Second-Order Cybernetics Maintaining Essential Variables Is Not Passive Matthew Egbert « 8 » Proponents of the enactive ap- sible failure to do so will necessarily result “ the survival standard itself is inadequate for proach agree that the homeostat’s double in the homeostat’s death” (§10). However, at the evaluation of life. If mere assurance of per- feedback architecture made an important the level of its physical body, an organism is manence were the point that mattered, life should contribution, but at the same time they always far-from-equilibrium with respect to not have started out in the first place. It is essen- struggle to overcome its limitations (Ike- its environment. Falling into an equilibrium tially precarious and corruptible being, an ad- gami & Suzuki 2008). Franchi briefly refers is the same as dying, because the organism venture in mortality, and in no possible form as to evolutionary robotics models inspired by would lose its ability to do the work of self- assured of enduring as an inorganic body can be. this ultrastability mechanism, but he does producing its own material identity, i.e., the Not duration as such, but ‘duration of what?’ is the not mention that further progress has been very process that ensures that the double question.” (Jonas 2001: 106) difficult. Ezequiel Di Paolo (2003) showed feedback loop between behavior and inter- that implementing Ashby’s mechanism is a nal homeostasis is intrinsically connected Tom Froese graduated with a D.Phil. in cognitive science significant step toward more organism-like within a whole. Only non-living matter can from the Univ. of Sussex. He is currently a postdoctoral robots, but the contingent link between (in- be in physical equilibrium with its environ- researcher at the National Autonomous University of ternal) homeostatic mechanisms and (ex- ment. Franchi conflates the living with the Mexico. He investigates the relationship between life, ternal) behavior is a continuing source of non-living in another way when he notes mind and sociality using a variety of methods, including problems. that the homeostat “will continue acting the evolutionary robotics, systems theory and phenomenology. « 9 » Switching the connections of an way it normally does until an outside force ultrastable system may in some cases lead compels it to change course of action” (§26). Received: 2 October 2013 to an adaptive re-organization of original This is precisely the way in which Newton’s Accepted: 18 October 2013 behavior, as Franchi’s model of two self-con- first law of motion describes the behavior of nected homeostat units shows, but it is far objects. However, living beings – subjects from guaranteed in more complex systems. – can change their behavior even in the ab- It is possible that the switch is not sufficient sence of a change in external conditions. For Biological Systems, to “break” the essential variables, thereby Maintaining Essential failing to provoke an adaptive response at Overcoming the autonomy/ all, or, if adaptation does occur, the new heteronomy dichotomy Variables Within Viability equilibrium may not reestablish the origi- « 12 » Franchi enticingly hints at the Limits Is Not Passive nal behavior. From Ashby’s perspective, this possibility of a “reevaluation of the tradi- failure to recover the original behavior is not tional distinction between heteronomous Matthew Egbert a problem, since he simply equates all equi- and autonomous behavior” (Abstract). Is University of Hertfordshire, UK libriums with survival – without asking how perhaps the relational view of life defended they might differ in terms of desirability (an by the enactive approach (e.g., Di Paolo mde/at/matthewegbert.com idea that is later echoed by Maturana’s claim 2009) such a reevaluation? As I see it, there that autopoiesis does not admit gradations). is a natural development of ideas from first- > Upshot •The target article proposes « 10 » However, this neutrality is not order cybernetics via second-order cyber- that Ashby’s investigations of the ho- found in nature. For example, inverted netics toward enactive cognitive science. meostat and ultrastability lead to a view goggle experiments can be safely conducted The early approach was too heteronomous, of living systems as heteronomous, pas- 109 without any irreversible consequences to the while the second-order approach was too sive “sleeping” machines and thus are participants. How then is the appropriate re- autonomous, but the relational stance of in fundamental conflict with concepts configuration achieved? There have been a the enactive approach formulates a dialectic of autonomy developed by Jonas, Varela couple of attempts at addressing the practi- between these two extremes (Froese 2011; and others. I disagree, arguing that (1) cal problems stemming from contingent de- Froese & Stewart 2013). Relatedly, it aims the maintenance of essential variables pendencies with improved evolutionary ro- for a middle ground between