idea of a physical system as an obvious and direct counterparat of abstract systems. If this is granted then Dennis' idea that a law should always be trated as something chartacteristic of specific systems needs some qualifiacations.
The process of scientific inquiry need not start with selecting specific physical systems in order to discover regularities that govern their behavior. It might as well be the other way around. We might discover a regularity (e.g. a pattern of behavior) even before we are able to define in an accurate enough way the class of systems of which that regularity is characteristic. If so, we may as well reverse Dennis statement and say that physical systems are not entities existing "on their own right"; they do exist only to the extent to which they display certain specific regularities.
There are two points I am trying to make: (1) Neither the idea of a physical system nor that of physical regularity can be defined independently from on another. Neither is prior to the other. There are no regularities which do not apply to specvific physical systems, and there are no physical systems which are not goverened by specific regularities. (2) Unless we are concerned with some purely mathematical ideas which are to provide some idealized account of some phenomena, neither the idea of physical systems of a specific kind nor that of regularities characteristic of those systems can be defined in a fully precise way.
COMMENT 2. Dennis points out that Weinberg's idea of grand reductionism might not suffice to account properly for how scientific laws might be related to one another. I fully agree, and I believe Weinberg would not question this view either. On the other hand, the way in which I understand Weinberg's opposition of petty and grand reductionism might not the same as that suggested by Dennis.
Weinberg's position seems to be the following. Yes we might not be able to produce all the calulations which are necessary in order to show that, say, laws of chemistry are formally derivable from QM. In other words if we insist on petty reducibility of chemistry to physics, we might not be able to prove that such reducibility obtains. So what? Should we conclude that our knowledge of quantum phenomena is incomplete and start looking for laws that apply to chemical phenomena not being derivable from the laws of QM? The fact that we do not knwow how to derive laws of chemistry from laws of QM, in other words the fact that we do not know how to reduce chemistry to QM in a "petty" way does not prove that QM is not a sufficient foudational basis for chemistry, i.e, that the latter is reducible to the former in the "grand" sense of the word.
If I am correct Weinberg treats the idea of grand reductionism as normative rather than descriptive. Any physical theory, say QM, as we know it today might well be incomplete. Yet, in order to improve it one has to discover phenomena which the theory handles in a wrong way. The fact that we might not know how to effectively apply QM in order to account for some specific events (e.g. some chemical processes) neither gives good reason for questioning adequacy of QM nor justifies declaring chemistry non-reducible to QM. Moreover it gives us no cues how to improve either QM or chemistry. Any research program which rejects the idea of physicalism (grand reducibility of all laws of science to laws of physics) is bound to be cognitively idle.
Is the above consistent with the way in which Dennis seeks to account for the idea of grand reductionism when he offers the following example: "Two chess players who share the same fundamental physical properties and the same conditions will be in the same mood and have the same intentions. This it seems to me, is what modern physicalism says, and what Wienberg defends"?
There is a rather subtle point to be discussed at this juncture. If two chess players are to be treated as some concrete persons (not just as some abstract entities) they certrainly never share the same physical properties and the same conditions. Consequently, what Dennis maintains is true but unfortunately it is true for purely logical reason: the hypothesis of the conditional is false. But this remark does not seem to undermine the point Dennis is trying to make. If am correct Dennis believes that whatever might be the intentional (psychological) characteristic of the players it has its physical countertpart. Well, I do not have any firm opinion on this matter. The idea of psycho-physical parallelism might be perfectly correct. But it might be wrong as well. There is no compelling reason to claim that whatever is in our mind must be somehow determined by the state of our body.
But let us leave the above question to philosophers. What matters from the methodological point of view is that the idea of psycho-physical paralelism is idle. It is idle just in the same way as the the idea that there might be laws of chemistry that are not reducible to physics. It does not result in any research program worth being pursued.
SOME ADDITIONAL REMARKS
I am afraid that only too often the debate on physicalism, reductionism, etc. hides another problem. The problem is: what is science? More specifically: what is a scientific law? what counts as a scientific discovery? The claim that all scientific dysciplines are reducible to physics need not be applied as a descriptive sentence. It may as well serve as a stipulation: a discipline or any its part which is not reducible to physics is not scientific. A regulartity which is not reducible to physical laws is not a law. A discovery which is irrelevant to our knowledge of physical aspects of the word is not a scientific discovery.
Two things I consider fully unreasonable. Firstly, to deny the scientific status of such disciplines as medicine, economics, linguistic, sociology, psychology, etc. just because the regularities they discover in no obvious way are reducible to physical laws. Secondly, in a sense to maintain just the opposite: to maintain that all the genuine laws those disciplins are able to discover are dermined by physical properies of the examined objects.
We have to face two things: (1) unless we take dogmatic and very narrow point of view, the fact that a discipline is not reducible to physics in no way proves that this discipline is not scientific. (2) If the reducibility issue is discussed it might be good to keep in mind that there is a rather substantial difference between the way in which dysciplines close to physics, say chemistry and social dyscyplines, say economics, are related to physics. Let me dwell on the second point.
All laws of chemistry are causal. They are causal in the sense that they describe results one may achieve acting in a specific way. Moreover they are causal in the sense that in order to decide whether a specific law is applicable to a specific situation (whether it will both acurately and reliably enough account for a specific chemical proces) the applicability of the law in question should be examined in terms of causal conditions that guarantee or prevent such applicability. Now, in spite of what we often incline to believe, the laws of economics are not purely causal. They are (also) intentional.
