No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Competition Affects Word Learning in a Developmental Robotic System
Abstract
It is well-established that toddlers can correctly select a novel referent from an ambiguous array in response to a novel label, or fast-map (Golinkoff, Bailey, Hirsh-Pasek & Wenger, 1992; Carey, 1978). There is a growing consensus this ability depends at least in part on a mutual exclusivity-type process; that is, if all-but-one objects in an array have a label (“known objects”), the novel label must refer to the novel object (for a review, see Halberda, 2006). However, the precise mechanism underlying this phenomenon is debated. Horst, Scott & Pollard (2010) shed light on this issue by investigating the effect of contetxt on word learning. The authors familiarised 2-year-old children with novel label-object mappings by asking them to select a single novel object from arrays containing two, three or four known objects. At test, when presented with arrays of just novel target objects, only children who had encountered the novel label/object mappings with fewer competitors recalled these mappings; children who saw arrays with more competitors did not recall mappings at levels greater than expected by chance. The authors argued that attention to competitor objects is fundamental to the mutual exclusivity process and, by extension, word learning.
Here, we use a computational model to explore whether this behaviour can emerge from simple associative processes, or whether higher-level reasoning is required. We present a developmental robotic replication of the target empirical study using a variant of the connectionist Epigenetic Robotic Architecture (Morse, de Greeff, Belpeame, & Cangelosi, 2010) implemented in the iCub humanoid robot (Metta et al., 2010). The replication demonstrates that the apparently complex reasoning demonstrated by the children in the target empirical study emerges from low-level inhibitory processes in the computational model; more generally the current study indicates that word learning may depend on relatively simple bottom-up perceptual processing rather than complex, top-down reasoning.
... Once this set has been established, we are able to begin the experiment. Replicating the fast-mapping experiments (Twomey, Morse, Cangelosi, & Horst, 2014), we present combinations of multiple objects (either all known, or including a single novel object) and ask iCub where one of them is (verbally asking by the name of the object) (see Fig. 3). ...
... For example, performance at fast mapping declines rapidly when more distractors (known objects) are present when asked to find an X (where X is novel word). Fig. 4 highlights the close match between the robot and child data for this task (see Twomey et al., 2014, for full details of this experiment). ...
... Similarly retention and extension have been explored following presentations of narrow or wide variance in the novel object on subsequent naming trails leading to differences in generalization and category formation (Twomey et al., 2014), and recent work (Horst & Morse,unpublished results;see Fig. 5 for a summary of results) has explored what happens when features such as color are held constant across known and novel objects, in all cases demonstrating a very close fit between this model and the child data in a variety of fast-mapping experiments (Fig. 6). ...
Most theories of learning would predict a gradual acquisition and refinement of skills as learning progresses, and while some highlight exponential growth, this fails to explain why natural cognitive development typically progresses in stages. Models that do span multiple developmental stages typically have parameters to “switch” between stages. We argue that by taking an embodied view, the interaction between learning mechanisms, the resulting behavior of the agent, and the opportunities for learning that the environment provides can account for the stage-wise development of cognitive abilities. We summarize work relevant to this hypothesis and suggest two simple mechanisms that account for some developmental transitions: neural readiness focuses on changes in the neural substrate resulting from ongoing learning, and perceptual readiness focuses on the perceptual requirements for learning new tasks. Previous work has demonstrated these mechanisms in replications of a wide variety of infant language experiments, spanning multiple developmental stages. Here we piece this work together as a single model of ongoing learning with no parameter changes at all. The model, an instance of the Epigenetic Robotics Architecture (Morse et al 2010) embodied on the iCub humanoid robot, exhibits ongoing multi-stage development while learning pre-linguistic and then basic language skills.
... However, because that model does not instantiate autonomous processes of visual exploration and attention, it is still unclear how the word learning system, as opposed to the modeler, comes to view novel and familiar stimuli as more or less salient. The more autonomous model of Twomey and colleagues captures related data with older children, but because either the modeler selects the novel object for the robot (Twomey et al., 2013), or the robot looks at and processes all test objects (Twomey, Morse, Cangelosi, & Horst, 2014), it does not inform questions regarding the basis for children's bias. ...
Theories of cognitive development must address both the issue of how children bring their knowledge to bear on behavior in-the-moment, and how knowledge changes over time. We argue that seeking answers to these questions requires an appreciation of the dynamic nature of the developing system in its full, reciprocal complexity. We illustrate this dynamic complexity with results from two lines of research on early word learning. The first demonstrates how the child's active engagement with objects and people supports referent selection via memories for what objects were previously seen in a cued location. The second set of results highlights changes in the role of novelty and attentional processes in referent selection and retention as children's knowledge of words and objects grows. Together this work suggests that understanding systems for perception, action, attention, and memory, and their complex interaction, is critical to understand word learning. We review recent literature that highlights the complex interactions between these processes in cognitive development and point to critical issues for future work.
... In response to the question where is the dax (dax is a novel word), it points at the unknown object. With this iCub demonstrates "fast-mapping", which is also observed in young children when they learn to map words to objects relying on only a few exposures and certain learning constraints [77] ...
Recent
years
have
seen
a growing
interest
in applying insights from developmental psychology to build artificial intelligence and robotic systems. This endeavour, called developmental robotics, not only is a novel method of creating artificially intelligent systems, but also offers a new perspective on the development of human cognition. While once cognition was thought to be the product of the embodied brain, we now know that natural and artificial cognition results from the interplay between an adaptive brain, a growing body, the physical environment and a responsive social environment. This chapter gives three examples of how humanoid robots are used to unveil aspects of development, and how we can use development and learning to build better robots. We focus on the domains of word-meaning acquisition, abstract concept acquisition and number acquisition, and show that cognition needs embodiment and a social environment to develop. In addition, we argue that Spiking Neural Networks offer great potential for the implementation of artificial cognition on robots.
ResearchGate has not been able to resolve any references for this publication.