Seminar #45: Cognitive Surfing – Exploiting coupling to hostile environments
*** Friday Jan 16th, 16:30-18:00, Room Pev1 1A01 ***
NOTICE SPECIAL DAY AND LOCATION
The first Alergic and Life and Mind seminar of the Spring term will be a combined talk by Dr Manuel Bedia (Computer Science Dept. University of Zaragoza, Spain and Visiting Researcher, Department of Informatics, University of Sussex).
He will be speaking on:
Cognitive Surfing: Exploiting coupling to hostile environments.
Traditional perspectives in cognitive science put all the weight of performance on mechanisms internal to an agent who must attempt to find a viable solution to a problem. Environmental noise and uncertainties play a negative role in this view.
In this talk, I will discuss three simple, but general examples of coupled agent-environment systems using the framework of dynamical systems theory in order to question these intuitions. In all cases, agents try to achieve their goals using strategies that exploit local conditions in their interactions with the environment, but not maximizing the quality or efficiency of the solutions.
1) In the first case (modeled as an ordinary differential equations system) an agent searches solutions with maximum versatility (high probability to find alternative nearby solutions). We show that strategies keeping the maximum number of options open, optimize the coupled system when the environment changes in an unpredictable way.
2) In the second case (modeled as a finite-difference system) an agent pursues strategies that minimize the time to exchange the current solution by another one. We demonstrate that, when it is known how the external situation changes the internal state of the agent, more flexible strategies (even when more poorly adjusted to the world) optimize the coupled system.
3) In the third case (a discrete-time stochastic differential system) an agent combines its strategy with the dynamics of the (hostile) environment: the effect of the mutual interaction results in a successful cooperation improving the agent’s behavior. If we understand the system as a coupled game, we have two losing games that become winning when combined. This counterintuitive result points out how the nature of stochasticity can induce regularity in non-linear coupled systems. We demonstrate that a combination of two negative strategies can give rise to a positive result (i.e., bad decisions getting unwanted states, could have good outcomes).
Finally, and based on the results above, we elaborate a new conceptual framework for studying cognitive phenomena. More specifically, we advance a new hypothesis that could explain some functional aspects of emotions in an easier way than the somatic marker theory proposed by Antonio Damasio does. Damasio proposes a model of how the brain organizes the interaction between emotional experiences (stored as somatic markers) and decision-making mechanisms, improving our capacity to make good decisions if somatic markers are adequately implemented (“good” marker if the experience was “good” and vice versa). We argue that this hypothesis could be invalidated if storing bad experiences with good somatic markers does not influence the result of the system. The possibility is discussed based on the results of the models.