Weinberg puts forward arguments of a very different nature in his plea for reductionism. When he speaks of what he calls 'grand reductionism' it is hard not to get sympathetic for this idea. Indeed most of the beautiful stories of science, some of which are mentioned by Weinberg, are, at first sight at least, connected to a dynamic of what Weinberg calls grand reductionism. Weinberg puts forward the idea of grand reductionism in the following way: 'the view that all of nature is the way it is (with certain qualifications about initial conditions and historical accidents) because of simple universal laws, to which all other scientific laws may in some sense be reduced'. Weinberg also uses very much the concept 'explain' in his argumentation. We cite: " Though wrong. Thales and his pre-Socratic successors were not just being silly. They had somehow come upon the idea that it might be possible to *explain* a great many complicated things on the basis of some simple and universal principle - everything is made of water, or everything is made of atoms, or everything is in flux, or nothing ever changes, or whatever. Not much progress could be made with such purely qualitative ideas. Over two thousand years later Isaac Newton at last proposed mathematical laws of motion and gravitation, with which he could *explain* the motion of the planets, tides, and falling apples." Its not unimportant to remark this, because I indeed believe that it is about 'explanation' and 'understanding'. Darwin has *explained* the 'origin of the species', by a mechanism 'natural selection' that we all can *understand* how it works. This search for 'understanding' and 'explanation' is the driving force of all science, and it should be cherished and treated with respect. It is however not 'just' a search for reduction. What happens when this search for understanding and explanation are successful is more complicated than what people will usually understand under reductionism. Weinberg should be aware of the fact that it is not just by chance or out of mere stupidity that often reductionism is understood as what he calls 'petty reductionism'. The dynamical process that takes place in these great moments of the history of science, and that brings us 'understanding' and 'explanation', and also takes place each day in the life of a researcher on a little scale, I would be tempted to indicate with 'synthesis', much more than with 'reductionism'. Scientific research, but in a more general way this is true for the global cognitive exploration of humanity, takes place as an interplay of synthesis and analysis. Analysis stands for the detailed and specialised exploration, always deeper and deeper, and deliberately fragmenting and idealising and cutting it out of the whole, while synthesis stands for the opposite movement of integration, taking it back to its place in the whole. Like Weinberg speaks about grand reductionism, in these parts of his texts were I would explicitly be in favour of his plea, it is more synthesis that he means. One of the aspects of such a synthesis is that the old views are often shown to be just wrong in most of their important aspects, such that it is hard to interpret these old views as now 'reduced' to the new ones. Sometimes they still show up as 'special cases' of the new view, like the example of Newtonian mechanics in relation to special relativity theory. We can 'understand' why Newtonian mechanics worked so well, because special relativity becomes mathematically isomorphic with Newtonian mechanics if we take the limit of the velocity of light to infinity. And since the velocity of light is very big in reality, this 'explains' why Newtonian mechanics worked so well. But this well known example can also be very misleading: indeed, the nature of the reality that is described by special relativity, certainly when it is regarded as a special theory of general relativity, is very different in so many aspects of the nature of reality as it was described by Newtonian physics. Fritz has referred to some aspects of this difference by announcing the necessity for 'bridge laws' that are ad hoc [FR Oct 17]. And we must not classify this lack of coherence (as Fritz calls the reductionist programme, see again [FR Oct 17]), as something we have to try to avoid. No, the real revolutionary changes that science has brought us 'always' have to do essentially with a rather profound changes in our interpretations, so the rise of new world views, completely different from the old ones and certainly not being able to reduce them. We only have to compare the world view that incorporated an explanation for the origin of species before the rise of modern science (but Weinberg would certainly not agree about the fact that there was also 'explanation' present there) and the world view in which we now, after the revolution provoked by Darwin, understand the origin of species. One does not have to underestimate the explanatory power, at that time and not now of course, of the pre-modern western world view, synthesis of the Christian beliefs and the Greek 'pre'- scientific proposals. Of course we have good reasons to believe that the cognitive exploration of mankind of our surrounding reality does not just consist of ever changing world views, but that there is also real progress. We know more about the nature of reality now, and we can explain more and we understand more. And reductionism and even petty reductionism has played a major role in this recent conquest of humanity. But also the limitations of the reductionist programme, that recently have been, and partly this is due to fashion, which has to make us prudent, have been put into evidence, adds to a greater 'understanding' and explanatory power of science. The insight that reductionism, be it petty or grand, may not be the ultimate situation of the nature of reality, is a synthetic type of insight. I feel myself very much now for a view about reality that is more or less the following: reality has a layered structure, were different the different layers are of course not neatly separated, they are causally and dynamically connected. We can give names to the different major layers, but of course this is already a big over simplification: the pre-material layer, the material layer, the biological layer, the social layer and the cognitive layer (for example, but I do not want to go into to much details, and just use this model to make clear what I want to state about reductionism). It seems that it is proper to speak of some fundamental order or hierarchy, in the sense that the later layers, that have evolved out of the forgoing layers, could be called 'higher' (pre-material has evolved material has evolved biological has evolved social has evolved cognitive). The different layers have their laws and regularities, that are not independent of the laws and regularities of the other layers. There is of course an old fashioned reductionist connection: the jaws of an ant will have to obey the physical laws of Newtonian mechanics (in principle even the laws of general relativity). But they will also have to obey the biological requests of collecting food and the social necessities of protecting the ant population against enemies (this example was told to me by Francis Heylighen, one of our collaborators in the Centre, so I just want to mention this). This means that generally the laws on a higher level also have an influence backwards in the direction of the lower levels: this influence was called by the late D. T. Campbell by the name of 'downward causation (see Campbell D. T., Levels of Organisation, Downward Causation, and the Selection-Theory Approach to Evolutionary Epistemology, in, Scientific Methodology in the Study of Mind, E. Tobach and G. Greenberg (ed.), (Erlbaum, Hillsdale, NJ), 1, 1990). In the old fashioned reductionist view the laws and regularities of the higher layers (biological, social and cognitive) would be classified with what Weinberg calls the historical accidents. It is a great synthetic idea - and a new idea of this century - that this would not be the case, and that these higher levels have their own regularities that are connected fundamentally with their specific nature. This does not mean that the layer traditionally described by physics is not the most fundamental one, but it does throw a different light on the possibility of reducing all of reality to the laws and principles that are encountered in this layer. It changes the explanatory power of such an operation, and it also opens the possibility - a very interesting and again synthetic hypothesis that was first explored this century - that perhaps the laws and regularities that we encounter in higher layers could throw some light on the nature of the laws and regularities of the lower layers (this is perhaps similar to the thought expressed by Dennis [DD Oct 17], were he puts forward the idea that laws go together with systems). We have been questioning whether physics could be a theory (or approach) that could ever lend a theory for the description of all of reality. In this sense I do think that physics, but most of all the mathematical structures that are used in physics, are special. Einstein once said that one of the greatest mysteries of science is connected to the question of why mathematics works so well. Even if we would proceed towards a model - like for example the layered model of reality, but there could be others - were reductionism would not be a basic principle, it could well be (and myself I am tempted to believe this) that we will be able to construct theories of a most general nature and theories using the methodology of physics of trying to axiomatise by mathematics. The reason for this is, in my opinion, that mathematics explores the nature of reality from the other side, always taking into account an axiomatic approach from the start, and leaving all flexibility of adaptation to the different possibilities of reality in also typically layers that are not described by standard physics.
