The Cybernetic Brain
Page 17
Ashby's strained optimism of 1957 was misplaced. A year later, on 29 September 1958, we find him writing (pp. 6058–60): "The new DAMS . . . having fizzled out, a new idea occurs to me today—why not make a small DAMS, not for experimental purposes but purely for demonstration. . . . The basic conception is that all proofs are elsewhere, in print probably; the machine is intended purely to enable the by-stander to see what the print means & to get some intuitive, physical, material feeling for what it is about. (Its chief virtue will be that it will teach me, by letting me see something actually do the things I think about.) Summary: Build devices for demonstration." The drift in this passage from DAMS to demonstration machines is significant. After a break, the same journal entry continues poignantly: "The atmosphere at Namur (Internatl. Assoc. for Cybs., 2–9 Sep.) showed me that I am now regarded more as a teacher than as a research worker. The world wants to hear what I have found out, & is little interested in future developments. Demonstration should therefore be my line, rather than exploration. In this connexion it occurs to me that building many small machines, each to show just one point, may be easier (being reducible) than building a single machine that includes the lot. Summary: Build small specialist machines, each devised to show one fact with perfect clarity." A formally beautiful but personally rather sad technosocial adjustment is adumbrated in this note. In it, Ashby responds to two or possibly three resistances that he felt had arisen in his research. The one that he failed to mention must have been his lack of technical success in developing DAMS as a synthetic brain. The second was the escalating cost and lack of commensurate institutional support for developing DAMS, as just discussed. And the third was what he perceived, at least, to be a developing lack of interest in his research in the European cybernetics community. How far he was correct in this perception is difficult to judge; it is certainly true, however, that youngsters like Stafford Beer and Gordon Pask were bursting onto the scene by the late 1950s—Beer was thirty-four in 1958, Pask thirty-two; Ashby was becoming a grand old man of cybernetics at the age of fifty-four. And all of these resistances were accommodated by Ashby's strategy. Technically, building small demonstration machines presented him with a finite task (unlike the never-ending difficulties with DAMS as a research machine), reduced the cost to a bearable level, and, socially, positioned Ashby as a pedagogue.
In important respects, Ashby went through with this plan. Especially at the University of Illinois in the 1960s, his demonstration machines became legendary, as did his qualities as a pedagogue.37 It is certainly not the case that he gave up his research after 1958—his "hobby" was always his raison d'être—but his major subsequent contributions to cybernetics and systems theory were all in the realm of theory, as foreshadowed in the first quotation above. As a full professor at a major American university, Ashby's funding problems appear to have been significantly alleviated in the 1960s, and there is one indication that he returned then to some version of DAMS as a research project. In an obituary, Oliver Wells recalled that Ashby's "love of models persuaded von Foerster to have constructed what was called the 'The Grandfather Clock' which was designed as a seven foot noisy model of state-determined complex 'systems' running through trajectories of cycles of stabilisation and 'randomness' " (Wells 1973). One has to assume that nothing significant emerged from this project; like the English DAMS, it was never the subject of anything that Ashby published.
The stars were in a strange alignment for Ashby in the late 1950s. Immediately after the deflationary post-Namur note he added an interstitial, undated note which reads: "Here came the Great Translation, from a person at B. H. to Director at B. N. I. [the Burden] (Appointment, but no more till May '59)" (p. 6060). But now we, too, can take a break and go back to madness.
Madness Revisited
At the beginning of this chapter I noted that Ashby's career in Britain was based in mental institutions and that he was indeed active in research related to his profession, publishing many papers on explicitly psychiatric topics. I want now to discuss the relation between the two branches of Ashby's work, the one addressed to questions of mental illness and the cybernetic work discussed in the preceding sections.
My starting point is Ashby's 1951 assertion, already quoted, that his cybernetics, as developed in his journal, "was to me merely a delightful amusement, a hobby I could retreat to, a world where I could weave complex and delightful patterns of pure thought." This assertion deserves to be taken seriously, and it is tempting to read it as saying that his cybernetic hobby had nothing to do with his professional research on pathological brains and ECT. It is also possible to read his major works in cybernetics, above all his two books, as exemplifications of this: there is remarkably little of direct psychiatric interest in them. The preceding discussions of the homeostat and DAMS should likewise make clear that this aspect of Ashby's work had its own dynamic. I nevertheless want to suggest that this reading is untenable, and that there were in fact interesting and constitutive relationships between the two branches of Ashby's oeuvre—that psychiatry was a surface of emergence and return for Ashby's cybernetics, as it was for Walter's.
We can start by noting that in the 1920s Englishmen took up many hobbies, and theorizing the adaptive brain is hardly the first that comes to mind. If in 1928 Ashby had taken up stamp collecting, there would be nothing more to say. But it is evident that his professional interests structured his choice of hobby. If his cybernetics, as discussed so far, was an attempt to understand the go of the normal brain, then this related to his professional concerns with mental illness, at minimum, as a direct negation rather than a random escape route. More positively, Ashby's materialism in psychiatry, shared with Golla and Walter, carried over without negation into his hobby. The hobby and the professional work were in exactly the same space in this respect. And we should also remember that in medicine the normal and the pathological are two sides of the same coin. The pathological is the normal somehow gone out of whack, and thus, one way to theorize the pathological is first to theorize the normal. The correlate of Ashby's interest in adaptation, in this respect, is the idea going back at least to the early twentieth century, that mental illnesses can be a sign of maladaptation (Pressman 1998). Simply by virtue of this reciprocal implication of the normal and the pathological, adaption and maladaptation, it would have been hard for Ashby to keep the two branches of his research separate, and he did not.
