Matthew Schlesinger

Learning to Reach by Constraining the Movement Search Space

Trial-and-error learning strategies play a central role in sensorimotor development during early infancy. However, learning to reach by trial-and-error normally requires a slow and laborious search through the space of possible movements. We propose a computational model of reaching based on the notion that early sensorimotor control is driven by the generation of exploratory movements, followed by the selection and maintenance of adaptive movement patterns. We find that, instead of exhaustively exploring the full search space of movement patterns, the model exploits several emergent constraints that limit the initial size of the movement search space. These constraints exploit both mechanical and kinematic properties of the reaching task. We relate these results to the development of reaching during infancy, and discuss recent findings that have identified similar constraints in young infants.

Artificial Life and Piaget

Abstract Artificial Life is the study of all phenomena of the living world through their reproduction in artificial systems. We argue that Artificial Life models of evolution and development offer a new set of theoretical and methodological tools for investigating Piaget’s ideas. The concept of an Artificial Life Neural Network (ALNN) is first introduced, and contrasted with the study of other recent approaches to modeling development. We then illustrate how several key elements of Piaget’s theory of cognitive development (e.g., sensorimotor schemata, perception-action integration) can be investigated within the Artificial Life framework. We conclude by discussing possible new directions of Artificial Life research that will help to elaborate and extend Piaget’s developmental framework.

Mapping the Landscape of Human-Level Artificial General Intelligence

We present the broad outlines of a roadmap toward human-level artificial general intelligence (henceforth, AGI). We begin by discussing AGI in general, adopting a pragmatic goal for its attainment and a necessary foundation of characteristics and requirements. An initial capability landscape will be presented, drawing on major themes from developmental psychology and illuminated by mathematical, physiological and information processing perspectives. The challenge of identifying appropriate tasks and environments for measuring AGI will be addressed, and seven scenarios will be presented as milestones suggesting a roadmap across the AGI landscape along with directions for future research and collaboration.

The Neural Basis for Visual Selective Attention in Young Infants: A Computational Account

Recent work by Amso and Johnson (Developmental Psychology, 42(6), 1236—1245, 2006) implicates the role of visual selective attention in the development of perceptual completion during early infancy. In the current article, we extend this finding by simulating the performance of 3-month-old infants on a visual search task, using a multi-channel, image-filtering model of early visual processing. Model parameters were systematically varied to simulate developmental change in three neural components of visual selective attention: degree of oculomotor noise, growth of horizontal connections in visual cortex, and duration of recurrent processing in parietal cortex. While two of the three components—horizontal connections and recurrent parietal processing—are each able to account for the visual search performance of 3-month-olds, recurrent parietal processing also suggests a coherent pattern of developmental change in visual selective attention during early infancy. We conclude by highlighting plausible neural mechanisms for modulating recurrent parietal activity, including the development of feedback from prefrontal cortex.

Infants' developing expectations of possible and impossible tool-use events between ages 8 and 12 months

Infants’ developing causal expectations for the outcome of a simple tool-use event from ages 8 to 12 months were investigated. Causal expectations were studied by comparing infants’ developing tool-use actions (i.e. as tool-use agents) with their developing perceptual reactions (i.e. as tool-use observers) to possible and impossible tool-use events. In Experiment 1, tool-use actions were studied by presenting infants, ages 8 and 12 months, with tool-use object-retrieval problems. In Experiment 2, a second age-matched sample of infants watched a comparable series of possible and impossible tool-use events in which a tool was used to retrieve a goal-object. Two core related findings were made. First, infants’ causal action and causal perception develop in parallel. In both action and perception, supporting tool-use develops before surrounding tool-use. Second, infants’ tool-use action develops before their causal perception of comparable tool-use events. The findings support the constructivist hypothesis that infants’ causal actions may develop before and inform their causal perceptions.