Let me try to open another can of worms.
Downward causation, mentioned in the latest entry by Dirk, goes against physical reductionism and the analytical approach, because it makes claims about events that go beyond what can be inferred from the knowledge of pieces and their interrelations. If downward causation can be explained physically, it ceases to be downward and becomes "upward". I would like to examine this argument in detail.
Unless other participants in the discussion object, let me introduce into our consideration computer models of consciousness. In the area of artificial intelligence systems are built based on introspection of our perception, planning, discovery, learning, etc. Those models, sometimes implemented as so called autonomous intelligent robots, are more concrete and practical than physical models.
All events in our consciousness, including free decisions and intentionality, provide perrenial arguments against reductionism.
Our consciousness has been a challenge to scientific reductionism and now it defies computer-based reconstructions of mind. To me, in everything that occurs in our consciousness there is a distinct component which cannot be captured neither in physical nor in computer models. We can call this component ``intrinsic contents'' or `` conscious contents'' and we distinguish it from information contents.
Reduction of mind to a computer system meets the same problem as scientific models. Intrinsic contents are equally redundant in both cases. The functioning of a computer model is limited to states of computer memory and to instructions which transform memory cells. Those are the only building blocks used when we try to represent conscious mind in a computer system.
No contents, additional to memory elements and computational steps, is needed for a computer to execute the program. At each level, from the machine level to representations in high-level languages, the problem is the same. On the machine level instructions occur in the form of sequences of 0s and 1s. The processor uses those instructions to transform elements of memory, also expressed as sequences of 0s and 1s. A program in a higher-level language uses more complex data structures and instructions that operate on those data. But the instructions are formally defined and no extra contents is needed to conduct the entire computation step after step. I have in mind the actual computation, not representing computations by humans who try to understand it or explain to other humans.
Robots are computers equipped with sensors, manipulators and programs which interpret the data coming from sensors, make decisions and direct the physical actions of manipulators. In addition to making computations, robot's computer sends to sensors and manipulators strings of characters (ultimately 0s and 1s), which trigger behavior of simple processors embedded in those devices. The computer receives back from the sensors the data formatted as 0s and 1s. And that is all to the functioning of a robot.
Let us consider a hypothetical attempt at a conscious robot. Consciousness is often viewed as a central decision-making system -- a necessary integrator of sensing, planning and acting in humans and animals. Consciousness picks up goals, recognizes the situations, and decides on actions. Computer systems in the domains of artificial intelligence and autonomous robotics use central decision making in numerous ways. But the operation of those systems consists of the same computational steps applied to the same memory elements as considered earlier. Whenever we try to represent intrinsic contents we can only produce yet another mechanism that performs formal instructions, ultimately rearranging 0s and 1s. The functioning of such a system does not require any intrinsic conscious contents. Every detail can be described without them. Thus, computer reconstructions of internal contents of consciousness, if successful, would demonstrate redundancy of consciousness and its intrinsic contents.
It is not possible to demonstrate that robots do not possess intrinsic contents of consciousness, but we know they can do without them. We do not know how such internal states could emerge in a computer system nor what would be their role in the system's behavior.
Dear panelists, The recent contributions give me a welcome opportunity to attempt a clarification of some points of my own. They also prompt me to ask for clarification concerning some issues raised by the others.
1. [RW,23 Oct]
I agree with the greater part of what is observed by Ryszard about the connection between physical systems and laws. In particular, physical systems are characterized, to some ex- tent, by the laws they obey. I add "to some extent"; because if systems were to be defined completely via the laws they are subject to, it would become an analytical statement that an electron, say, obeys the laws of quantum electrodynamics, say. And that is clearly not the way the concept "electron" is used in physics. So I think that part of the meaning of a term like "electron" is given by criteria which are insensitive to the exact theory under discussion. But I agree with Ryszard that there is an intimate connection between laws and the systems they apply to. This seems to reinforce the point I made in my previous contribution, namely that it makes no sense to speak of laws as detached from phy- sical systems; and that it therefore is impossible to have "grand reductionism", on the level of laws, without "petty reductionism", on the level of systems. When I said in my previous contribution that laws cannot exist in their own right, I also had in mind the following. Physical systems are commonly regarded as being part of the world's ontology; whereas the same is much less natural to assume for laws. In an empiricist philosophy of science laws are not considered as "things existing out there", as in a kind of code of law to which nature has to conform. From this perspective, if reductionism is ever to succeed in establis- hing connections between the "layers" of the world (an approp- riate picture introduced by Dirk), it has first of all to suc- ceed on the level of systems (and this is what Weinberg calls petty reductionism, if I understand him correctly) and only then on the level of laws. So I still think Weinberg's insis- tence that we should disregard petty reductionism and strive for grand reductionism is rather unfortunate.
2. [RW, 23 Oct]
I agree with Ryszard that Weinberg's intuition is that physics provides a foundation, e.g. for chemistry, even if we cannot derive chemical laws from physical ones. This is exactly what I tried to make more precise by means of the notion of super- venience (a notion which originates, I think, from discussions in the philosophy of mind). The idea of physicalism, expressed by means of supervenience, is that the physical state of the universe fixes the states assigned by chemistry, biology, etc., even if we are not able (yet) to formulate the relations between laws etc. The physical level is assumed to be causally closed, and complete. Note that this is intended as an attempt to make clear what physicalism means, not as an assertion on my part that physicalism is necessarily true---though I am inclined to defend physicalism as long as no clear counterevidence is offered.
