|Essay on Robert Brandom, Making it Explicit
July 31, 2006
The Evolution of AWhy?@
1. The descent of content. Which way is up?
I have recently completed the bracingBand long overdueBexperience of reading Making it Explicit, and trying to superimpose my own views on it to see where the disagreements stand out. Mainly, there is agreement. Claim after claim, page after page, I find myself agreeing wholeheartedly with Brandom, and this is not just on matters about which philosophers in general agree. Brandom and I are on the same page about many issues that divide the profession. Moreover, on a few points where Brandom explicitly objects to positions I have maintained, I think he is, in the main, right, and I will acknowledge this in more detail below. What then is left to argue about? There is a fundamental difference of direction in our work. But I will try to show that our residual disagreementsBwhich at first make us appear poles apart on some very central thesesBmight quite easily be nudged into consonance if Brandom can take on what I view as a few friendly amendments.
We both learned a lesson from Wilfrid Sellars that still hasn=t sunk in with many of our colleagues. I quote, not for the first time, what I consider to be the pithiest expression of it in Sellars:
My solution is that A>. . . .= means - - -@ is the core of a unique mode of discourse which is as distinct from the description and explanation of empirical fact, as is the language of prescription and justification. (Chisholm and Sellars, 1958, p527Bdiscussed briefly by me in 1987, p[341)
The ineliminable, foundational normativity of all talk of meaning or intentionality was first insisted upon by Sellars, and Brandom=s version of the reason for this is comprehensive and detailed. Brandom chooses to adopt my Aintentional stance@ way of characterizing this unique mode of discourse, since the evaluative or normative presuppositions can be readily seen to be built right into the rules of that game. He then draws a distinction between simple intentional systems (what I call first-order intentional systems, entities whose behavior is readily interpretable by ascribing beliefs, desires and other intentional states to them) and interpreting intentional systems (what I call higher-order intentional systems, capable of ascribing intentional states to others and to themselves). Now which kind comes first?
The question is ambiguous, and once we sort it out, we=ll have exposed a central point of discord between Brandom and me. Of course Brandom and I agree on the obvious evolutionary fact: we interpreting intentional systems are a later-evolved, and much more sophisticated sort of agentBindeed human beings with language are the only species that uncontroversially meets the requirements, in spite of much hopeful research with apes, dolphins and other promising species.
Shouldn=t the earlier, simpler variety be viewed as more fundamental? As Brandom says, Ait is clear that there were nonlinguistic animals before there were linguistic ones, and the latter did not arise by magic.@ (p155) As he notes, Stalnaker and I follow this course, attempting to make sense of the Anonlinguistic@ or Ainstrumental@ belief of such animals as the fundamental notion. Brandom appreciates at least some of the grounds for the move--Athis is a laudable aspiration and it may seem perverse to spurn it.@ (155)Bbut he insists on adopting the Acontrary order of explanation @ :
the intentionality of nonlinguistic creatures is presented as dependent on, and in a specific sense derivative from, that of their linguistically qualified interpreters, who as a community exhibit a nonderivative, original intentionality. ( p152)
This might seem to suggest, bizarrely, that nonlinguistic creatures had no intentionality until we humans came along to interpret them
, at which point they acquired derivative intentionality, but of course that is not what Brandom means. What he means is that nonlinguistic creatures, long before human beings came along to interpret them, had a sort of cognitive prowess that is only with strain
interpretable as a sort of intentionality, modeled on the sort of intentionality that we can make sense of only as a phenomenon occurring within a full-fledged linguistic community: Athe understanding of intentionally contentful states that permits us to stretch the application of that notion and apply it in a second-class way to nonlinguistic animals (simple intentional systems)@ (p151). The order of explanation
is thus first to explain how linguistic communities carry on, relying on norms implicit in their social practices, and then use that account to provide a model, bristling with caveats and demurrals, for the more rudimentary phenomena exhibited by non-linguistic animals and other simple intentional systems.
This is certainly not a position I have ever defended in detail, but I=ve made a few gestures in this direction, and I agree with most of it. Now that Brandom has done such a fine job making out the case for it, I am happy to subscribe to itBup to a point. It is not just that (of course) it takes language-using human beings to create the very terms and practices of ascription that render the intentional stance visible and available to science as a model for understanding non-language using agents. That would be like the fact that you could hardly speak of some chemical reaction as a Adance of the molecules@ if there weren=t any dancing people to serve as exemplars. It is rather that the presuppositions of those terms and practices run deep, and until we understand them in their home territory, the deontic score-keeping game of asking for and giving reasons, we will not be able to understand the normativity that alone can give traction to a regress-stopper that we need. I take this to be the chief import of Brandom=s title, making it explicit
: until we language-users came along and made Ait@ explicit, it
didn=t really have intentionalityBat least not the full-fledged intentionality we need to understand before we can extend the concept to our distant ancestors and other languageless varieties.
