Holistic Assurance Methods to Support Technical/Ethical Reasoning in AI (HAMSTER-AI)

19 Pages Posted: 1 Dec 2022

See all articles by T. Scott Ankrum

T. Scott Ankrum

The MITRE Corporation; ACM

Catherine Petrozzino

The MITRE Corporation

Stuart S. Shapiro

MITRE Corporation

Ellen Laderman

The MITRE Corporation

Cathleen Zhang

affiliation not provided to SSRN

Date Written: September 30, 2021

Abstract

To date, solutions to the AI ethics challenge have mostly focused on high-level principles or model adjustments in order to address targeted equity issues. AI solution developers have had limited actionable solutions to address ethics holistically. To address this need, we propose establishing actionable ethics through the development of a Lifecycle Ethical Analysis (LEA) process that relies on existing methods which are adapted to support ethics in AI.

In this paper, we examine the use of counterfactual analysis in an “off-label” way as a method to support the Problem Definition phase of the AI lifecycle and structured assurance as a method to support the Data and Modeling phase. We hypothesize that these methods can be adapted to support ethics in AI and serve to generate an ethics profile for a project. We further propose that these methods will support translation between ethical reasoning and technical reasoning, which is essential for empowering AI technical staff to address their projects’ ethical posture. We ground this analysis by applying it to an on-going AI project.

Our results show that we could successfully adapt counterfactual analysis and structured assurance to support AI ethics analysis, and further, that these methods provided the AI project team with insight on ethics for their project that they otherwise lacked. An area identified for improvement was better informing the project team on how to supply evidence in building the ethics assurance case.

Keywords: artificial intelligence, AI, ethics, assessment, life cycle

Suggested Citation

Ankrum, Scott and Petrozzino, Catherine and Shapiro, Stuart and Laderman, Ellen and Zhang, Cathleen, Holistic Assurance Methods to Support Technical/Ethical Reasoning in AI (HAMSTER-AI) (September 30, 2021). Available at SSRN: https://ssrn.com/abstract=4248186 or http://dx.doi.org/10.2139/ssrn.4248186

Scott Ankrum

The MITRE Corporation ( email )

7515 Colshire Blvd.
McLean, VA 22102
United States
2407317581 (Phone)
20874 (Fax)

ACM ( email )

1601 Broadway, 10th Floor
New York, NY 10019-7434
United States

Catherine Petrozzino

The MITRE Corporation ( email )

7515 Colshire Blvd.
McLean, VA 22102
United States

Stuart Shapiro (Contact Author)

MITRE Corporation ( email )

202 Burlington Road
Bedford, MA 01730
United States

Ellen Laderman

The MITRE Corporation ( email )

7515 Colshire Blvd.
McLean, VA 22102
United States

Cathleen Zhang

affiliation not provided to SSRN

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
22
Abstract Views
169
PlumX Metrics