A constructivist connectionist model of transitions on false-belief tasks

How do children come to understand that others have mental representations, e.g., of an object’s location? Preschoolers go through two transitions on verbal false-belief tasks, in which they have to predict where an agent will search for an object that was moved in her absence. First, while three-and-a-half-year-olds usually fail at approach tasks, in which the agent wants to find the object, children just under four succeed. Second, only after four do children succeed at tasks in which the agent wants to avoid the object. We present a constructivist connectionist model that autonomously reproduces the two transitions and suggests that the transitions are due to increases in general processing abilities enabling children to (1) overcome a default true-belief attribution by distinguishing false- from true-belief situations, and to (2) predict search in avoidance situations, where there is often more than one correct, empty search location. Constructivist connectionist models are rigorous, flexible and powerful tools that can be analyzed before and after transitions to uncover novel and emergent mechanisms of cognitive development.     

Logic programs and connectionist networks

One facet of the question of integration of Logic and Connectionist Systems, and how these can complement each other, concerns the points of contact, in terms of semantics, between neural networks and logic programs. In this paper, we show that certain semantic operators for propositional logic programs can be computed by feedforward connectionist networks, and that the same semantic operators for first-order normal logic programs can be approximated by feedforward connectionist networks. Turning the networks into recurrent ones allows one also to approximate the models associated with the semantic operators. Our methods depend on a well-known theorem of Funahashi, and necessitate the study of when Funahashi's theorem can be applied, and also the study of what means of approximation are appropriate and significant.     

The neural basis of cognitive development: a constructivist manifesto

How do minds emerge from developing brains? According to "neural constructivism," the representational features of cortex are built from the dynamic interaction between neural growth mechanisms and environmentally derived neural activity. Contrary to popular selectionist models that emphasize regressive mechanisms, the neurobiological evidence suggests that this growth is a progressive increase in the representational properties of cortex. The interaction between the environment and neural growth results in a flexible type of learning: "constructive learning" minimizes the need for prespecification in accordance with recent neurobiological evidence that the developing cerebral cortex is largely free of domain-specific structure. Instead, the representational properties of cortex are built by the nature of the problem domain confronting it. This uniquely powerful and general learning strategy undermines the central assumption of classical learnability theory, that the learning properties of a system can be deduced from a fixed computational architecture. Neural constructivism suggests that the evolutionary emergence of neocortex in mammals is a progression toward more flexible representational structures, in contrast to the popular view of cortical evolution as an increase in innate, specialized circuits. Human cortical postnatal development is also more extensive and protracted than generally supposed, suggesting that cortex has evolved so as to maximize the capacity of environmental structure to shape its structure and function through constructive learning.     

Catastrophic forgetting in connectionist networks

All natural cognitive systems, and, in particular, our own, gradually forget previously learned information. Plausible models of human cognition should therefore exhibit similar patterns of gradual forgetting of old information as new information is acquired. Only rarely does new learning in natural cognitive systems completely disrupt or erase previously learned information; that is, natural cognitive systems do not, in general, forget ‘catastrophically’. Unfortunately, though, catastrophic forgetting does occur under certain circumstances in distributed connectionist networks. The very features that give these networks their remarkable abilities to generalize, to function in the presence of degraded input, and so on, are found to be the root cause of catastrophic forgetting. The challenge in this field is to discover how to keep the advantages of distributed connectionist networks while avoiding the problem of catastrophic forgetting. In this article the causes, consequences and numerous solutions to the problem of catastrophic forgetting in neural networks are examined. The review will consider how the brain might have overcome this problem and will also explore the consequences of this solution for distributed connectionist networks.     

Associative change in connectionist networks: an addendum

The results of a recent study have provided direct support for the suggestion that conditional learning in rats is best characterized by a 3-layer connectionist network (M. J. Allman, J. Ward-Robinson, & R. C. Honey, 2004). In the 2 experiments reported here, rats were used to investigate the nature of the changes that occur when a stimulus compound is presented, whose components activate hidden units associated with food and no food, and either food or no food is presented. The results of both experiments, while controlling for the possible contribution of associations between these hidden units (within-layer links), provide evidence that the distribution of associative change between units in the hidden layer that are activated by the stimulus compound and those in the output layer (between-layer links) are unequal. They also indicate that associative change is more marked on trials on which no food was presented than on trials on which food was presented.     

