Project Details
Projekt
Learning From Environment Through the Eyes of Children
Subject Area
Developmental and Educational Psychology
Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
Term
since 2024
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 539642788
The paradigm of social science research in child development often revolves around observations, interviews, and questionnaires, constraining our understanding to adult perspectives and third-party evaluations. Our interdisciplinary proposal, situated at the intersections of developmental psychology and computer science, aims to revolutionize this paradigm by capturing children's visual experiences from a first-person perspective. This is informed by the recognition that learning is a complex computational problem that cannot be fully understood without considering the dynamic interplay between children and their environment. Leveraging recent advancements in deep learning architectures, we propose a robust methodological framework that seamlessly integrates wearable technologies and artificial intelligence (AI) to explore, quantify, and interpret the daily experiences of children aged one to five in naturalistic settings. We aim to operationalize three Work Packages (WPs) that focus on the following goals: WP1: Gaze and Cognition: Using wearable eye-trackers, we aim to characterize children's eye gaze patterns during free play and link these to various cognitive and psychological metrics, such as cognitive control, memory, and behavioral tendencies like hyperactivity and inattention. WP2: Environmental Quality: We plan to assess the quality and information density of children's home environments through artificial intelligence algorithms that can analyze video feeds captured via head-mounted cameras. This would give us an objective measure of environmental enrichment and available learning opportunities. WP3: Parent-Child Interaction Dynamics: Exploiting time-series analysis techniques, we will investigate the temporal relationships between the information sampled by children and their parents during interactions, aiming to discern predictive patterns that may relate to children’s long-term memory formation. Our methodology will employ head-mounted cameras to capture children and caregivers perspective during their interactions and everyday live situations at home. Moreover, we will predominantly implement state-of-the-art deep learning algorithms (such as CLIP, Whisper, and others) to perform (semi-)automatic data coding, substantially reducing the need for labor-intensive manual coding to test our hypothesis at scale. To establish external validity, we will correlate AI-derived metrics with traditional evaluation methods like partly hand coded videos, surveys and psychological tests. This proposal aligns with SPP Research Area 4, aiming for a new multimodal modeling approach in surveys for social sciences. Merging developmental psychology and computer science expertise, our project offers a novel, tech-enabled perspective on child development. We anticipate our insights will fill existing gaps in early childhood understanding and spur next-gen social science methods.
DFG Programme
Infrastructure Priority Programmes
