Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reversing the Lens: Using Explainable AI to Understand Human Expertise

Roussel Rahman, Aashwin Ananda Mishra, Wan-Lin Hu

TL;DR

The paper investigates how humans learn to solve complex, uncertain problems by applying explainable AI–style graph analyses to archival operator data from tuning a particle accelerator. By representing 27 tuning parameters as graph nodes and using community detection and hierarchical clustering, the authors find a robust three-community structure that persists across novices, intermediates, and experts, while the hierarchical organization tightens with experience. These results support a bounded rationality, divide-and-conquer view of human problem solving and demonstrate the utility of graph-based methods for quantitatively studying human cognition and informing AI reasoning in real-world, high-stakes environments. The work has implications for designing human–AI collaboration and for developing reasoning strategies that scale to complex physical systems.

Abstract

Both humans and machine learning models learn from experience, particularly in safety- and reliability-critical domains. While psychology seeks to understand human cognition, the field of Explainable AI (XAI) develops methods to interpret machine learning models. This study bridges these domains by applying computational tools from XAI to analyze human learning. We modeled human behavior during a complex real-world task -- tuning a particle accelerator -- by constructing graphs of operator subtasks. Applying techniques such as community detection and hierarchical clustering to archival operator data, we reveal how operators decompose the problem into simpler components and how these problem-solving structures evolve with expertise. Our findings illuminate how humans develop efficient strategies in the absence of globally optimal solutions, and demonstrate the utility of XAI-based methods for quantitatively studying human cognition.

Reversing the Lens: Using Explainable AI to Understand Human Expertise

TL;DR

The paper investigates how humans learn to solve complex, uncertain problems by applying explainable AI–style graph analyses to archival operator data from tuning a particle accelerator. By representing 27 tuning parameters as graph nodes and using community detection and hierarchical clustering, the authors find a robust three-community structure that persists across novices, intermediates, and experts, while the hierarchical organization tightens with experience. These results support a bounded rationality, divide-and-conquer view of human problem solving and demonstrate the utility of graph-based methods for quantitatively studying human cognition and informing AI reasoning in real-world, high-stakes environments. The work has implications for designing human–AI collaboration and for developing reasoning strategies that scale to complex physical systems.

Abstract

Both humans and machine learning models learn from experience, particularly in safety- and reliability-critical domains. While psychology seeks to understand human cognition, the field of Explainable AI (XAI) develops methods to interpret machine learning models. This study bridges these domains by applying computational tools from XAI to analyze human learning. We modeled human behavior during a complex real-world task -- tuning a particle accelerator -- by constructing graphs of operator subtasks. Applying techniques such as community detection and hierarchical clustering to archival operator data, we reveal how operators decompose the problem into simpler components and how these problem-solving structures evolve with expertise. Our findings illuminate how humans develop efficient strategies in the absence of globally optimal solutions, and demonstrate the utility of XAI-based methods for quantitatively studying human cognition.
Paper Structure (12 sections, 2 figures)

This paper contains 12 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Modeling and Explaining Behavior in Complex Environments
  3. Methods Used
  4. Experimental Task: Tuning Particle Accelerators
  5. Dataset and Participants
  6. Graph Construction and Analysis
  7. Community Detection
  8. Hierarchical Clustering
  9. Results
  10. Consistent Communities in Graphs across levels of experience
  11. Evolving Hierarchical Structures within Communities with Experience
  12. Conclusions

Figures (2)

  • Figure 1: Graphs of FEL tuning for three groups of operators. The node sizes represent the PageRank values, and the distances between nodes represent the edge weights. Communities (denoted by colors) were identified using the Louvain algorithm and verified using spectral clustering.
  • Figure 2: Hierarchical clustering of FEL tuning graphs for three experience groups. The height shows the distance between elements in the embedding space. The elements are clustered together based on this distance.