This is the abstract of a talk prepared for the International interdisciplinary seminar on new robotics, evolution and embodied cognition (IISREEC).12th to 15th November 2002, Lisbon, Portugal
Abstract: There is a deep-seated puzzle in the field of early cognitive development. Young infants appear to show by looking experiments that they have concepts about the physical world: objects, number, causality. However, toddlers of 2 or 3 seem to lack the very same knowledge when tested using reaching actions. This is an important debate because it gets to the heart of what is “innate” and what is learned through experience, but it remains unresolved.
From a dynamic systems view, questions about what children “really” know are untenable. Rather, we maintain that knowledge is constituted in action, that is, from the melding of the immediate cues and task with the child's short- and longer-term history of acting in similar situations. Dynamic field models are a powerful way to express this integration of immediate cues and memory to make a decision to act.
In this talk, I will show how these models resolve the apparent contradictions of what infants and toddlers know. I show that experiments supporting claims of both young infants' knowledge of object properties and toddlers' ignorance of object properties can be understood strictly in terms of their task dynamics. I show that the metric properties of the stimuli distances, overlap, and timing all matter profoundly. Moreover, these interactions are highly nonlinear, such that small changes in the metrics can lead to outcomes supporting either “knowing” or “not-knowing.” I suggest that clever task design can produce nearly any desired outcome.
The model simulations give further support to “knowing” as a process rather than a product, and to the inseparability of the dynamics of mind, body, and world