Online research in philosophy

Recent theoretical work has identified a tightly-constrained sense in which genes carry representational content. Representational properties of the genome are founded in the transmission of DNA over phylogenetic time and its role in natural selection. However, genetic representation is not just relevant to questions of selection and evolution. This paper goes beyond existing treatments and argues for the heterodox view that information generated by a process of selection over phylogenetic time can be read in ontogenetic time, in the course of individual development. Recent results in evolutionary biology, drawn both from modelling work, and from experimental and observational data, support a role for genetic representation in explaining individual ontogeny: both genetic representations and environmental information are read by the mechanisms of development, in an individual, so as to lead to adaptive phenotypes. Furthermore, in some cases there appears to have been selection between individuals that rely to different degrees on the two sources of information. Thus, the theory of representation in inheritance systems like the genome is much more than just a coherent reconstruction of information talk in biology. Genetic representation is a property with considerable explanatory utility.

Why do societies collapse? We use an individual-based evolutionary model to show that, in environmental conditions dominated by low-frequency variation (“red noise”), extirpation may be an outcome of the evolution of cultural capacity. Previous analytical models predicted an equilibrium between individual learners and social learners, or a contingent strategy in which individuals learn socially or individually depending on the circumstances. However, in red noise environments, whose main signature is that variation is concentrated in relatively large, relatively rare excursions, individual learning may be selected from the population. If the social learning system comes to lack sufficient individual learning or cognitively costly adaptive biases, behavior ceases tracking environmental variation. Then, when the environment does change, fitness declines and the population may collapse or even be extirpated. The modeled scenario broadly fits some human population collapses and might also explain nonhuman extirpations. Varying model parameters showed that the fixation of social learning is less likely when individual learning is less costly, when the environment is less red or more variable, with larger population sizes, and when learning is not conformist or is from parents rather than from the general population. Once social learning is fixed, extirpation is likely except when social learning is biased towards successful models. Thus, the risk of population collapse may be reduced by promoting individual learning and innovation over cultural conformity, or by preferential selection of relatively fit individuals as models for social learning. © 2009 Elsevier Inc. All rights reserved.

The deep formal and conceptual link existing between artificial life and artificial intelligence can be highlighted using conceptual tools derived by Karl Popper's evolutionary epistemology.Starting from the observation that the structure itself of an organism embodies knowledge about the environment which it is adapted to, it is possible to regard evolution as a learning process. This process is subject to the same rules indicated by Popper for the growth of scientific knowledge: causal conjectures (mutations) and successive refutations (extinction). In the field of machine learning such a paradigm is represented by genetic algorithms that, simulating biological processes, emulate cognitive processes. From a practical viewpoint, that perspective allows to identify the two different kinds of learning considered by artificial intelligence, knowledge acquisition and skill improvement, and to get a different view of the problem of heuristic knowledge in learning systems. From a theoretical point of view, these considerations can shade a new light on an old epistemological problem: why do we live in a learnable world?

What do people learn when they do not know that they are learning? Until recently, all of the work in the area of implicit learning focused on empirical questions and methods. In this book, Axel Cleeremans explores unintentional learning from an information-processing perspective. He introduces a theoretical framework that unifies existing data and models on implicit learning, along with a detailed computational model of human performance in sequence-learning situations.

Because evolution in natural systems happens so slowly, it is dif- ficult to design inquiry-based labs where students can experiment and observe evolution in the way they can when studying other phenomena. New research in evolutionary computation and artificial life provides a solution to this problem. This paper describes a new A-Life software environment – Avida-ED – in which undergraduate students can test evolutionary hypotheses directly using digital organisms that evolve on their own through the very mechanisms that Darwin discovered.

Learning is the acquisition of new knowledge and skills. It spans a range of processes from practice and rote memorization to the invention of entirely novel abilities and scientific theories that extend past experience. Learning is not restricted to humans: machines and animals can learn, social organizations can learn, and a genetic population can learn through natural selection. In this broad sense, learning is adaptive change, whether in behavior or in belief.

It is often thought that there is one key design principle or at best a small set of design principles, underlying the success of biological organisms. Candidates include neural nets, ‘swarm intelligence’, evolutionary computation, dynamical systems, particular types of architecture or use of a powerful uniform learning mechanism, e.g. reinforcement learning. All of those support types of self-organising, self-modifying behaviours. But we are nowhere near understanding the full variety of powerful information-processing principles ‘discovered’ by evolution. By attending closely to the diver- sity of biological phenomena we may gain key insights into (a) how evolution happens, (b) what sorts of mechanisms, forms of representation, types of learning and development and types of architectures have evolved, (c) how to explain ill-understood aspects of human and animal intelligence, and (d) new useful mechanisms for artificial systems.

Many analogies exist between the process of evolution by natural selection and of learning by reinforcement and punishment. A full extension of the evolutionary analogy to learning to include analogues of the fitness, genotype, development, environmental influences, and phenotype concepts makes possible a single theory of the learning process able to encompass all of the elementary procedures known to yield learning.

Gene targeting has generated a great deal of data on the molecular mechanisms of long-term potentiation and its potential role in learning and memory. However, the interpretation of some results has been questioned. Compensatory mechanisms and the contribution of genetic background may make it difficult to unequivocally prove the existence of a causal (genetic) link between LTP and learning.

The paper considers the problems involved in getting neural networks to learn about highly structured task domains. A central problem concerns the tendency of networks to learn only a set of shallow (non-generalizable) representations for the task, i.e., to miss the deep organizing features of the domain. Various solutions are examined, including task specific network configuration and incremental learning. The latter strategy is the more attractive, since it holds out the promise of a task-independent solution to the problem. Once we see exactly how the solution works, however, it becomes clear that it is limited to a special class of cases in which (1) statistically driven undersampling is (luckily) equivalent to task decomposition, and (2) the dangers of unlearning are somehow being minimized. The technique is suggestive nonetheless, for a variety of developmental factors may yield the functional equivalent of both statistical AND informed undersampling in early learning.