Atlas of Human-AI Interaction (v1): An Interactive Meta-Science Platform for Large-Scale Research Literature Sensemaking

TL;DR

This paper presents the Atlas of Human-AI Interaction, an open-source platform that uses LLM-powered extraction to convert empirical findings from over 1,000 HAI papers into a structured knowledge graph. It yields 2,037 empirical findings expressed as cause-effect triplets and provides three interactive views (3D graph, cause-effect Sankey, and paper view) to enable large-scale literature sensemaking and cross-domain integration. Expert evaluation with 20 researchers demonstrates Atlas’s utility for gap discovery and interdisciplinary connections, illustrating AI’s potential to transform literature synthesis. Together, the methodological, structural, and practical contributions offer a scalable framework for evidence-based design and computational meta-science in HCI and beyond.

Abstract

Human-AI interaction researchers face an overwhelming challenge: synthesizing insights from thousands of empirical studies to understand how AI impacts people and inform effective design. Existing approach for literature reviews cluster papers by similarities, keywords or citations, missing the crucial cause-and-effect relationships that reveal how design decisions impact user outcomes. We introduce the Atlas of Human-AI Interaction, an interactive web interface that provides the first systematic mapping of empirical findings across 1,000+ HCI papers using LLM-powered knowledge extraction. Our approach identifies causal relationships, and visualizes them through an AI-enabled interactive web interface as a navigable knowledge graph. We extracted 2,037 empirical findings, revealing research topic clusters, common themes, and disconnected areas. Expert evaluation with 20 researchers revealed the system's effectiveness for discovering research gaps. This work demonstrates how AI can transform literature synthesis itself, offering a scalable framework for evidence-based design, opening new possibilities for computational meta-science across HCI and beyond.

Figures (5)

• Figure 1: Graph generation pipeline comprising five stages: (1) research abstract collection, (2) findings triplet extraction, (3) triplet normalization and validation, (4) semantic entity clustering, and (5) structured graph construction with community detection.
• Figure 2: The figure shows four views of the 3D graph Atlas (Left to Right): (Top) a complete network visualization (Full Graph View), the detailed display of the "human:designer" node with its direct connections and relationship information (Node View), (Bottom) the detailed view of the "Machine Learning Models and Explainability" cluster (Cluster View), and the detailed view of the relation between human:#trust and co:collaboration (Edge View).
• Figure 3: Three visualization modes of the Atlas of Human-AI Interaction: (Left) 3D Graph View showing the complete network of empirical findings with nodes representing research concepts and edges showing cause-effect relationships; (Middle) Cause-Effect View using a Sankey diagram to emphasize directional causal flows from AI systems to documented effects; (Right) Paper View providing a searchable repository with detailed paper information and extracted findings for individual study exploration.
• Figure 4: Sample graph data illustrating two nodes: ai>explanation and human> #trust, connected by an edge annotated with the original finding and source paper reference.
• Figure 5: In this figure we present pariticapnts' respective ratings of our 7 Likert-rated items from the survey section. Participants rated their agreement with these items from 1 (Not at all) to 7 (Extremely).