I do not think that research programs that do not aim at deri- ving laws at one level from laws at another level are "cogni- tively idle". We can have evidence for relations between le- vels, and knowledge about this, without being involved in the derivation of theoretical reduction relations. Think of the medical doctor who gives his patient some drug against schi- zophrenia, without any theory in the background about psycho- logical laws and their possible deductive relations to laws of brain processes. The relations that we know and use have a piecemeal character and have not been subsumed under general laws. It can even be doubted whether general psychological laws can exist. A research program that aims at finding such isolated relations seems worth-while pursuing. If the research program succeeds, this gives evidential support to the validity of physicalism.
I have a question concerning what Ryszard writes about the relation between intentional or psychological states and phy- sical ones. I understood the beginning of his 2nd comment as a defence of physicalism, with respect to the relation between QM and chemistry. He says that any research program (concerning the relation between QM and chemistry) which does not embrace physicalism is cognitively idle. But with respect to the relation between psychology and physics (or chemistry) he seems to defend the opposite point of view, namely that it is cognitively idle to pursue a research program investigating this relation. The difference seems to be that in the case of psychology, as opposed to chemistry, we have to do with in- tentional states. But I do not see why the difference cau- sal/intentional is correlated with a difference physically reducible/irreducible, or with a difference in cognitive value of research programs. It seems to me that some of the most exciting present-day research programs in psychology and arti- ficial intelligence are based on the attempt to understand in- tentional states in terms of the underlying physical/chemical processes. Probably I haven't understood Ryszard's intentions here.
3. [JZ, 27 Oct]
The viability of physicalism depends strongly on whether it is possible to meet the challenge of explaining consciousness. I think substantial steps forward have been made (Dennet, Chur- chland, etc.), but I am not at all an expert in the field and suppose that the question is still very much open. Jan offers an argument that seems to exclude the possibility of a physi- cal explanation of consciousness from the outset. He makes the distinction between intrinsic contents and information con- tents of memory, and says that machines can only have the latter. I would like to ask what the distinction exactly is; and how we can be sure that we have intrinsic contents that cannot be described in the way the contents of a robot's me- mory are described.
4. [DA, 28 Oct]
I very much sympathize with Dirk's idea of a "layered" struc- ture of reality. That is, I subscribe to the idea that there are various levels of reality, with their own laws and con- cepts. But I am inclined to combine this picture with the idea of physicalism---I wonder whether Dirk would be opposed to that. For me, the various levels are relatively autonomous in the sense that we are not able to derive general deductive relations between them and that, moreover, it would be nonsen- sical to discuss subjects like animal behavior in terms of elementary particles ; but still, I assume (until counterar- guments convince me of the untenability of this position) that all levels depend (supervene) on the basic physical one. This dependence is reflected in the fact that there exist concrete relations between the levels: Dirk's ants will behave diffe- rently, and betray different intentions, if I mix some drug in their food. Within this picture I can understand "downward causation" as a way of speaking about the relations between higher and lower levels ("the jaws of the ant require such and such a material constitution, in order to be solid enough for chewing"). Be- cause the ant is still completely determined by its physical make-up, I have no problem in reconciling this with physica- lism. But would Dirk perhaps disagree and maintain that the physical layer is not self-contained and causally closed, and that it is literally open to non-physical causal influences "from above" (the higher layers of reality)?
Dear panelists,
This is to comment on selected points of the latest Dirk's contributions. I will not address Jan's contribution; I plan to do this in my next letter. Neither Dirk nor Jan support Weinberg's (advocated also by Dennis) idea of reducibility (grand reducibility, to be sure) of all the scientific disciplines to physic. Dennis seem to be the only person among us who supports Weinberg's idea of grand reducibility.
Let me repeat, however, what I stubbornly keep saying: As long as we insist on advocating a metaphysical version of either physicalism or anti-physicalism we have little chance, if any, to prove that the view to which we subscribe is right, while the opposite is wrong. This is not to say that I ignore the relevance of metaphysical considerations, They motivate our most general ideas (worldviews) which serve us as general frameworks within which our scientific theories are formed. On the other hand, what really matters is not metaphysics but the methodology motivated by metaphysical views.
If physicalism is to have any relevance to the process of science formation, it must not reduce to the metaphysical view that the
(MF) PHYSICAL LAWS ARE THE ULTIMATE LAWS OF ALL THE NATURAL PHENOMENA.
It must also be promoted in its epistemic version:
(E) PHYSICAL LAWS ARE FUNDAMENTAL FOR OUR UNDERSTANDING OF THE PHENOMENA WE SEEK TO ACCOUNT FOR BY DEVELOPING SCIENCE.
Actually, it must also be promoted in its methodological version:
(MD) IF YOU DO SCIENCE, SEEK TO FORMULATE ALL THE REGULARITIES YOU ARE ABLE TO DISCOVER WITH THE HELP OF TERMS WHICH, AT LEAST IN PRINCIPLE, CAN BE DEFINED IN THE LANGUAGE OF PHYSICS.
One of the central claims of Dirk's (DA Oct 27) contribution was the claim to the effect that (MF) does not formally imply (E). It does not. But the best way to kill the idea of physicalism is to keep it apart from its epistemic or methodological versions. This is why, as I am going to argue, questioning Dirk position, that the idea of understanding (explanation) and the idea of reducibility are hand in glove with one another.
Note that the methodological version (MD) of physicalism appeals to the idea of reduction in a direct, though implicit, manner. Of special relevance for our considerations is the fact that reducibility of a language to another falls under the category which Weinberg dismisses as petty reducibility. On the other hand, unless we do not know how to execute that petty kind of reduction, the idea of grand reducibility remains a slogan deprived of any practical consequences.
Weinberg, who opposes grand reductionism to petty reductionism might easily put himself in a somewhat uncomfortable position of a person who is saying the right things yet ones to which nobody must pay attention. This, I believe, explain his tendency (criticized by Dirk) not to separate clearly metaphysical version of physicalism from its epistemic version; the two are being discussed by him as if they were equivalent.
