Alignment Divergence Metrics: Measuring Human-AI Concordance in Document Content Analysis

Chunling Niu, Soheila Sadeghi, Marta S. Del Rio-Guerra, Kelly Bradley, Meerna E. Ammari, Arthur E. Hernandez

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper introduces the Alignment Divergence Metrics (ADM) framework quantifying concordance between human and AI analytical outputs in document analysis. We propose five dimensions for assessment: Content Identification, Thematic Organization, Interpretive Depth, Contextual Sensitivity, and Inferential Judgment, each with tailored mathematical formulations. ADM precisely identifies human-AI analytical divergence points, supporting targeted improvements. This standardized approach contributes to more reliable AI tools, responsible research deployment, empirically-grounded alignment theory, and informed policy development-addressing the gap in evaluations that typically focus on task performance rather than alignment with human analytical processes.

Original languageEnglish
Title of host publicationProceedings - 2025 IEEE Conference on Artificial Intelligence, CAI 2025
Pages1235-1238
Number of pages4
ISBN (Electronic)9798331524005
DOIs
StatePublished - 2025
Event3rd IEEE Conference on Artificial Intelligence, CAI 2025 - Santa Clara, United States
Duration: May 5 2025May 7 2025

Publication series

NameProceedings - 2025 IEEE Conference on Artificial Intelligence, CAI 2025

Conference

Conference3rd IEEE Conference on Artificial Intelligence, CAI 2025
Country/TerritoryUnited States
CitySanta Clara
Period5/5/255/7/25

Bibliographical note

Publisher Copyright:
© 2025 IEEE.

Keywords

  • divergence metrics
  • document content analysis
  • Human-AI alignment
  • multi-dimensional evaluation
  • text analysis concordance

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Computer Vision and Pattern Recognition
  • Information Systems and Management
  • Modeling and Simulation

Fingerprint

Dive into the research topics of 'Alignment Divergence Metrics: Measuring Human-AI Concordance in Document Content Analysis'. Together they form a unique fingerprint.

Cite this