What is the problem? It can perhaps best be seen by following Brandom=s vision of the contrast between his position and mine. (We will later soften the contrast.)
Dennett=s strategy of treating the normative significances of intentional states as instituted by the attitudes of interpreters does not by itself involve a commitment to reducing the normative to the nonnormative, insofar as it is the proprieties
of attitudes that are invoked. That reductive commitment comes in later, in explaining these proprieties. Understanding those proprieties in terms of predictive success, as Dennett does (a strategy different from that to be pursued here) gives an objective basis to the norms governing the adoption of the intentional stance. It puts Dennett in a position to say that talk of the predictive utility of adopting that stance is just a wayBindeed, the only one available to usBof specifying an important kind of objective pattern in behavior. (p58)
Brandom appreciates the need to ground his norms. They aren=t just prejudices or fashions that have unaccountably caught on, and he can=t resort to such anti-naturalist moves as God commands them
. But still, he does not want to commit himself to any of the familiar kinds of naturalism either, which leaves him with a delicate task:
the challenge: to maintain the stance stance1
toward both simple and interpreting intentional systemsBthat is, to acknowledge that the normative status of being such intentional systems is intelligible only by reference to the normative attitude of taking or treating something as such a system, that is interpreting it as oneBwhile at the same time securing the distinction between original and derivative intentionalityBand so not allowing the notion of intentional normative status to collapse into that of the normative attitude of intentional interpretation. (p60)
Oversimplifying somewhat, the distinction Brandom sees between my naturalistic way of Acollapsing@ intentional normative status and his way of avoiding the regress can be captured by the supposed contrast between the violated norms of faulty design
(my way) and the violated norms of a social transgression
Roughly, it=s the difference between being stupid and being naughty. At the same time, he must show that a violation of the social conceptual norms is not just a faux pas
; it=s a mistake that really matters. I think it is this last obligation that in the end he cannot fulfill without falling back, eventually, on something like my way of dealing with the problem. As he notes, I attempt to meet this requirement with predictive utilityBlife is short and time flies and decisions must be made now
by finite minds under considerable pressureBbut he finds my brand of pragmatismBin contrast with his own, actually quite similar, brand of pragmatismBto be unsatisfactory. What does he put in its place? Community.
2. Original and derived intentionality
To understand how this ingenious move is supposed to work, consider the threatened regress: ANoises and marks on paper do not mean anything all by themselves.@ (p60) They don=t have intrinsic intentionality but only derived intentionality. Derived from what? From the intentionality of interpreting intentional systems
of course. And where do the interpreting intentional systems get their intentionality? Here is where our answers differ. Brandom says: From each other!
AOn this line, only communities
, not individuals, can be interpreted as having original intentionality.@ ( p61) And membership in a (linguistic) community while in some sense optional for us, carries with it a commitment to the conceptual norms established and constituted by that community.
I, in contrast, have claimed that our own
intentionality is just as derived as that of our shopping lists and other meaningful artifacts, and derives from none other than Mother Nature, the process of evolution by natural selection. Consider a cartoonish dialogue about my version.
We come upon a scrap of paper with the symbols A2 qt mk@ on it.
AWhat might it mean?@
AAsk the agent who wrote itBthe robot R1D1.@
R1D1 utters: Ait=s my shopping list and it means buy two quarts of milk.@
ABut where did R1D1 obtain its intentionality (assuming it has some)?@
AFrom me, R1D1's designer. I made it to do my shopping for me, among other things.@
AAnd where did you get your intentionality?@
AFrom my designer, Mother Nature!@
ABut what kind of purposes and intentions does >she= have?@
None, in one sense, but plenty, in another: we can uncover the Afree-floating rationales@ of the design Adecisions@ that evolution has made, and thereby uncover the purposes and rationales of my organs and dispositions just as reverse engineering can uncover the purposes and rationales that account for the derived intentionality of my robot=s organs and dispositions. The fact that evolution doesn=t have purposes or intentions is not a bug but a feature: it permits us to end the regress of ulterior purposes with a whimper, not a bang. (The problem with God as the Intelligent Designer is that it forces the question of who designed God. The answer that nobody had to design God is just as unsatisfactory as Searle=s mysticalBor at least mystifyingBinsistence that it=s just a brute fact that human beings have original intentionality. That is why Searle=s vision of original intentionality is ultimately incoherent; it begins with an unacceptably marvelous and unexplained prime mover. Brandom=s account at least has the virtue of offering an explanation of how it is that the norms on which original intentionality depend play their roles. But he then has to face his own version of the question: whenceBand whyBcame communities?)
