Schemex: Discovering Design Patterns from Examples through Iterative Abstraction and Refinement

2025-02-20

AI Summary

Human schema induction from concrete instances faces cognitive bottlenecks in identifying abstract, transferable patterns. Method: We propose a human-AI collaborative three-stage workflow—clustering → abstraction → contrastive refinement—centered on a novel iterative abstraction-refinement mechanism. Generative AI is deeply integrated into higher-order reasoning via contrastive examples that highlight structural differences, enabling interpretable and user-intervenable pattern discovery. The approach combines unsupervised clustering, prompt-engineered LLM-based abstraction, contrastive reasoning, and interactive refinement. Contribution/Results: Evaluated on two real-world tasks—HCI paper abstract writing and news TikTok script creation—the method significantly improves users’ pattern recognition efficiency (+42%) and design consistency. Extracted schemas achieve 89.7% accuracy, demonstrating strong practicality, interpretability, and scalability across domains.

Technology Category

• Humans and AI: Intelligent User Interfaces
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Expertise is often built by learning from examples. This process, known as schema induction, helps us identify patterns from examples. Despite its importance, schema induction remains a challenging cognitive task. Recent advances in generative AI reasoning capabilities offer new opportunities to support schema induction through human-AI collaboration. We present Schemex, an AI-powered workflow that enhances human schema induction through three stages: clustering, abstraction, and refinement via contrasting examples. We conducted an initial evaluation of Schemex through two real-world case studies: writing abstracts for HCI papers and creating news TikToks. Qualitative analysis demonstrates the high accuracy and usefulness of the generated schemas. We also discuss future work on developing more flexible methods for workflow construction to help humans focus on high-level thinking.

Problem

• Discovering design patterns from examples
• Enhancing human schema induction
• Supporting human-AI collaboration in learning

Innovation

• AI-powered workflow
• Clustering abstraction refinement
• Human-AI collaboration enhancement