A question which appears at this juncture is whether (E) can be kept separated from (MD). I do doubt it. The idea of understanding surely is pragmatical. To understand something is to reduce that something to the ideas which we consider primitive in the sense that they do not need any further explanation. In order to understand how the regularities examined in a specific domain are related to laws of physics, in other words how one can possibly explain them by appealing to laws of physics, one should be aware how the processes accounted for in terms characteristic of the domain in question can be adequately described with the help of (and thus reduced to) physical terms alone. The reducibility in question can be called *explanatory* for it is based on the informal ideas of justification, explanation, understanding rather than on an idea of a formal (mathematical) proof. Surely, I have to be more specific on what I mean by "explanatory reduction". I plan to to take this topic un in my next letter.
As you might gather from the above remarks, my views diverge from those expressed by Dirk (DA, Oct 27). The gist of that difference is I believe the following: Dirk, as in fact many theorists, tends to think about reducibility as formal relation, i.e. one that can be adequately defined in terms of logical (mathematical) inference. Unlike Fritz, I do not think that the idea of logical reducibility (and this is the kind of reducibility Dirk seem to advocate) is entirely useless ("dead" to use Fritz' word). Yet, I am not so much far away from Fritz' position as the above declaration may suggest. Indeed, on various occasions (see also CFrame) I have stressed the relevance of expertise, tacit knowledge, (i.e. ideas that cannot be adequately formalized) for reasoning in science. Fritz' "bridge laws" (the notion familiar from the hey-day of logical empiricism) are in fact another way of calling attention to the fact that reasoning in physics is not purely mathematical. There are some differences however between my and Fritz's position.
The bridge-law view encounters two difficulties. The auxiliary hypotheses applied to handle a specific situation, may involve presuppostions which taken literary are false. No planet -- to produce an example -- is a mass point. Nevetheless, some theoretical models of planetary motion are based on this presupposition. The "bridge laws" (as I understood the term) are hypotheses to the effect that a certain specific physical body possesses specific mathematical prpoperties (a planet is a mass point, its position is twice differentiable function, its mass reamin unchanged etc., etc.) There is no way to set a battery of bridge laws good for every occasion. Their selection differ from one application to another, change from one theoretical model set to handle a specific problem to another such model. Consequently, this is up to experts to judge on the base of their exepertize whether a specific way of forming a theoretic model is correct or not.
On numerous occasions (but not on all) physicists do agree which extra-theoretical hypothses (bridge laws) are acceptable in the context of specific theoretical cosiderations and which are not. The less divergent are their opinions, the more plausible seems to be the idea that they all share the same tenets which, if stated explitly, will transform their informal reasonings into formal proofs. I am afraid, unless the idea of a formal proof is treated in more or less "Pickwickian sense", this hope must be an illusion. It is an illusion for much the same reason for which it would be an illusion to belive that, say, measuring a distance between to points with a calibrated rod consists in applying always the same easy to describe procedure. In fact, given any specific pair of point the experimenter adapts the measuring procedure to the physical peculiarities characteristic of the physical situation in which the measurement is executed.
The way in which a theoretician reasons when they are dealing with a specific problem resembles the above describe situation to a large extend. If the same question Q is asked about two different physical situations S and S', one should not expect that all the parameters a theoretician seeks to control looking for the right model for handling Q are exactly the same in case S as in case S'. Eventually it may turn out that apparently "the same" question should be handled by different models. But even it turns out that the same model M is good for both S and S', it need not mean that the reasoning that justifies the selection of M is for both cases the same.
The above concludes my comments on Dirk's contribution. Let me sum them up as follows: the idea of understanding (explanation) which Dirk wants to keep separate from the idea of reducibility cannot be kept that way unless we ignore pragmatic (explantory) elements involved in any scientific reasoning.
==========
REQUEST: My task of the coordinator of the discussion is not easy. I have to guess in the right which problems you consider of special significance, to state them in a clear and acceptable for everybody way, to set some hirerarchy of them, and to focus your attention on the most urgent.
You might not be happy with the way in which I do this. I would appreciate both your criticism and assistance. Please offer your suggestions concerning the text I have called CFrame (the conceptual frame of the panel): propose its revisions, improvements and extensions. Unless we try hard to keep our exchange of views under control and develop it in an orderly manner we easily can pile up a set of contributions which hardly will form any consistent and complementary whole.
Will the language of a universal theory be physics?
Dear panellists,
In would like to react to some of the statements that have been made in [JZ Oct 29], [RW Oct 30], and [DD Oct 30].
I agree with Dennis in his remark to Jan's intervention. Explaining consciousness is one of the challenges posed to physicalism and reductionism. Would Jan agree about the fact however that this problem is not settled at all now? I would like to refer to the extensive pieces of analysis that are brought forward in the discussions inside the AI community and more notably by people as Daniel Dennet, Roger Penrose, William Calvin and Francisco Varela (who all have different opinions and their arguments to defend them). I have followed parts of these discussions with much interest and I have the strong feeling that we lack a lot of pieces of scientific information to even start to settle this question. I have my own opinion about this matter (again different from the four that I have mentioned, and also different from Jan's one I suppose) and would like to elaborate very much on this issue. But first I would like to ask Jan to state his opinion in a more detailed way if this is possible. Am I right if I state that Jan believes that consciousness needs a real extra-physical component to be explained (what he calls intrinsic contents?).
I do agree about Ryszards statement that the idea of understanding (explanation) cannot be kept separate from the idea of reducibility unless we ignore pragmatic (explanatory) elements involved in any scientific reasoning. And in this way I do have to explain better which are my doubts as to the idea of grand reductionism (and petty reductionism) brought forward by Weinberg. Nobody has stated this explicitly yet, but it can be well inferred from Weinbergs article and it is super-clear for those who know Weinbergs work, that Weinberg believes (and wants to defend) grand reductionism as reductionism to the theories of elementary particles as they are constructed and proposed now. Weinberg is very much against the Santa Fee way of thinking (he calls this place a heaven for anti- reductionists) were people are most of all expressing this idea that 'different layers' of reality have their own laws and regularities that 'perhaps' cannot be obviously reduced to the 'oldest' layer of pre- material reality.