As should be clear from the little dialogue above, my strategy has been to help myself to communicating (or Acommunicating@) robots and their makers, and not make a big deal about the fact that this presupposes a community of communicatorsBthe robot-makersBso that I can draw attention to the way in which the robot=s intentionality (which I claim is well-nigh indistinguishable from ours in practice) depends on design considerations about which the robot-makers are the (corrigible) authorities. (They may have forgotten, or never understood, the good reasons why they made their robots the way they did.) In our own case, the parallel curiosity directs our attention to evolution. Just as we reverse-engineer the robot in order to understand its (derived) intentionality, so we reverse-engineer ourselves in order to understand the source of our own (derived!) intentionality. The only Aoriginal@ intentionality anywhere is the mere as-if
intentionality of the process of natural selection viewed from the intentional stance. Contrast this with Brandom=s declaration: AOriginal, independent
, or nonderivative intentionality is an exclusively linguistic affair.@ (p143)
Concentrating on the need to stop the regress of derived intentionality with a diminuendo
, not a crescendo
, I passed overBglissando
, to continue the leitmotifBthe issue of a community
of communicating robot-makers as if it might be a local accident of history, rather than a constitutive requirement, but I now see no reason not to agree that Brandom is right about thisBbut not that the story ends where he ends it. I am willing, that is, to put in a major fermata
at the place in the development where community comes in, and help myself to much of Brandom=s book to provide an analysis of this. In fact, I spoke favorablyBbut I admit, only in passingBof just such a strategy some years ago, and in that passage foreshadowed both my current agreements and disagreements with Brandom, which I will at long last attempt to make explicit:
[Michael] Friedman. discussing the current perplexity in cognitive psychology, suggests that the problem
is the direction of the reduction. Contemporary psychology tries to explain individual
cognitive activity independently from social
cognitive activity, and then tries to give a micro
reduction of social cognitive activityBthat is, the use of a public languageBin terms of a prior theory of individual cognitive activity. The opposing suggestion is that we first look for a theory of social activity, and then try to give a macro
reduction of individual cognitive activityBthe activity of applying concepts, making judgments, and so forthBin terms of our prior social activity. (1981, p15-16).
With the idea of macro-reduction in psychology I largely agree
, except that Friedman=s identification of the macro level as explicitly social is only part of the story. The cognitive capacities of non-language-using animals (and Robinson Crusoes, if there are any) must also be accounted for, and not just in terms of an analogy with the practices of language users. The macro level up
to which we should relate microprocesses in the brain in order to understand them as psychological is more broadly the level of organism-environment interaction, development, and evolution.
There is no way to capture the semantic properties of things (word tokens, diagrams, nerve impulses, brain states) by a micro-reduction. Semantic properties are not just relational but, you might say, super-relational, for the relation a particular vehicle of content, or token, must bear in order to have content is not just a relation it bears to other similar things (e.g., other tokens, or parts of tokens, or sets of tokens, or causes of tokens) but a relation between the token and the whole lifeBand counterfactual lifeBof the organism it >serves= and
that organism=s requirements for survival and
its evolutionary ancestry. (Dennett, 1981, as reprinted in The Intentional Stance
, 1987, p65)
3. Are persons made of parrots?
Understanding can be understood, not as the turning on of a Cartesian light, but as practical mastery of a certain kind of inferentially articulated doing, responding differentially according to the circumstances of proper application of a concept, and distinguishing the proper inferential consequences of such application. (Brandom, p120)
This is Brandom=s brand of pragmatism
, and I agree that you have to derive representation from inference, not vice versa. This is not recognized by many in philosophy though it has been common understanding in Artificial Intelligence for many years. William Woods= AWhat=s in a link@ (1975) is the locus classicus
. I myself have not always understood the implications of it, and I accept Brandom=s use of me as a bad example of the formalist approach in AIntentional Systems,@ (1971, discussed by Brandom p99ff). Now that I see the work that Amaterial inference@ can do, I am ready to agree wholeheartedly with Brandom about this. I do not agree with him about the origins of norms, however.
Where do the norms come from and why? Let=s consider Brandom=s answer (in a typically hard-to-parse sentence, it must be said).