the extreme of within viability limits is not a passive botics models, in particular by more closely computationalism, which explicitly repre- process for living systems and (2) the tying the desired behavior to the satisfaction sents goals, and the extreme of Ashby’s (and purpose of Ashby’s investigations of the of the homeostatic conditions (Iizuka & Di Maturana’s) insistence on absolute non-con- homeostat was to investigate adaptiv- Paolo 2008; Iizuka et al. 2013). However, tingency. Crucial for these two theoretical ity, a subject that is related to, but clearly these attempts have met with only partial shifts is to go beyond traditional theories distinct from, autonomy. As such, I find success. that portray life’s ultimate purpose, no mat- Ashby’s work on adaptivity to be neither « 11 » We need to move beyond the tra- ter whether by active or passive means, as in opposition to nor in direct support of ditional fixation on equilibrium dynamics in equilibrium, stability, or survival. As Hans modern concepts of biological autono- order to make real progress on these issues. Jonas recognized, only when we face up to my and suggest that a productive way According to Franchi, an Ashbyan organ- the essential precariousness of living exis- forward involves the investigation of the ism is continuously required to behave so as tence can we hope to understand the mean- intersection between these two funda- to go to equilibrium, since “its always pos- ing of life: mental properties of living systems. http://www.univie.ac.at/constructivism/journal/9/1/093.franchi « 1 » Stefano Franchi distinguishes be- the homeostat can generate but, other than « 6 » Nowhere is the necessity of activ- tween models of objects, which are used to that, he seems to agree: ity for the maintenance of essential vari- evaluate the accuracy of our understanding ables more evident than in living systems. of real-world systems, and models of con- “ the main source of resistance to Ashby’s the- All organisms are far-from-equilibrium cepts, which are used to refine, explore or sis about life lies […] in its effort to derive action dissipative structures that require ongo- explain theoretical constructs. He observes, from non-action [and this] means something ing acquisition of energetic and material rightly, that the two approaches are not in- philosophically far more disturbing: the homeo- resources to counteract their degradation compatible and can in fact be complemen- stat will act as much as is needed in order to go (Gánti 2003; Schrödinger 1944). If we think tary, provided that they are not conflated.2 back to sleep. It will go to extraordinary lengths, about archetypal essential variables, such as I agree with Franchi that the homeostat is it will take whatever action it can – just in order body temperature or the levels of sugar and most productive when it is considered as to go back to rest (i.e., to equilibrium).” (§10, oxygen in the blood, it is clear that these all a model of a concept rather than a model emphasis in original) require ongoing organismic activity if they ICS of the brain. However, I may disagree with are to be maintained within viability limits. Franchi over what concept the homeostat is « 4 » Here, Franchi is suggesting that To assume the maintenance of these within NET R intended to be a model of. because the homeostat “derives action from viability limits as being passive is a mistake. E B « 2 » I see the homeostat as a system non-action,” it is passive; but there is some « 7 » This leads to why I disagree with CY for investigating adaptivity. “How does the confusing language used here that I believe the proposition in §§13–14 that Ashby’s R brain produce adaptive behaviour?” is the leads to an incorrect view of the homeostat work opposes the relatively recent frame- DE opening sentence of Ross Ashby’s Design as passive. In particular, there is a confla- works developed for studying autonomy. R for a Brain (1960), and, in my view, the goal tion with the idea of (1) a variable being at a Ashby developed a framework for defining, ND-O of the book. Franchi appears to have a dif- steady state (a concept that is more general describing and studying adaptive behaviour. ferent interpretation, seeing the homeostat than that of equilibrium and that allows for Hans Jonas, and later Francisco Varela and CO as targeting a broad view of life: “Ashby’s stability, i.e., an absence of change, at con- Humberto Maturana, developed conceptual N SE device is a proxy for a view of life as gen- ditions that are far away from equilibrium) frameworks for understanding how a system I S eralized homeostasis, whose principles it and (2) the system being inactive, passive, can be autonomous, how its own needs can T P embodies” (§11). I will argue below that or “asleep.” The homeostat is capable of the emerge and how these can be satisfied by the E C if the homeostat is seen as targeting “just” former, but is described in the quote above system itself.