The three different aspects of the problem called by Ryszard (MF) PHYSICAL LAWS ARE THE ULTIMATE LAWS OF ALL THE NATURAL PHENOMENA. (E) PHYSICAL LAWS ARE FUNDAMENTAL FOR OUR UNDERSTANDING OF THE PHENOMENA WE SEEK TO ACCOUNT FOR BY DEVELOPING SCIENCE. (MD) IF YOU DO SCIENCE, SEEK TO FORMULATE ALL THE REGULARITIES YOU ARE ABLE TO DISCOVER WITH THE HELP OF TERMS WHICH, AT LEAST IN PRINCIPLE, CAN BE DEFINED IN THE LANGUAGE OF PHYSICS. should indeed best be distinguished. However in my opinion for good science (and what is good science, this is of course another problem that we could reflect about, in my opinion good science does not have only to do with truth, but also with beauty, simplicity and fruitfulness) the three aspects mentioned by Ryszard should go along. I must however state that, the more I reflect about it, I do probably believe in some form of reductionism, although I find it a bad word choice in this sense, since it immediately will be classified as what we don't want it to mean, and that is the reason that I prefer to use words as explanation and understanding. I state it more clearly: I would advocate the search for a theory (or set of theories) that explain and make us understand reality as a whole, and I think that this is a possible goal. But I would not agree so much with the words PHYSICAL LAWS and IN THE LANGUAGE OF PHYSICS in all of the three aspects that Ryszard points out. To explain what I mean I can come back now to Dennis remark on my first intervention.
If we think about the layered structure model of reality, were each of the layers has its own laws and regularities, then I even would not exclude 'physicalism' from the start (and in this way it is possible that Ryszards words PHYSICAL LAWS and IN THE LANGUAGE OF PHYSICS can remain in his three aspects). I could very well be possible that the 'oldest' layer, the one described by physics - or we should be more prudent and say the one intended still to be described by physics - is the most fundamental one, and that all laws and regularities of the younger (and higher?) layers could ultimately be 'reduced' to (or explained by and understood by) the laws and regularities of the oldest one. Dennis states that he tempts to believe so and till the contrary is shown.
I mentioned the concept of 'downward causation' as introduced by Campbell, and this would perhaps indicate that I don't believe that such an ultimate reduction of the laws and regularities of all layers to the laws and regularities of the oldest layer could be found. I would like to clarify my opinion about this problem. If Dennis states: "would Dirk perhaps disagree and maintain that the physical layer is not self- contained and causally closed, and that it is literally open to non- physical causal influences "from above" (the higher layers of reality)?", then I would agree with him and say that I don't believe this (till the proof of the contrary of course), such that indeed a physicalist attitude is in principle still possible.
What I do believe however, and I should now reflect deeper about the way in which this belief gives us arguments against reductionism (perhaps not at all if I follow Ryszard) and physicalism, is that in all the layers the laws and regularities come 'partly' into existence during the construction process of this layer itself. In this sense one could say - if this belief is true of course - that the laws and regularities of a higher layer could probably not always be connected sense fully to the laws and regularities of a lower layer. This is not only a matter of principle I belief. In this sense I do belief that a theory that would try to explain the behaviour in all the layers will not be physics. One of the reasons, but there may be others, is that I don't believe that the concepts, entities, laws and ways of behaviour of the physical layer are best suited for such a programme. In this way I would like to mention - since also Ryszard has brought forward already the property of intentionality that is so well present in the reality of the social layer - this nice exercise of Daniel Dennet in his book 'Kinds of Minds' (Basic Books 1996) were he aims at a description of the physical reality from an Intentional Systems Approach, remarking that he does not make a 'scientific mistake' here. This also reminds me about the ideas of the philosopher and physicist Alfred North Withehead, who was very eager in stating that it is better to try to 'reduce' the 'simple' to the 'complex' as the other way around. I am more and more tempted myself, for example, to consider dead matter as a special case of living matter and this together, but on the same explanatory level, with living matter as a complex behaviour of dead matter. This is a good example of what I wanted to make clear in my first intervention, about the difference between reductionism and synthesis. If we would follow this idea, and explain dead matter as a special case of living matter, then we have explained without 'only' reducing it in the petty way (perhaps we have reduced anyhow in the logical and even philosophical way, as Ryszard would say). I have a strong feeling that the situation is more or less the following: reality allows a lot of patterns of behaviour of entities and interactions (different types of interactions, formation of bigger entities out of smaller, creation of new entities out of others, creation of new interactions ect...). In the different layers we are confronted with different possibilities of these ways of behaviour, and the physical layer gives us a rather narrow window as related to these patterns of behaviour. A theory that would describe all layers and also how they are connected would have to give a description of all these patterns and explain how some are special cases of the others (and why they are) and how indeed also some are built up out of others (and this would be a fruitful aspect of petty reductionism). I do believe that the complexity and chaos studies have partly started such a programme (but I do want to say immediately that some scientists advocating these fields of research use it in a bad way, in the sense really trying to be more 'post modern' than the post modernists and negating all possibilities of unification, and I would not agree with these ones). What I do believe is that the language of such a unifying theory will still me mathematics - something I tried to point out in perhaps not such a clear way in my first intervention, but it will not be physics as we now know it. I believe that fundamental principles that have been more easily identified in other layers of reality, because they were more obvious there - as for example Darwin's principle - will come into in on the fundamental level of principles and laws and entities.
Dear panelists,
In this letter I present and discuss some selected ideas of "Pluralistic Ontology and Theory Reduction in Physical Sciences" by Fritz Rohrlich, the paper quite long ago (it was published in Brit. J. Phil. Sci, 1998). I have been recently going through this paper once more, in order to discover once more how inspiring it is and how much relevant to our discussion.
The central claim of this paper is, I believe, the following: The cosmic evolution of the universe results in creation more and more complex objects (atoms, molecules, stars, planetary systems, organism, social systems, etc. etc.) Eventually a hierarchy of object of different complexity -- from the *finest* [I apply stars to mark technical terms of Rohrlich's paper] to the most *coarse* one -- is formed. Our knowledge expands by forming concepts (and associated words) that correspond to object located at of different level of the ontological hierarchy. The objects themselves are identified by our cognitive apparatus in terms of certain properties characteristic of the ontological level to which they belong. In this way our scientific knowledge divides into various cognitive levels associate with corresponding to them ontological levels.
In order to understand properly Rohrlich views the following two things should be explained.
