Title: The statistics of experience in learned and evolved behavioral plasticity.
Author: David W. Stephens
Affiliation: University of Minnesota (Deparment of Ecology, Evolution and Behavior, College of Biological Science)
Web:
Experience affects behavior via a wide variety of mechanisms and at many biologically significant time scales. This presentation will develop a simple model that asks when it pays to attend to experience. It will review data from my laboratory that supports this model at two different time scales. Our model, which we call the flag model, considers the role to two statistical variables that, in theory, make experience valuable: uncertainty and reliability. Broadly speaking the model claims that it pays to attend to experience when uncertainty & reliability are high. We have tested this model at both behavioral and evolutionary time scales. We have shown for example that our model correctly predicts when captive blue jays will learn to use an artificial signal. Studies using the techniques of experimental evolution show that our model also correctly predicts the action of natural selection on behavioral plasticity. In addition, we can use the model to ask which type of experience an animal should attend to. Using this approach, my laboratory has developed the first experimental study of the evolution of prepared learning. I will discuss extensions and limitations of the model, and its relationship to other approaches to the problem of behavioral plasticity.
Title:
Rationality and decision making: Linking ecology, foraging and choice
Author: Marco Vasconcelos
Affiliations: University of Minho and University of Oxford (Behavioral Ecology Research Group, University of Oxford and Animal Learning and Behavior Lab, Universidade do Minho-Portugal)
Web:
http://users.ox.ac.uk/~kgroup/people/alexkacelnik.shtml http://escola.psi.uminho.pt/unidades/lca
Rationality principles are the bedrock of normative theories of decision making in biology and microeconomics, but while in microeconomics consistent choice underlies the notion of utility, in biology the assumption of consistent selective pressures justifies modelling decision mechanisms as if they were designed to maximise fitness. In either case, violations of such principles, defined variously as failure to maximise some substantial benefit or as showing inconsistent preferences, populate a growing catalogue of putative 'cognitive biases' in humans and other animals. This literature's main theoretical unity is the violation of predictions from normative models, resulting in such examples often being invoked to object to the relevance of optimality modelling of behaviour. Based on a large body of experimental evidence, I propose an alternative approach, in which paradoxical observations in experimental situations serve to reflect on selective pressures in ecological circumstances, and information-processing mechanisms are included in the development of normative models capable of novel predictions. The contention is that learning mechanisms can (and should) be used to develop better normative models and that blending classical optimal foraging and contemporary learning theories helps unravelling behavioural mechanisms and the causes of sub-optimal behaviour.
Title: Aprendizaje discriminativo de estimulos visuales complejos en ratones adultos (Visual discrimination learning of complex visual stimuli in adult mice)
Author: Mario Trevino Villegas
Affiliation: Universidad de Guadalajara (Instituto de Neurociencias)
Web:
http://www.ineuro.cucba.udg.mx/articulo.php?id=175
The mouse is receiving growing interest as a model organism for studying visual perception. However, little is known about how discrimination and learning interact to produce visual conditioned responses. Here, we adapted a two-alternative forced-choice visual discrimination task for mice and examined how training with equiprobable stimuli of varying similarity influenced conditioned response and discrimination performance as a function of learning. Our results indicate that the slope of the gradients in similarity during training determined the learning rate, the maximum performance and the threshold for successful discrimination. We also found that the amount of trials that the mice made lateral choices increased with stimulus similarity and also occurred in conditions of high discriminability. Interestingly, while lateralization occurred at the individual level, it was absent, on average, at the population level. Biased choice sequences obeyed the generalized matching law and increased task efficiency when stimulus similarity was high. Furthermore, a mathematical analysis revealed that strongly-biased mice used information from past rewards but not past choices to make their current choices. Finally, we found that the strength of the side-bias measured during the first day of training predicted individual differences in the average learning behavior. This framework provides useful analysis tools to study individualized visual-learning trajectories in mice and introduces a new training protocol with quantitative measures to study perceptual learning and visually-guided behavior in freely moving mice.