Ashbian Homeostasis as Non-autonomous Adaptation
Stefano Franchi · Self-Adaptive and Self-Organizing Systems · 2015
The paper argues that the approach to adaptation developed by W.R. Ashby in Design for a Brain and partially revived in evolutionary robotics is orthogonal to the traditional distinction between autonomy and heteronomy that underlies much recent work in cellular biology, evolutionary robotics, ALife, and general AI. Ashby's theory of adaptation claims that homeostasis is the only principle grounding sub-cognitive as well as all cognitive functions. Ashby's generalized homeostasis thesis does not fit the traditional conception of autonomy nor its converse (heteronomy). The traditional conception generates a dualism opposing human autonomous life to its non-human heteronomous counterpart. The dualism is usually overcome by claiming that human life is actually heteronomous although it does not appear so, as the masters of suspicion once claimed (Freud, Marx, and Nietzsche), or, conversely, by claiming that non-human life is actually fully autonomous in spite of its outlook (as suggested by Hans Jonas and his enactivist followers). Ashby's theory offers a genuine alternative by rejecting the very opposition of autonomy vs. Heteronomy that both strategies rely on. The joint effect of Ashby's unorthodox philosophical stance and the practical shortcomings of the simulating technology at his disposal has prevented a true evaluation of his theory. We can assess the philosophical and technical viability of the general homeostasis thesis Ashby advocated through the construction of virtual cognitive agents (i.e. Simulated robots in a physically plausible environment) that provide an integral and faithful digital replica of the architecture of Ashby's original homeostat that reproduces the homeostat's dynamics and its stochastic rewriting mechanism through a Ctrnn-like network architecture, whose outline implementation is then discussed.