Project
From MESAwiki
Contents |
Summary
Although a developing infant and its caregiver may appear to be a simple social system, their behaviors are complex and reciprocal, driven by a rapid play of cognitions, emotions, and perceptions. The system is dynamic on several levels: not only does the dyad’s behaviors shift from moment to moment, and respond to rich, changeable environments; but the infant changes from week to week—acquiring new behaviors, refining skills and sensitivities, and gaining control over thought and action. Concurrently, caregivers adapt their behaviors towards infants to accommodate infants’ new skills. In this interplay of social action and reaction, within a changing, dynamic system, infants gradually learn to interact effectively with adults. Within months infants will begin using language, and struggle to master other complex skills their culture holds to be important. These capacities—language and cultural learning— rest on a fundamental social capacity: sharing attention with other people. We will test and refine a new theory of how attention-sharing develops, using state-of-the-art behavioral and computational methods.
Attention-sharing is noticing what another individual is attending to, and shifting one’s own attention to it because the other is attending to it. This is a distinctive human social behavior. Although much descriptive evidence has been reported on the development of attention-sharing (mostly on older infants’ tendency to follow their caregiver’s gaze), we know very little about how attention-sharing emerges in the first half-year of life. We also do not know how caregivers behave with infants to provide the raw data for infants to learn attention-sharing. The central problem is that we lack a theory of the emergence of attention-sharing skills. This proposal is designed to find the antecedents of attention-sharing in infants. One part of the proposal is a longitudinal behavioral study of infants and parents, to find out “what matters” for acquiring the first attention-sharing skills. The other part is a series of computer and robotic simulations of infant-parent interactions, using data from the longitudinal study as well as learning software based on a new theory of the emergence of early attention-sharing behaviors.
Intellectual Merit: This multidisciplinary project builds on our previous behavioral and modeling efforts, and goes further in several ways. First, the longitudinal study of infants’ attention-following uniquely combines (1) month-by-month measures of infants’ attention-following skills; (2) in-home observation of caregivers’ behaviors with the infant; and (3) laboratory measures of learning and perceptual processes that might promote attention-following. Second, the computational models make use of advanced graphical simulation and robotics technology to systematically study the dynamics of infant learning and infant-parent social interactions. Third, the behavioral and modeling efforts are closely related: for example, the computer simulations use data from “real” parents to “teach” simulated infants; also, the learning processes tested in the longitudinal study were confirmed by preliminary computer simulations. Fourth, the theory we are building incorporates learning, perception, and emotional processes of young infants, as they work on behavioral input from caregivers. Our interdisciplinary approach will test, in simulated infant-caregiver dyads, how these factors let infants learn attention-sharing skills. The results will provide unprecedented understanding of a major early social skill.
Broader Impacts: The proposed longitudinal study will yield a database that answers many questions about cognitive and social development during the first year. The use of state-of-the-art embodied computational modeling technology will open up new avenues for studying the dynamics of social systems in general, and infant social development in particular. If our approach is successful, we expect more research groups to employ related models to study other problems in human social dynamics. Also, a deeper understanding of how attention-sharing skills develop is crucial for understanding deficits of early social behavior. For example, children with autism show profound attention-sharing difficulties, and the results of this research could point towards new diagnostic and treatments methods. Finally, the results will point out new ways to develop “autonomous systems” (e.g., computerized teaching assistants; “helper” robots) that become socially intelligent and interactive via human-like developmental processes. Thus, this proposal will lead to models that transcend current machine learning approaches to social development.
Observational project
We are currently gathering observational data analyzing naturalistic interactions between parents and infants. We are particularly interested in the emergence of shared attention between 3 to 12 months:
- How do infants' attention skill emerge during the first year?
- How are caregivers' and infants' attention distributed in free play?
- What actions do caregivers use to get infants' attention?
- What kind of events precede shared attention?
- How does these interactions change with infants' age?
Our recent findings suggest the development of infant learning relationships between caregiver gaze, caregiver hand motion cues and infant gaze following events in scripted play interactions. Currently, we are further investigating the significance of such attention sharing events in free play interactions.
Modelling project
In this study we develop embodied computational models of typical and atypical emergence of shared attention. Our models simulate the fine-grained interactions between infants, their caregivers, and objects in the environment to provide a learning account of the emergence of gaze following.
A typical modeling meeting