The objectivity of conceptual norms requires that any attitude of taking, treating, or assessing as correct an application of a concept in forming a belief or making a claim be coherently conceivable as mistaken, because of how things are with the objects the belief or claim is about.@ (p63)
Why? What is it about conceptual norms that imposes this requirement? Do communities just happen to have a correctness fetish (the original political correctness!)? The answer that I would want to give is shunned or ignored by Brandom: the social conceptual norms don=t just happen; they are a requirement for a working system
of communication. We as a species don=t communicate just for the fun of itBthough communication is
a lot of fun, and that, too, is no accident.. Communication has to pay for itself like every other complex adaptation, and this imposes design considerations among which are those that are met by this requirement.
Brandom accepts part of this: Ait is clear that intentionality has a representational dimension and that to understand intentional contentfulness one must understand representation,@ but he shrinks from the idea that Aa suitable notion of representation can be made available in advance of thinking about the correct use of linguistic expressions and the role of intentional states in making behavior intelligible.@ (p60)
It helps understand the order of explanation here if we think of this as an engineering problem. Consider a so-called data-structure in a computer (as set up and maintained by a programBperhaps an airline reservation program). In what (strained) sense could it be about
Chicago? If it serves to maintain, update, and provide information about Chicago to the functions it was designed to serve. It has to work
: the numbers of the flights listed have to be flights into O=Hare Airport, and this fact has to be reliably maintained over time, and the data fields dealing with current weather have to track what the weather is in the Chicago area right now, etc. If it does work, it is about Chicago in the only way that could matter here. Derived intentionality at its best: handsome is as handsome does.2
The fact that these data-structures are typically written in such a way that they readily generate symbol strings that we human beings can read (e.g.
, they link to ASCII characters for Apartly cloudy
, wind SW at ten MPH@ ) is an irrelevant feature. If a data-structure in a robot serves to maintain, update, and provide information about a wall in the robot=s vicinity, and the robot needs to hug the walls to find its way out of a maze and uses the data-structure to guide itself (non-linguistically, with even less sophistication than an insect), then that data-structure is about that wall in a fundamental sense that has nothing to do with linguistic expression.
And even though a robotBor an insectBis not a member of a linguistic community that chastises its members for violations of the community=s norms, its data structures are still in jeopardy of a sort of punishment: extinction for cause if they don=t do what they are supposed to do. The robot=s (intelligent) designer may be the critic who tells herself to go back to the drawing board and come up with a better system of data-structures, or she may have been intelligent enough to design the robot=s information-processes to correct themselves
, extinguishing the versions that don=t work and replacing them with better versionsBshe=s designed a learning robot, in other words. The same goes for the insect, of course. If the insect cannot learn to adjust its behavior when the world departs from the conditions in which its hardware evolved, it will succumb, sphexishly
(Hofstadter, 1982, Hofstadter and Dennett, 1982, Dennett, 1984 ). The dividing line between Alearning@ by a lineage over evolutionary time, and learning, by individual trial-and-error, generate-and-test, is not principled. Some mammals can walk at birthBthey almost literally hit the ground runningBand others need to learn how to control their legs. Evolution is supremely opportunistic in apportioning the R & D process between lineages and individuals.
Does the same verdict apply at the level of community instruction? Here, I think, is where Brandom thinks there is a significant watershed: there is all the difference in the world, he thinks, between norms maintained by communities of norm-protecting language-users and norms maintained by the relatively blind processes of natural selection or operant conditioning and its kin. But here he must confront his own version of the regress problem: even if we assume that normal adults appreciate the moves in the gameBrecognizing the speech acts that honor, flout, exploit, defend, abandon the norms of the linguistic communityBhow did they get themselves into this savvy condition? How, in short, does linguistic correction bootstrap itself into existence? John Haugeland (1998) confronts the same issue in his own version of Pittsburgh normativity, and Asolves@ it by speaking of an innate and pre-linguistic Acensoriousness@ that we humans are endowed with. A gift from God, or did it evolve, and if so, why? See my review
, Dennett (2000).
Why did Awhy@ evolve? That is, why did the communicative community arise, and what sustains it? Brandom=s pragmatist emphasis draws attention to the functional setting, explaining the proximal motivation for score-keepingBthe sanctions enforce the playing of the gameBwhile ignoring the distal motivation: why does the game exist at all? What pays for this elaborate expenditure of energy?