The most obvious link between the two branches of Ashby's research is that most of Ashby's early cybernetic publications indeed appeared in psychiatric journals, often the leading British journal, the Journal of Mental Science. And, as one should expect, all of these papers gestured in one way or another to the problems of mental illness. Sometimes these gestures were largely rhetorical. Ashby would begin a paper by noting that mental problems were problems of maladaptation, from which it followed that we needed to understand adaptation, which would lead straight into a discussion of tilted cubes, chicken incubators, beads and elastic, or whatever. But sometimes the connections to psychiatry were substantial. Even Ashby's first cybernetic publication, the 1940 essay on dynamic equilibrium, moves in that direction. Ashby there discusses the "capsule" which controls the fuel flow in a chicken incubator and then asks what would happen if we added another feedback circuit to control the diameter of the capsule. Clearly, the capsule would not be able to do its job as well as before, and the temperature swings would be wilder. Although Ashby does not explicitly make the point, this argument about "stabilizing the stabilizer" is of a piece with the conventional psychiatric idea that some mental fixity lies behind the odd behavior of the mentally ill—mood swings, for example. What Ashby adds to this is a mechanical model of the go of it. This simple model of the adaptive brain can thus be seen as at once a model for thinking about pathology, too. Likewise, it is hard not to relate Ashby's later thoughts on the density of connections between homeostat units, and their time to reach equilibrium, with lobotomy. Perhaps the density of neural interconnections c
an somehow grow so large that individuals can never come into equilibrium with their surroundings, so severing a few connections surgically might enable them to function better. Again, Ashby's understanding of the normal brain immediately suggests an interpretation of mental pathology and, in this case, a therapeutic response.
Ashby often failed to drive home these points explicitly in print, but that proves very little. He contributed, for example, the entry "Cybernetics" to the first Recent Progress in Psychiatry to appear in Britain after World War II (Fleming 1950).38 There he focused on pathological positive feedback in complex machines—"runaway"—as a model for mental illness, leading up to a lengthy discussion of the stock ways of curing such machine conditions: "to switch the whole machine off and start again," "to switch out some abnormal part," and "to put into the machine a brief but maximal electric impulse" (Ashby 1950b, 107). We saw this list before in the previous chapter, and when Walter produced it he was not shy of spelling out the equivalences to sleep therapy, lobotomy, and ECT, respectively. Given a pulpit to preach to the psychiatric profession, Ashby could bring himself to say only, "These methods of treatment [of machines] have analogies with psychiatric methods too obvious to need description" (1950b, 107).
To find more specific and explicit connections between Ashby's cybernetics and his professional science, it is interesting to begin with a paper I mentioned before, his 1953 essay "The Mode of Action of Electro-convulsive Therapy" (Ashby 1953a). As I said, the body of this paper is devoted to reporting biochemical observations on rats that had beeen subjected to electroshock, and the theoretical introduction accordingly lays out a framework for thinking about ECT and brain chemistry. But Ashby also throws in a second possible interpretation of the action of ECT:
There is a possibility that E. C. T. may have a direct effect on the cortical machinery, not in its biochemical but in its cybernetic components. . . . It has been shown [in Design for a Brain]that one property such systems [of many interacting elements] will tend to show is that their responses . . . will tend to diminish. When the stimulus is repeated monotonously, the phenomenon is well known under the name of "habituation." We can also recognise, in everyday experience, a tendency for what is at first interesting and evocative to become later boring and uninspiring. Whether the extreme unresponsiveness of melancholia is really an exaggeration of this process is unknown, but the possibility deserves consideration. What makes the possibility specially interesting is that the theory of such statistical systems makes it quite clear that any complex network that has progressed to a non-responding state can, in general, be made responsive again by administering to it any large and random disturbance. The theory also makes clear that such a disturbance will necessarily disturb severely the system's memory: the parallel with E. C. T.'s effect on memory is obvious. Whether, however, E. C. T. acts in essentially this way is a question for the future.