Learning in small connectionist networks does not generalize to large networks

Summary A number of recent models of human information processing have been based on connectionist architectures. Such models are designed to illustrate specific psychological principles and are usually implemented in small networks. The assumption implicit in the work is that principles illustrated in small networks can be applied easily to brain-size networks by scaling up as required. To consider the scaling question, we used a back-propagation algorithm to compare learning in both large and small networks and found that learning depended on the size of the network. In small networks, increasing (the rate-of-learning parameter) beyond 1 increased the rate of learning; in large networks, the same manipulation reduced the rate of learning. The example illustrates the difficulty of generalizing across network size and calls into question the assumption that principles illustrated in small networks can be applied to the brain by expansion of the network.Formerly at Queen's University at Kingston     

Re-thinking stages of cognitive development: an appraisal of connectionist models of the balance scale task

The present paper re-appraises connectionist attempts to explain how human cognitive development appears to progress through a series of sequential stages. Models of performance on the Piagetian balance scale task are the focus of attention. Limitations of these models are discussed and replications and extensions to the work are provided via the Cascade-Correlation algorithm. An application of multi-group latent class analysis for examining performance of the networks is described and these results reveal fundamental functional characteristics of the networks. Evidence is provided that strongly suggests that the networks are unable to acquire a mastery of torque and, although they do recover certain rules of operation that humans do, they also show a propensity to acquire rules never previously seen.     

Finding MAPs using strongly equivalent high order recurrent symmetric connectionist networks

Belief revision is the problem of finding the most plausible explanation for an observed set of evidences. It has many applications in various scientific domains like natural language understanding, medical diagnosis and computational biology. Bayesian Networks (BN) is an important probabilistic graphical formalism widely used for belief revision tasks. In BN, belief revision can be achieved by finding the maximum a posteriori (MAP) assignment. Finding MAP is an NP-Hard problem. In previous work, we showed how to find the MAP assignment in BN using High Order Recurrent Neural Networks (HORN) through an intermediate representation of Cost-Based Abduction. This method eliminates the need to explicitly construct the energy function in two steps, objective and constraints. This paper builds on that previous work by providing the theoretical foundation and proving that the resultant HORN used to find MAP is strongly equivalent to the original BN it tries to solve.     

Evaluation and development of a connectionist theory of configural learning

A configural theory of associative learning is described that is based on the assumption that conditioning results in associations between the unconditioned stimulus and a representation of the entire pattern of stimulation that was present prior to its delivery. Configural theory was formulated originally to account for generalization and discrimination in Pavlovian conditioning. The first part of the article demonstrates how this theory can be used to explain results from studies of overshadowing, blocking, summation, and discrimination learning. The second part of the article shows how the theory can be developed to explain a broader range of phenomena, including mediated conditioning, reinforcer devaluation effects, the differential outcomes effect, acquired equivalence, sensory preconditioning, and structural discriminations.     

Hemispheric asymmetries in cognitive modeling: connectionist modeling of unilateral visual neglect

Neglect is an acquired cognitive disorder characterized by a lack of processing of one side of a stimulus or representational space. There are hemispheric asymmetries in its cause and in its effects, but implemented computational models of neglect have tended not to incorporate this fact. The authors report a series of neural network simulations of the line-bisection task. They test the hypothesis that simple, neuroanatomically realistic principles of connectivity in the nervous system can produce emergent behaviors that capture a wide range of quantitative and qualitative data observed in neglect patients presenting with general visuospatial neglect. They demonstrate that exploring low-level architectural principles in implemented computational models is both a productive avenue of research and offers the most parsimonious explanations of behaviors observed in patients.     

Causal cognitive architecture 1: Integration of connectionist elements into a navigation-based framework

The brain-inspired Causal Cognitive Architecture 1 (CCA1) tightly integrates the sensory processing capabilities found in neural networks with many of the causal abilities found in human cognition. Causality emerges not from a central controlling stored program but directly from the architecture. Sensory input vectors are processed by robust association circuitry and then propagated to a navigational temporary map. Instinctive and learned objects and procedures are applied to the same temporary map, with a resultant navigation signal obtained. Navigation can similarly be for the physical world as well as for a landscape of higher cognitive concepts. There is good explainability for causal decisions. A simulation of the CCA1 controlling a search and rescue robot is presented with the goal of finding and rescuing a lost hiker within a grid world. A simulation of the CCA1 controlling a repair robot is presented that can predict the movement of a series of gears.     