(1) Rohrlich does not presuppose that *observables* --- object to which we have experimental access are located at a specific ontological levels of the external reality; they need not to be "directly oservable" in the sense familiar from the early time of logical empiricism. They may appear at various levels. The access to them may require resorting to sophisticated techniques, which in turn may require using notions available at a specific cognitive level. Observables, to be sure, are independent of the observational techniques (instruments, modes of observation).
(2) The correspondence between ontological and cognitive levels is established via *mature theories*. This notion plays a central role in Rohrlich considerations. A mature theory is a theory confirmed by numerous applications and does not changes any more. All elements of a mature theory (its formal structure, its informal semantics, its scope of applicability) are fixed and preserved with time. A typical example of a mature theory is Newtonian Mechanics, NM. We do not revise NM any more. The theory was superseded by more accurate ones (Special and then General Relativity) but not discarded; from that time we know its *validity domain* within which we may apply it safely.
Rather than treating a theory which has reached the mature stage as a partially adequate descriptions of the external world, Rohrlich (if I convey his position correctly) treats it as a cognitive counterpart of an "ontological levels", or perhaps I should say "ontological regions" of reality. [I guess, a specific ontological level may split into various "ontological regions".] One may ask: is the idea of an ontological level as something corresponding to a mature theory compatible with the idea of ontological levels determined by complexity of objects formed by cosmic evolution? There is nothing subjective in the proces of cosmic evolution while, obviously, the formation of mature theories involves various pragmatical elements?
I guess Rohrlich answer to this question might be the following. Cosmic evolution is an objective process, also growing complexity of objects that emerge in the process is an objective phenomenon. But the only way to expand our knowledge to new ontological region is to discover something which differs perspectively in an essential way from the already familiar objects. And this can be done by a elaborate inquiry process which through numerous experiment resulting in discovering various *empirical regularities*, through numerous *conjectures* that are steps toward forming adequate conceptual schemes of the phenomena examined eventually result in forming well established, firmly vindicated mature theory. Once this is done we might be confident that a correspondence between our cognitive ideas and physical entities of a specific ontological level has been established.
There are still various question marks left. What might be semantic correspondence that obtains between a mature theory and the relevant ontological level [or perhaps "region" -- I hesitate which words conveys better Rohrilch's ideas]? Theories, as Rohrlich's makes clear have only *formal ontological content*. The *factual ontological content* of scientific knowledge is established through empirical laws. One concludes that the relation between theories and ontological levels/regions of the universe is somehow fixed via the relations that hold between theories and empirical laws. Theories must save empirical laws.
[This exposition is not the right place to present my own ideas, so let me mention briefly only the following. Empirical laws, as well as any well established empirical evidence serves to vindicate theories. The links between a theory and real phenomena (theoretical concepts and factual counterparts of them) are determined in an indirect way via theoretical models that are formed with the help of the theory in search of solution to a specific question(s) concerning specific factual phenomena. I am afraid that without the idea of theoretical model that serves to handle factual problems concerning specific factual situations, the question of how a theory corresponds to factual events must remain enigmatic.]
To sum up the above presented Rohrlich ideas : Ontological levels/regions of the universe are products if cosmic evolution. But for us, people seeking to learn the regularities of the universe defined by mature theories and the corresponding empirical laws.
The unity of nature consist in the fact that all the existing objects are formed from some fundamental ones rather that in that the properties of more complex objects are definable in terms of properties characteristic of the fundamental objects. Rohrlich rejects the idea of ontological reduction consisting in forming a theory that accounts for the phenomena taking place on higher levels of ontological hierarchy in terms of phenomena that obtain at the fundamental level. In his opinion the emergence of a new ontological level always consists in emergence of something qualitatively new. We cannot account for the qualities characteristic of a certain ontological level in terms of finer objects and qualities characteristic of them. But we can to do this by resorting to idea of successive approximations. Objects of a new ontological level and properties corresponding to them emerge (Rohrlich resorts to this analogy) in much the same way as the irrational number pi results from a series of rational approximations 3, 31/10, 314/100, ... The accuracy of approximation can be viewed a certain parameter p which in the limit takes the value 0.
Rohrlich suggests that transition from a lower (finer) ontological level to the next more complex (coarse) is analogous to that which results in arriving to the irrational number pi through its finite approximations. In this context Rohrlich mentions such parameters as (v/c)^2 or Planck constant h. He also discuses an example which is of a quite different kind, for it is not related to any mature theory: By putting more and more dots on a sheet of paper one produces certain distribution of different shades of gray of increasing density. These shades conspire to make us suddenly recognize the face of Winston Churchill. In an analogous way, Rohrlich explains, light ray -- the central notion of geometrical optics -- emerges from a fine structure of superpositions of waves in electromagnetic theory.
I hope that the above gives a fairly accurate account of those parts of Rohrlich's paper that concern his idea of pluralistic ontology. There is no, as Rohrlich strongly emphasizes, the `only true ontology' rather through the process of formation of mature theories we continuously discover some new otologies all of the are equally "true". I will not try to present the arguments to which Rohrlich resorts in order to show that the ontological levels (and thus the corresponding mature theories) are not reducible to one another. One thing should perhaps be added. In Rohrlich's opinion, the problem of reducibility, hardly makes any sense when it is addressed to anything else than mature theories.
The above concludes my presentation of what I believe are the main ideas of "Pluralistic Ontology and Reduction in Physical Sciences". A few comments on those ideas I am going to offer below are mainly intended to even more define Rohrlich's views by opposing the to what I am going to call the "standard view".
COMMENTS:
The common wisdom is that our cognitive faculties are defective. Sometimes we "discover" something which actually do not exists. Such an apparent discovery (caloric may serve as an example) may result in producing a theory which eventually will be discarded. But there are more possibilities of arriving at a wrong notion of phenomena we seek to learn. Besides "discovering" something which do not exist we may overlook something which does and moreover under some circumstances may affect the behavior of things we are interested in. We may also develop a wrong idea of some regularity. Wrong but not "entirely wrong". The common wisdom to which I appeal tell us that truth can be more or less complete, that something can be approximately true, or partially true.