This passage is remarkable in at least two ways. First, it does not belong in Ashby's essay at all. If taken seriously, it undercuts the entire rationale for the biochemical investigations reported there. Second, and more important in the present context, it makes an explicit connection between Ashby's cybernetics and his work on DAMS on the one hand, and his interest in ECT and its functioning on the other, and we can return to DAMS here.39 A journal entry of 25 August 1951 records that "while working with DAMS I found I was unconsciously expecting it to 'run down,' then I realised what was happening, & that my expectation was not unreasonable, was a new idea in fact." Then follows the first discussion of "habituation" in DAMS (though Ashby does not use the word here): "there is therefore a tendency for the neons to change their average 'readiness' from 'more' to 'less.' " And Ashby immediately moves from this observation to a consideration of the antidotes to habituation: "After this initial reserve of changeable neons has been used up the system's possibilities are more restricted. The only way to restore the possibilities is to switch the set off, or perhaps to put in some other change quite different from those used during the routine. This fact can obviously be generalised to a principle." As just mentioned, there was a stock equation in the cybernetics of this period between switching off a machine and sleep therapy for mental illness, though Ashby does not comment on this in his note. However, there then follows a quick sketch of the argument that in its response to a new and different input, DAMS will regain its prehabituation sensitivity to the old one, after which Ashby concludes: "Summary: A multistable system tends to lose reactivity, which will often be restored by applying some strong, but unrelated stimulus, at the cost of some forgetting. ? Action of E. C. T. (Corollary p. 3464)" (pp. 3434–3437).
This is the argument Ashby relied upon above but did not provide in his 1953 essay on the functioning of ECT, but here we find it right in the heartland of his hobby, engaging directly with his major cybernetic project of the early 1950s, DAMS. And it is revealing to follow this story a little further in his journal. The reference forward from the last note takes us to a journal entry dated 12 September 1951, which begins, "From p. 3464, it is now obvious how we make DAMS neurotic: we simply arrange the envt. so that it affects two (or more) essl. variables so that it is impossible that both should be satisfied." Page 3464 in fact takes us to a discussion of Clausewitz, which I will come back to in the next section. In this entry, though, Ashby draws a simple circuit diagram for DAMS as subject to the conflicting demands of adapting to two different voltages at once (fig. 4.10) and comments that "both E.V.'s will now become very noisy," seeking first to adapt to one voltage and then the other, "and the system will be seriously upset. It is now very like a Masserman cat that must either starve or get a blast in the face. The theme should be easily developed in many ways" (pp. 3462–63). We thus find ourselves explicitly back in the psychiatric territory I associated in the previous chapter with Grey Walter and the CORA-equipped tortoise, now with DAMS as a model of neurosis as well as normality and of the functioning of ECT.40
Ashby's journal entry refers forward to another dated 22 September 1951, where Ashby remarks that DAMS will simply hunt around forever when posed an insoluble problem, but that "the animal, however, . . . will obviously have some inborn reflex, or perhaps several, for adding to its resources. . . . A snail or tortoise may withdraw into its shell. . . . The dog may perhaps simply bite savagely. . . . A mere total muscular effort—an epileptic fit—may be the last resort of some species. . . . My chief point is that the symptoms of the unsolvable problem, whether of aggression, of apathy, of catatonia, of epilepsy, etc are likely to be of little interest in their details, their chief importance clinically being simply as indicators that an unsolvable problem has been set" (pp. 3479–81). Here Ashby covers all the bases, at once addressing a whole range of pathological clinical conditions, while dismissing the importance of symptoms in favor of his cybernetic analysis of the underlying cause of all of them— and, in the process, perhaps putting down Grey Walter, for whom epilepsy—"a mere total muscular effort"—was a major research field in its own right.
Habituation and dehabituation, then, were one link between Ashby's cybernetics and his psychiatry, and, indeed, it is tempting to think that the possibility of this link explains some of the energy Ashby invested during the 1950s in this otherwise hardly exciting topic. But it is worth emphasizing that it was by no means the only possible link that Ashby discerned. To get at the range of his thinking it is enough to look at his published record, and here we can focus on a 1954 paper, "The Application of Cybernetics to Psychiatry" (Ashby 1954).41 This tentatively outlines several different ways of thinking cybernetically about mental illness. I will just discuss a couple.42
Figure 4.10."How DAMS can be made neurotic." Source: Ashby's journal, entry dated 12 September 1951 (p. 3463). (By permission of Jill Ashby, Sally Bannister, and Ruth Pettit.)
One carried further Ashby's theorizing of the chemistry of electroshock. As mentioned at the beginning of this chapter, Ashby's own measurements had shown, he believed, that electroshock was followed by
"a brisk outpouring of steroids." Here the question addressed was this: The level of steroids in the brain is presumably a quantity which varies continuously, up or down. Insanity, in contrast, appears to be dichotomous—one is either mad or not. How then can a continuous cause give rise to a discontinuous effect? "What is not always appreciated is that the conditions under which instability appears are often sharply bounded and critical even in a system in which every part varies continuously. . . Every dynamic system is potentially explosive.. . . These facts are true universally. . . . They are necessarily true of the brain" (1954, 115–16). And Ashby had, in fact, addressed this topic mathematically in a 1947 paper (Ashby 1947). There he considered a complex system consisting of interlinked autocatalytic chemical reactions of three substances, with rates assumed to be controlled by the presence of some enzyme, and he showed by numerical computation that there was an important threshold in enzyme concentration. Below that threshold, the concentration of one of the reacting chemicals would inevitably fall to zero; above the threshold, the concentration would rise to unity. This mathematical result, then, showed in general how discontinuous effects can emerge from continuous causes, and, more specifically, it shed more light on the possible go of ECT—how the outpouring of steroids might conceivably flip the patient's brain into a nonpathological state.43