AI as an External Cognitive Aid: Scaffolding Reflective Parenting through Source-Grounded LLMs

Abstract

This talk demonstrates how source-grounded Large Language Models (LLMs), specifically Google’s NotebookLM, can be practically applied as external cognitive aids to support the cognitively intensive domain of parental decision-making. The presentation will cover three main points:

• A "curation-as-methodology" approach, arguing that the human-led curation of a high-quality knowledge base (~50 expert sources from developmental psychology and behavioral science) is a critical cognitive task for building user trust.
• A qualitative analysis of how the resulting "AI parenting assistant" functions as a cognitive tool, mapping its features (source-grounded summarization, thematic querying) directly to the cognitive science principles of Cognitive Offloading, Scaffolding, and Metacognition.
• A demonstration of findings from practical test scenarios (e.g., "How do I handle a tantrum?") run against the curated notebook, evaluating the quality of the AI-generated, source-grounded guidance.

Attendees will gain:

• A practical framework for utilizing pre-existing, off-the-shelf LLMs to study and support human cognitive processes, offering a research path that bypasses custom tool development
• Actionable insights into the critical role of source-grounding and human-led curation in designing trustworthy AI assistants for sensitive, high-stakes domains