Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating a Methodology for Increasing AI Transparency: A Case Study

David Piorkowski, John Richards, Michael Hind

TL;DR

The paper evaluates Richards et al.'s human-centered FactSheet documentation methodology in a healthcare AI field study, assessing usability by non-HC-trained developers, usefulness for diverse consumers, and the balance of benefits and costs. Through 16 participants and four AI models, the study finds the process yields high-quality, reusable FactSheets and is adaptable in real-world settings, though consumer needs remain role-specific and may require iterative tailoring. Consolidation of information into a single source, easier navigation, and reduced coordination are key benefits, while time to create templates and locate prior documentation are notable costs. Overall, the results support the viability of a repeatable, customize-by-need documentation workflow and highlight opportunities for tooling to automate fact capture and tailor content to varied users.

Abstract

In reaction to growing concerns about the potential harms of artificial intelligence (AI), societies have begun to demand more transparency about how AI models and systems are created and used. To address these concerns, several efforts have proposed documentation templates containing questions to be answered by model developers. These templates provide a useful starting point, but no single template can cover the needs of diverse documentation consumers. It is possible in principle, however, to create a repeatable methodology to generate truly useful documentation. Richards et al. [25] proposed such a methodology for identifying specific documentation needs and creating templates to address those needs. Although this is a promising proposal, it has not been evaluated. This paper presents the first evaluation of this user-centered methodology in practice, reporting on the experiences of a team in the domain of AI for healthcare that adopted it to increase transparency for several AI models. The methodology was found to be usable by developers not trained in user-centered techniques, guiding them to creating a documentation template that addressed the specific needs of their consumers while still being reusable across different models and use cases. Analysis of the benefits and costs of this methodology are reviewed and suggestions for further improvement in both the methodology and supporting tools are summarized.

Evaluating a Methodology for Increasing AI Transparency: A Case Study

TL;DR

The paper evaluates Richards et al.'s human-centered FactSheet documentation methodology in a healthcare AI field study, assessing usability by non-HC-trained developers, usefulness for diverse consumers, and the balance of benefits and costs. Through 16 participants and four AI models, the study finds the process yields high-quality, reusable FactSheets and is adaptable in real-world settings, though consumer needs remain role-specific and may require iterative tailoring. Consolidation of information into a single source, easier navigation, and reduced coordination are key benefits, while time to create templates and locate prior documentation are notable costs. Overall, the results support the viability of a repeatable, customize-by-need documentation workflow and highlight opportunities for tooling to automate fact capture and tailor content to varied users.

Abstract

In reaction to growing concerns about the potential harms of artificial intelligence (AI), societies have begun to demand more transparency about how AI models and systems are created and used. To address these concerns, several efforts have proposed documentation templates containing questions to be answered by model developers. These templates provide a useful starting point, but no single template can cover the needs of diverse documentation consumers. It is possible in principle, however, to create a repeatable methodology to generate truly useful documentation. Richards et al. [25] proposed such a methodology for identifying specific documentation needs and creating templates to address those needs. Although this is a promising proposal, it has not been evaluated. This paper presents the first evaluation of this user-centered methodology in practice, reporting on the experiences of a team in the domain of AI for healthcare that adopted it to increase transparency for several AI models. The methodology was found to be usable by developers not trained in user-centered techniques, guiding them to creating a documentation template that addressed the specific needs of their consumers while still being reusable across different models and use cases. Analysis of the benefits and costs of this methodology are reviewed and suggestions for further improvement in both the methodology and supporting tools are summarized.
Paper Structure (23 sections, 1 figure, 7 tables)

This paper contains 23 sections, 1 figure, 7 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Documentation Process
  4. Related Work
  5. Evaluation Methodology
  6. Study Participants and Field Study Context
  7. Interview Protocols
  8. Data Analysis
  9. Results
  10. RQ1: Usability of Documentation Process for FS Team Members
  11. How FactSheet Teams Used Educational Resources
  12. FactSheets Documentation Process Steps
  13. How the Process Helped FactSheet Teams Document AI Models
  14. RQ2: Documentation Usefulness for Consumers
  15. Evaluation of Usefulness
  16. ...and 8 more sections

Figures (1)

  • Figure 1: Consumer responses for evaluating the FactSheet quality. The responses on the right side of the dark vertical line represent responses of 'slightly agree' or higher.