What drives the decisions engineering designers make, and can language models map those decisions and their causes from designers’ own words?
Characterizing Design Decisions with LLMs
Use LLMs to identify design decisions and the factors that influence them from qualitative text (e.g., interviews, design notebooks, reports). One outcome is constructing causal influence diagrams that represent decisions and relationships.
Objectives
- Build an ingestion and extraction pipeline for design texts (interviews, logs, reports)
- Develop a schema for decisions, factors, and dependencies; prototype LLM-assisted labeling and evaluate vs. expert annotations
- Generate causal influence diagrams from extracted structures and validate with participants
- Release coding guidelines, samples, and a reproducible workflow