Consider a dog that doesn=t bark: the term Aevolution@ does not appear in the index of Making it Explicit
. This is a measure of how resolutely Brandom has set his face against the >collapse@ that I have proposed. Community
is Brandom=s skyhook (Dennett, 1995) and he can=t have it. He knows this, but he prefers not to dwell on what it would take to secure community as his base of operations. Not surprisingly, this obliges him to reinvent the wheels of others without realizing it. Thus he sees that misrepresentation or falsehood is parasitic on successful (and truthful) representation-- APurporting to represent is intelligible only as purporting to represent successfully
. (p72)Bbut ignores the development of this argument in Dawkins and Krebs (1978) classic work on the evolution of animal communication. Another term missing from the index is AMillikan,@ but Brandom speaks, for instance, of Asome consumer or target@ of a representation, a theme Millikan has developed in depth, treating subpersonal
consumers of (subpersonal) representations. The subpersonal machinery of cognition is another arena that Brandom refuses to enter, though many of the themes he explores are well-explored by Millikan and others working that territory. My point is not a procedural criticismBhe should have cited this workBbut a substantive one: by ignoring it, he creates the illusion (for himself and his readers) that his community-based
account of meaning and representation can be an autonomous alternative
to a naturalistic theory of the same topics. The themes he discusses so insightfully are not just accessible to naturalism; they are (already) explicable, to a significant degree, in naturalistic terms.
I think it is instructive to trace the roughly parallel paths of some of Brandom=s observations and their counterparts in naturalistic accounts. Starting in the physical world (from the physical stance
, in my terms), Brandom notes that Airon doesn=t conceive its world as wet when it responds by rusting.@ (p87) Why not? Because, as he says, >a normative dimension is required, which can underwrite a distinction between correct and incorrect application of concepts.@ He has just quoted Hegel=s wonderful (and as he notes, naturalistic) account of animal desire: Aan animal classifies some particular as food when it >falls to without further ado and eats it up.=@ (p86) but he neglects to note that this Aresponsive disposition@ in the animal, unlike the disposition of the iron, is responsive to a normative dimension. AnimalsBand plants, too, of courseBexploit basic, undesigned physical dispositions like the disposition of iron to rust in the transducers and other hardware that they are endowed with, so that in the main what they eat is edible and nutritious
. These innate endowments provide plants and animals with sources of informationBand, abnormally (in Millikan=s sense), misinformationBwhich subpersonal consumers utilize in the processes of governance that keep life going. But in Brandom=s terms, none of the transactions within the nervous system contain explicit
meanings because although the transactors may be loosely, metaphorically, modeled as homunculi
Bcognitive agents of one sort or anotherBthere isn=t enough of a social
system in place. There may be, as Minsky has put it (1985), a society of mind, but there is no community!
Consider the parafoveal change-detectors or anomaly detectors in the vision system that compete with each other to determine where our eyes dart next. ALook over here! Something interesting is happening in my sector.@ they all seem to shout, and the one that wins succeeds in attracting a saccadeBthe high-resolution foveas of the eyes jump to that point in the visual world for upgraded processing. But these signals are not speech acts; they are neither explicitly imperative (ALook over here!@) nor declarative (ASomething interesting is happening in my sector.@) but rather too primitive to count. One might even say they are degenerate, from Brandom=s point of view, like Wittgenstein=s ASlab!@ language game, which Brandom says is not an instance of a verbal
, but Amerely vocal@, language game (p172). (Cf. also the pioneering analysis of this issue by Jonathan Bennett, in his discussion of bee-dance Acommunication@ in Rationality
In what to me is a key example, Brandom describes a red-measuring instrument hooked up to a tape recorder that appropriately emits AThat=s red@ when a red thing is present. He compares it to a human observer who has been asked (hired?) to do the same thing. There is a world of difference, as he insists, and then he supplies a nice intermediate case: a parrot trained to do the same job. (p88) In spite of being alive, the parrot is more like the instrument, Brandom insists, than the human observer, because the parrot lacks Aa kind of understanding@Bthe kind of understanding Sellars rightly insisted is the key to content. Brandom says:
Insofar as the repeatable response is not, for the parrot, caught up in practical proprieties of inference and justification, and so of the making of further judgments, it is not a conceptual or cognitive matter at all. (p89)
Not at all? This needs to be adjusted. The parrot is some kind of cognitive agent, and even if it doesn=t explicitly engage in inference and justification, its cognitive processes include quality-control measures that track such explicit processes to a significant extent. The contrast Brandom wants to draw between persons and parrots is not as big a gulf as he makes out, but it is real, and it is important precisely because it helps us see a way out of our regress problem: in order for there to be such large agentsBpersonsBcapable of Sellarsian understanding, they need to be composed of smaller, stupider agents with much less understandingBnot so much subpersonal homunculi as psittaculi!
