Together with Ezequiel Di Paolo I presented this poster at ESCOP2011, we present some old and new preliminary results on evolutionary robotics to ground a richer notion of habit than the one currently used in cognitive psychology and neuroscience. Click on the image below to download a PDF version of the poster.
I just got back from the workshop «Systematicity and the post-connectionist era«, where I presented a talk entitled «From systematicity of thought to systemicity of habits«. Congratulations to the organizers for this extraordinary experience.
My talk started by assuming the real challenge of systematicity for dynamical approaches. The work of René Thom and Jean Petitot on morphodynamics and cognitive grammars serves as a powerful framework to solve this problem. In the second part of the talk I defend a contemporary re-appraisal of the notion of habit within the Piagetian framework, with illustrations from evolutionary robotics. You can download the pdf of the slides bellow:
Together with Ezequiel Di Paolo we embarked into a historical research on the notion of Habits as theoretical building blocks for cognitive science. Far from the simplified stimulus-response pairing conception of habits defended by behaviorism, we found that habits have long been a very rich conceptual category at the root of the sciences and philosophies of mind, until very recently. Here is a preliminary graph that summarizes some of our results (that would hopefully be published as a paper some time soon):
Phylogeny of the concept of Habit. The graph is still under development but captures the most important trends.
I am presenting this poster at the CogSys2010 Conference in Zurich. The poster shows some results of a simulation model I developed long ago (almost 5 years ago!) on a reinforcement learning task with evolutionary robotics, using homeostatic plasticity on the CTRNN controllers. I am currently writing a paper with Ezequiel Di Paolo on the details of the simulation model. The agents are capable to display a developmental process (connection weights are initialized at value 0), to solve an operant conditioning task and display the spontaneous formation of habits.
Xabier E. Barandiaran 