GroundWork makes arguments clear and editable for humans and AI
GroundWork: Bridging Human and LLM Reasoning Through a Structured Argumentation Model
Human-Computer Interaction
Summary
It can be hard to keep track of complex arguments when talking with AI models because the reasoning is hidden in the text. The authors created GroundWork, which turns arguments into structured diagrams that both people and AI can edit together. This helps users build, compare, and improve arguments more easily, with every part linked back to real sources. They tested it with experts from different fields and found it made the arguments easier to follow and update collaboratively.
What this means in practice
- •For legal professionals: Organize and revise legal arguments collaboratively with AI while linking claims explicitly to case documents.
- •For medical decision makers: Build and test evidence-based medical arguments involving both expert input and AI reasoning with clear source traceability.
Authors
Sjoerd Vink, Byungmoo Kim, Michael Behrisch, Adam Coscia, Remco Chang
Abstract
Large language models (LLMs) are increasingly integrated into workflows that require reasoning over text documents, yet interaction remains largely conversational. This leaves argumentative structure implicit or embedded in text, making it difficult to maintain, compare, and revise arguments as analyses evolve. We present GroundWork, a mixed-initiative system that makes argument structure a shared, editable state for humans and LLMs. GroundWork operationalizes the Toulmin model of argumentation as a typed graph with explicit operations for human and LLM edits. It exposes this structure through coordinated corpus views and an editable workbench for constructing, comparing, testing, and tracing source-grounded arguments. We evaluate GroundWork through an expert study with four participants spanning consulting, product development, law, and immunology, and three use-case scenarios in intelligence, law, and medicine. Our findings show that GroundWork makes source-grounded arguments inspectable, comparable, and revisable as shared human-LLM state.