A connectionist perspective on the development of reading skills in children

The development of decoding skills has traditionally been viewed as a stage-like process during which children's reading strategies change as a consequence of the acquisition of phonological awareness. More explicit accounts of the mechanisms involved in learning to read are provided by recent connectionist models in which children learn mappings initially between orthography and phonology, and later between orthography, phonology and semantics. Evidence from studies of reading development suggests that learning to read is determined primarily by the status of a child's phonological representations and is therefore compromised in dyslexic children who have phonological deficits. Children who have language impairments encompassing deficits in semantic representations have qualitatively different reading problems centring on difficulties with reading comprehension and in learning to read exception words.     

Comparative cognitive development

This paper aims to compare cognitive development in humans and chimpanzees to illuminate the evolutionary origins of human cognition. Comparison of morphological data and life history strongly highlights the common features of all primate species, including humans. The human mother-infant relationship is characterized by the physical separation of mother and infant, and the stable supine posture of infants, that enables vocal exchange, face-to-face communication, and manual gestures. The cognitive development of chimpanzees was studied using the participation observation method. It revealed that humans and chimpanzees show similar development until 3 months of age. However, chimpanzees have a unique type of social learning that lacks the social reference observed in human children. Moreover, chimpanzees have unique immediate short-term memory capabilities. Taken together, this paper presents a plausible evolutionary scenario for the uniquely human characteristics of cognition.     

Thinking styles and cognitive development

Using R. J. Sternberg's (1988, 1997) theory of thinking styles and W. G. Perry's (1970) theory of cognitive development, the author investigated the nature of thinking styles as they relate to cognitive development. Eighty-two Hong Kong university students (44 male, 38 female) responded to the Thinking Styles Inventory (R. J. Sternberg & R. K. Wagner, 1992) and the Zhang Cognitive Development Inventory (L. F. Zhang, 1997). Statistical analyses provided varying degrees of support for the prediction of an overlap between the thinking style and cognitive development constructs. In general, students who reasoned at a higher cognitive developmental level tended to use a wider range of thinking styles than students who reasoned at a lower cognitive developmental level. Implications of results are discussed in relation to education and research.     

Self-directed activity and cognitive development

Both in constructivist theories of development and in educational applications of such theories, self-directed activity is aleged to play a critical role in the developmental process. The purpose of the present study was to subject this popular assertation to empirical examination. Subjects were fourth- and fifth-graders who showed no formal operational reasoning. Subjects assigned to an experimental condition were matched with yoked-control partners of the same sex and grade. An additional group of subjects was assigned to a simple control condition. Experimental and yoked-control subjects were given opportunities to solve problems requiring formal operational reasoning over a three-month period. The procedure for experimental subjects differed from that for yoked-controls only in that experimental subjects selected the particular information-seeking activities they would engage in. Yoked-control subjects carried out the same activities that had been chosen by their experimental partners. Both groups showed significant progress in the construction of new reasoning strategies, but experimental subjects showed greater progress than yoked-controls. The enhanced progress of the experimental subjects, it is suggested, was due to the increased presence of anticipatory schemes regarding the outcomes of their actions. These anticipatory schemes allowed subjects to better “make use of”, in the cognitive sense—in other words, assimilate into a theoretical framework—the data yielded by the experiments, and thus they gained more from their experience.     

Behavioral versus cognitive classroom friendship networks

Researchers of social networks commonly distinguish between “behavioral” and “cognitive” social structure. In a school context, for example, a teacher’s perceptions of student friendship ties, not necessarily actual friendship relations, may influence teacher behavior. Revisiting early work in the field of sociometry, this study assesses the level of agreement between teacher perceptions and student reports of within-classroom friendship ties. Using data from one middle school teacher and four classes of students, the study explores new ground by assessing agreement over time and across classroom social contexts, with the teacher-perceiver held constant. While the teacher’s perceptions and students’ reports were statistically similar, 11–29% of possible ties did not match. In particular, students reported significantly more reciprocated friendship ties than the teacher perceived. Interestingly, the observed level of agreement varied across classes and generally increased over time. This study further demonstrates that significant error can be introduced by conflating teacher perceptions and student reports. Findings reinforce the importance of treating behavioral and cognitive classroom friendship networks as distinct, and analyzing social structure data that are carefully aligned with the social process hypothesized.