The above presented common wisdom view is actually commonly accepted in philosophy of science. From now on I will call it the "standard view". A part of the standard view is the tenet that Newtonian Mechanics, NM is a false theory -- it is based on some false presuppositions and some of the conclusions one can derive from NM directly or with the help of NM are bound to be false. This however in no way supports the belief (if anybody holds it) that NM does not have any claim to reality. Surely it does. We do know how to use NM --- within the scope of its applicability [the term *scope of applicability* corresponds to Rohrlich idea *limits of validity* but the two do not denote the same; roughly speaking the scope of applicability is the set of all those questions which can adequately be handled with the help of NM] --- in order to adequately account for numerous factual events; various questions concerning physical systems can be fully adequately answered by applying theoretical models formed with the help of NM. Note that if those questions define the limit of error within which they are to be answered, the answers to them have definite truth value: they are either true or false sentences not merely "approximately" or "partially" true.
The fact that according Rohrlich a mature well established theory must be true implies another important difference between his and standard view. The ontological perspective adopted by Rohrlich results in his arriving at questions which does not appear when one sticks to a more familiar semantic ideas. From the standard point of view there is no problem of reducibility of Newtonian Mechanics to say Special Relativity: one cannot reduce a theory which involves some false assumptions to a theory which do not involves them. The *correspondence issue*, i.e. the problem how to define the class of questions to which the two theories provide the same answers is not tantamount with the problem of reducibility.
Now the common wisdom to which I have appealed tells us that if there are two incompatible views one of them must be right and the other wrong. In this case I am not to subscribe to the common wisdom position. Rohrlich view and the standard view offer two different perspectives on how science is formed and accomplishes its main task: accounts for real phenomena. The differences between are chiefly of conceptual nature. There is nothing questionable in saying that for each mature and well established theory there is an aspect of reality which which that theory describes in fully adequately way.
Think about the factual events we seek to learn as available to us via some projections on our conceptual apparatus -- projections in much the same sense in which we speak about geometrical projections (mapping) onto a, say selected, surface. Are such projections something objective. Surely they are -- if the conceptual apparatus is given, they are fully determined by the properties of the events in question. The projected events are not tantamount with the events as such but they are also part of physical reality. NM, or another mature and well established theory can be viewed as an adequate (fully true) theory of the projection determined by its own conceptual apparatus.
If I prefer to stick to standard view rather to one advocated by Rohrlich it is not because I believe the latter is wrong and the former right. Rather I suspect that deriving from standard ideas creates more problems than it solves.
I could as well conclude my considerations at this point. But there is one more thing I wish to add. And this thing seems to me of great significance. Rohrlich deliberately refuses to deal with non-mature theories. In particular, in his opinion the problem of reducibility makes little sense if we try to discuss it dealing with non-mature theories.
The problem of reduction is primarily -- as I see it -- the problem of reduction concepts to concepts, not theories to theories. To reduce a theory to another theory is to adequately define its *central* concepts [the idea of a central concept is largely self-explanatory; Rohrlich applies it in its paper] in terms of central concepts of the latter -- "reduce" the former to the latter. Needless to say that given a specific set of concepts to be reduced and a specific set of concepts to which they are to be reduced, the problem of reducibility is the better defined the better defined are defined the concepts in question. The idea of a well defined concept might not be clear enough, not to mention that it cannot be defined in the same way for mathematical concepts, for theoretical concepts of factual theories, for empirical concepts and for common sense concepts. All these four categories of concepts are relevant for scientific investigations --- all of them play important role in scientific inquiry.
The point I have just made gives rise to a host of question. Surely, they cannot and they should not be handled in this contribution.
FINAL REMARK: I do hope that Fritz will find my synopsis of his paper acceptable. I am sorry if I distorted any of its tenets or presented them inaccurately. I hope I succeeded at least in this respect that I make it clear how important are his ideas for our panel debate.
Best wishes to everybody
Ryszard Wojcicki
Comments on Ryszard 10/30:
(1) "Understanding and reduction go hand in glove". I want to amplify on that. As I said earlier, one must distinguish *theory reduction* (TR) from *reductive explanation* (RE) (also called *explanatory reduction*). TR is the reduction of one theory to another, and since a theory is a complex involving language, ontology, math. structure, supporting empirical data, etc., it is essential that ALL of that is reduced. The reduction of the math. structure alone (which is what most physicists call "reduction", including Weinberg) is simply not sufficient. Such a reduction would mean that one can deduce the higher level from the lower level theory not knowing the higher theory ahead of time. I believe this fails because of the problem of language reduction.
RE is an explanation (and therefore implies a process resulting in understanding!) which uses a lower level theory. That's very different from TR. I am all for that. That's what scientists do all the time, and there is no problem with that. It does not have the constraints on it that TR has.
Thus, in my view, understanding only requires RE but not TR.
(2) I believe that MD is too strong.
(3) One must distinguish between idealization (a planet is represented by a point) and bridge laws which connect incommensurable terms ("heat" is "motion").
Comments on Ryszard 11/2:
Thank you for a very lucid summary of my position in "Pluralistic ontology...". I just want to add a few comments.
(1) I think of "scientific theory" as an abstract and general object, especially as far as its mathematical structure is concerned. Therefore, I distinguish it very clearly from the many "models" associated with it. A model is a specific system to which the theory can be applied and always involves IDEALIZATIONS. Example: Newtonian gravitation THEORY and the solar system MODEL. One can also use that same theory to discuss a model of the tides or of a galaxy, etc. There is an important difference between the validity of a model (due to idealizations) and the validity of the theory (in the above case, a breakdown because of too high a speed involved -special relativity-, or too high a gravitational energy involved - general relativity.)
(2) The concept of *emergence* needs a lot of attention. It is not well understood but extremely important. See for example the excellent article by Paul Humphreys in the Merch 1997 issue of Philosophy of Science.
Comments on Dirk 11/3:
(1) I fully agree that we don't understand consciousness and we can therefore not discuss how it fits into physicalism. However, there is no doubt that downward causation exists (for example from the mind to the physical level). Just think of intention as triggering physical action.
(2) Contrary to Weinberg, I am much closer to Santa Fe than to Weinberg. I have been very impressed, for example, by the very interesting work by Stuart Kauffman on Origins of Order.