Organisms are correctly seen as multi-cellular communities sharing
, for the most part, a common fate (they=re in the same boat). So evolution can be expected to favor cooperative arrangements in general .Your eyes may, on occasion, deceive youBbut not on purpose! (Sterelny, 2003) Running is sure to be a coordinated activity of the limbs, not a battle for supremacy. Nevertheless, there are bound to be occasions when subsystems work at cross purposes, even in the best-ordered communities of cells, and these will in general be resolved in the slow, old-fashioned way: by the extinction of those lineages in which these conflicts arise most frequently. The result is control systems that get along quite well without
any internal self-monitoring. The ant colony has no boss, and no virtual boss either, and gets along swimmingly with distributed control that so far as we can tell does not engage or need to engage in high level self-monitoring. According to the ethologist and roboticist David McFarland (1989), ACommunication is the only behavior that requires an organism to self-monitor its own control system.@ Organisms can very effectively control themselves by a collection of competing but >myopic= task-controllers that can interrupt each other when their conditions (>hunger= or need, sensed opportunity, built-in priority ranking, . . . ) outweigh the conditions of the currently active task controller. Goals are represented only tacitly, in the feedback loops that guide each task-controller, but without any global or higher-level representation. (One might think of such a task-controller as Auncommented@ codeBit works, but there is nothing anywhere in it that can be read off about what it does or why or how it does it.). Evolution will tend to optimize the interrupt dynamics of these modules, and nobody=s the wiser. That is, there doesn=t have to be anybody home
to be wiser!
But communication, McFarland thinks, is a behavioral innovation that changes that. Communication requires a central clearing house of sorts in order to buffer the organism from revealing too much about its current state
to competitive organisms. In order to understand the evolution of communication
, as Dawkins and Krebs (1978) showed, we need to see it as manipulation
rather than as purely cooperative behavior. The organism that has no poker face, that communicates state directly to all hearers, is a sitting duck, and will soon be extinct. What must evolve instead is a communication-control buffer that creates (1) opportunities for guided deception, and coincidentally (2) opportunities for self-deception (Trivers, 1985), by creating, for the first time in the evolution of nervous systems, explicit (and more Aglobally@ accessible) representations
of its current state, representations that are detachable from the tasks they represent, so that deceptive behaviors can be formulated and controlled.3
This in turn opens up structure that can be utilized in taking the step, described in detail by Gary Drescher (1991), from simple situation-action machines
to choice machines
, the step I describe as the evolutionary transition from Skinnerian to Popperian creatures.(Dennett, 1995.) (The previous two paragraphs are drawn from Dennett, forthcoming)
Supposing that we can analyze whole persons with all their Sellarsian understanding as composed of less comprehending interactors is not just a convenience to the theorist, for a reason that Brandom spells out: individuation by content is inescapably holistic because it involves both premises and conclusions. (p90). We can=t just reach into the snow is white
bin and pull out a snow-is-white
to install in our cognitive system=s belief box! There is no such thing as a snow-is-white
independently of the huge system in which it must reside. So items individuated by content cannot
be building blocks, independently identified. But we do need building blocks, so we can conclude that our blocks must be individuated by something Aless@ than semantic content. We need Janus-faced things that look
, from some angles, rather like propositional contents, and from other angles like scraps of machinery. Bring on the parrottings, understood derivatively as if they were (almost) speech acts. And expect that Brandom=s analyses will prove to be valuable tools for measuring the ways in which subpersonal agencies fall short.
For instance, at what level[s] does deontic status
of one or another attenuated sort arise? What is the difference between the guardian role of the immune system, for instance, and the Aticket taker or doorman@ Brandom usefully imagines? Could there be a sort
of social score-keeping among the semi-independent, cooperating/competing subsystems in the nervous system? Suppose we were to take Brandom=s nicely elaborated model of interpersonal communication and attempt to apply it dead literally to the communication between specialist agencies in the brain. What features would drop out and why? According to Brandom, it is only the Apractical inferential proprieties@ tracked in the score-keeping that Amake noises and marks mean what they mean,@ but aren=t there somewhat attenuatedBde-socialized
one might sayBinferential processes that can make brain-signals mean what they mean? Whatever the answer, it will be a valuable contribution to the theoretical understanding of the ways in which nervous systems work their unmagical magic.