Developing classification in action: II. Young chimpanzees (Pan troglodytes)

The development of spontaneous object manipulation in 5 chimpanzees (Pan troglodytes) from ages 15 to 54 months was investigated, focusing on formal properties of subjects’ acts and the objects they manipulated. Young chimpanzees’ manipulation progress from serial one-at-a-time acts on one object to parallel two-at-a-time acts on two or more objects. With age, simultaneous acts become increasingly transformational and identical or reciprocal to each other. Moreover, the class properties of objects manipulated simultaneously change. When presented with objects belonging to two different classes, subjects shift, with age, from manipulating different objects to manipulating identical or similar objects. In all these respects young chimpanzee’ development is similar to human infants’. In others it differs. Most especially, the onset age is later and the development is slower as well as less structurally complex.

Image Free-Viewing as Intrinsically-Motivated Exploration: Estimating the Learnability of Center-of-Gaze Image Samples in Infants and Adults

We propose that free viewing of natural images in human infants can be understood and analyzed as the product of intrinsically-motivated visual exploration. We examined this idea by first generating five sets of center-of-gaze (COG) image samples, which were derived by presenting a series of natural images to groups of both real observers (i.e., 9-month-olds and adults) and artificial observers (i.e., an image-saliency model, an image-entropy model, and a random-gaze model). In order to assess the sequential learnability of the COG samples, we paired each group of samples with a simple recurrent network, which was trained to reproduce the corresponding sequence of COG samples. We then asked whether an intrinsically-motivated artificial agent would learn to identify the most successful network. In Simulation 1, the agent was rewarded for selecting the observer group and network with the lowest prediction errors, while in Simulation 2 the agent was rewarded for selecting the observer group and network with the largest rate of improvement. Our prediction was that if visual exploration in infants is intrinsically-motivated—and more specifically, the goal of exploration is to learn to produce sequentially-predictable gaze patterns—then the agent would show a preference for the COG samples produced by the infants over the other four observer groups. The results from both simulations supported our prediction. We conclude by highlighting the implications of our approach for understanding visual development in infants, and discussing how the model can be elaborated and improved.

Decomposing infants' object representations : A dual-route processing account

The capacity for infants to form mental representations of hidden or occluded objects can be decomposed into two tasks: one process that identifies salient objects and a second complementary process that identifies salient locations. This functional decomposition is supported by the distinction between dorsal and ventral extrastriate visual processing in the primate visual system. This approach is illustrated by presenting an eye-movement model that incorporates both dorsal and ventral processing streams and by using the model to simulate infants’ reactions to possible and impossible events from an infant looking-time study (R. Baillargeon, “Representing the existence and the location of hidden objects: object permanence in 6- and 8-month-old infants”, Cognition, 23, pp. 21–41, 1986.). As expected, the model highlights how the dorsal system is sensitive to the location of a key feature in these events (i.e. the location of an obstacle), whereas the ventral system responds equivalently to the possible and impossible events. These results are used to help explain infants’ reactions in looking-time studies.

Investigating the Origins of Intrinsic Motivation in Human Infants

One of the earliest behaviors driven by intrinsic motivation is visual exploration. In this chapter, I highlight how the development of this capacity is influenced not only by changes in the brain that take place after birth but also by the acquisition of oculomotor skill. To provide a context for interpreting these developmental changes, I then survey three theoretical perspectives that are available for explaining how and why visual exploration develops. Next, I describe work on the development of perceptual completion, which offers a case study on the development of visual exploration and the role of oculomotor skill. I conclude by discussing a number of challenges and open questions that are suggested by this work.

An external focus of attention enhances manual tracking of occluded and visible targets.

The present study investigated the enhancement effects of an external focus of attention (FOA) in the context of a manual tracking task, in which participants tracked both visible and occluded targets. Three conditions were compared, which manipulated the distance of the FOA from the participant as well as the external/internal dimension. As expected, an external FOA resulted in lower tracking errors than an internal FOA. In addition, analyses of participants’ movement patterns revealed a systematic shift toward higher-frequency movements in the external FOA condition, consistent with the idea that an external FOA exploits the natural movement dynamics available during skilled action. Finally, target visibility did not influence the effect of focused attention on tracking performance, which provides evidence for the proposal that the mechanisms that underlie FOA do not depend directly on vision.