4. Beliefs and opinions: arresting Brandom=s flight from naturalism
I don=t think there has to be a disagreement between Brandom and me at this point. Assuming that his neglect of evolutionary considerations isn=t some sort of closet creationism (not that that is unknown among eminent philosophers dealing with these topics), he ought to be able to take on board most if not all of my friendly amendments. He still gets to maintain the ineliminably social or communitarian grounding of meaning, and thereby explain the second-class or derived application of the constitutive concepts to animals, to human parts, and, of course, to evolution itself. I agree with him about the conceptual dependence of all treatments of subpersonal parrottings as derivative from, dependent on, our prior understanding of personal-level moves in the deontic score-keeping game of inference, and I even agree that ABeliefs are essentially the sort of thing that can be expressed by making an assertion.@ (p153). How can I agree with this? By acceding, in a discussion of his book, to his specified use of the term Abelief@:
There clearly is a sense in which nonlinguistic animals can be said to have beliefs. But the sense of belief that Sellars
, Dummett, and Davidson are interested in (and that is the subject of this work) is one in which beliefs can be attributed only to language users. (p155).
So a belief for Brandom is very much what I call an opinion
(1978): the state of acceptance of a sentence collected (by a linguistic creature, naturally) as true, contrasting such a sophisticated state with the Alower brutish state@ of belief (p305). This makes commentary awkward since I use >belief= and >opinion= as contrasting terms, and Brandom uses >belief= more or less as I use >opinion= and contrasts beliefs so-characterized with Athe intentional states of nonlinguistic creatures,@ a category for which he doesn=t have a shorter term. I=m quite comfortable with Brandom=s insistence that such a language-involving acceptance state is not just the paradigm of a contentful state but that without which we couldn=t really make sense of attributing specific
contents in more attenuated
cases. So I agree that Awhat the frog=s eye tells the frog=s brain@ and what the dog or the chess-playing computer believes
(in my sense) has to be understood, as he insists, as a A
stretching@ usage that takes us out of the home cases into a world of derivative cases. AWhere intentional explanations are offered of the behavior of nonlinguistic creatures (those that are not understood as interpreters of others), the reasons are offered, the assertions are made, by the interpreter of a simple intentional system, who seeks to make its behavior intelligible by treating it as if
it could act according to reasons it offers itself.@ (p171).
Its reasons are what I call Afree-floating rationales@ that it does not represent to itself, and has no need of representing to itself. When, then does the need arise? This is one of the importantBfoundational--questions left unasked by Brandom=s resolutely upward bound account. And in the lacuna thus created he inserts a dubious apologia.
He assumes that Asuitable social creatures can learn to distinguish in their practice between performances that are treated as correct by their fellows (itself a responsive discrimination) and those that are not@ and he is careful to avoid circularity by insisting that Ano appeal will be made to instrumental rationality on the part of fledgling linguistic practitioners@ (p155) In other words, since he cannot help himself to the intentional stance yet, he has to follow a fundamentally Skinnerian line, behaviorism, replacing a rational
account of this learning process with a suitably innocent conditioning
account. And this decision, which he thinks he needs to take to avoid circularity or vicious regress
, creates a problem for him, parallel to the question that has bedeviled Fodor, Dretske and others under the guise of the (ill-posed) disjunction problem
: what can the theorist appeal to in either an individual=s history of learning by trial-and-error or a lineage=s history of evolutionary trial-and-error to distinguish errors from mere extensions of the category being discriminated? (See my AEvolution, Error and Intentionality,@ in The Intentional Stance
, 1987.) Brandom recognizes the problem: AHow is it possible for our use of an expression to confer on it a content that settles that we might all be wrong about how it is correctly used, at least at some times?@ (p137) His version has the advantage that he gets to appeal to community practices and community agreement about how the norms are to be applied, and he sees that his answer, like a good answer to the disjunction problem, has to find, as he says, some way of privileging a dispositionBexcept that he doesn=t actually say how he is going to do this.
He excuses himself from this task by retracting into a very modest goal: >explaining what the trick consists in, what would count as doing it, rather than how it is done by creatures wired up as we are.@ ( p155) In other words, he wants to give the Aspecs@ without a word on the implementation, the (reverse) engineering. This is a remarkable truncation, given not just the ambition but the gratifying success of his own reverse engineering of the game of asking and giving reasons. He blandly assures the reader that Athe abilities attributed to linguistic practitioners are not magical, mysterious or extraordinary@ (p156) but we are entitled to ask an emperor=s new clothes question: Why not?
AThere were no commitments before people started treating each other as committed@Ba phrase of Brandom=s that echoes Hobbes= celebrated attempt to reconstruct the origins of morality out of the state of nature, and Brandom can=t eschew discussing this transition by simply assuring us that it wasn=t magic. For instance, why haven=t other species evolved something like promising, for instance? It=s obviously good for something
. Hobbes presupposed language among his people in the state of nature
, and Brandom=s claim must be that the deontic status of language--with its sanctioned practices and hence implicit concepts of authority, commitment and entitlementBalready contains something with a variety of normativity missing in the state of (nonhuman) nature. (See p175.) Whether the commitments of language are properly the ancestors
of moral commitments or merely had to co-evolve with them is an interesting question, not raised by Brandom.
This is, I think, an unacceptable flight from naturalism, not just because it refuses to address an entirely appropriate question but because by doing so it actually distorts the analysis at the higher level. A deeper account of communication would be a better account. When organisms take on the new behavior of signaling (as contrasted with simply willy-nilly providing information about their state by being perceptible by their audience) the opportunity for strategic deception arises, and plays an important role in fixing precisely the sort of dependencies Brandom himself is arguing for. He asserts, convincingly, that AThese pragmatic inferential practices form a shell around the more basic semantic ones, which they presuppose.@ (159) but both also presuppose the underlying emergence of communication as a behavioral opportunity, with its own costs and benefits. Credible signaling must be hard-to-fake signaling, lest inflation set in and destroy the communication channel with noisy fakery. The panglossian (or polyannian) presumption of cooperation must not be built in.
These are quite uncontroversial themes in the considerable naturalistic literature on the evolution of communication. Adding them to his perspective would not commit Brandom to any doctrinaire scientism
that might seem to threaten the philosophical austerity of his project, and it would go some way to removing the suspicious residue of magic that clings to any account that purports to ground meaning in community and then, having done that, declares victory.4
Bennett, Jonathan, 1964, Rationality, London: Routledge & Kegan Paul.
Dawkins, R., and Krebs, J. R., 1978, AAnimal Signals: Information or manipulation?@ in J. R.
Krebs and N. B. Davies, eds., Behavioral Ecology An Evolutionary Approach, Oxford: Blackwell, pp282-309.
Dennett, Daniel C., 1978, Brainstorms. Cambridge, MA: MIT Press.
------------1981, AThree Kinds of Intentional Psychology,@ in R. Healey, ed., Reduction, Time and Reality, Cambridge: Cambridge Univ. Press, pp.37-61, reprinted in Dennett, 1987.
----------- 1984, Elbow Room: The Varieties of Free Will Worth Wanting., Cambridge, MA: MIT Press.
------------1987, The Intentional Stance, Cambridge, MA: MIT Press.
------------ 1999, Review of John Haugeland: Having Thought: Essays in the Metaphysics of Mind, in Journal of Philosophy, Volume 96I, pp. 430 435.
------------- forthcoming, AMy body has a mind of its own,@ in Ross, D., ed. Distributed Cognition and the Will. MIT Press
Drescher, Gary, 1991, Made-Up Minds: A Constructivist Approach to Artificial Intelligence, MIT Press.
Haugeland, John, 1998, Having Thought: Essays in the Metaphysics of Mind, Cambridge, MA: Harvard Univ. Press
Hofstadter, D. R. (1982). ACan creativity be mechanized?@ Scientific American, 247, 20 29.
Hofstadter and Dennett, 1982, The Mind=s I, New York: Basic Books.
McFarland, David, 1989, AGoals, No-Goals and Own Goals,@ in Alan Montefiore and Denis Noble, eds., Goals, No-Goals and Own Goals: A Debate on Goal-Directed and Intentional Behaviour. London: Unwin Hyman Ltd.
Millikan, Ruth, 2000, "Naturalizing Intentionality,"n Philosophy of Mind, Proceedings of the Twentieth World Congress of Philosophy Volume 9, Bernard Elevitch, Editor, Philosophy Documentation Center, 2000, pp.83-90.
Minsky, Marvin, 1985, The Society of Mind
, New York: Simon & Schuster.
Sterelny, Kim, 2003, Thought in a Hostile World: The Evolution of Human Cognition
, Oxford: Blackwell.
Trivers, Robert, 1985, Social Evolution
Woods, William, 1975, "What's in a Link: Foundations for Semantic Networks". In D. Bobrow and A. Collins (eds.), Representation and Understanding: Studies in Cognitive Science
, New York: Academic Press. Reprinted in R. Brachman and H. Levesque (eds.), Readings in Knowledge Representation
, San Mateo: Morgan Kaufmann
, 1985. Also reprinted in Allan Collins and Edward E. Smith (eds.), Readings in Cognitive Science
, San Mateo: Morgan Kaufmann, 1988.