Automating the Risk of Bias

59 Pages Posted: 27 Nov 2019 Last revised: 2 Feb 2020

Date Written: November 14, 2019

Abstract

Artificial intelligence (“AI”) is a transformative technology that has radically altered decisionmaking processes. Evaluating the case for algorithmic or automated decision-making (“ADM”) platforms requires navigating the tensions between two normative concerns. On one hand, relying on ADM platforms may lead to more efficient, accurate, and objective decisions. On the other hand, early and disturbing evidence suggests that ADM platform results may demonstrate biases, undermining proponents’ claims that this special class of algorithms will democratize markets and increase inclusion.

State law assigns decision-making authority to the boards of directors of corporations. State courts and lawmakers accord significant deference to the board in the execution of its duties. Among its duties, a board must employ effective oversight policies and procedures to manage known risks. The board of directors and senior management of firms integrating ADM platforms must monitor operations to mitigate litigation, reputation, compliance and regulatory risks that arise as a result of the integration of algorithms.

Evidence demonstrates that heterogeneous teams may identify and mitigate risks more successfully than homogeneous teams. These teams overcome cognitive biases such as confirmation, commitment, overconfidence and relational biases. In the wake of the recent financial crisis firms adopted structural and procedural governance methods adopted by firms to mitigate various types of risks; these approaches may prove valuable in mitigating the risk of algorithmic bias. More specifically, this Article explores the literature on gender balance in leadership and on boards and proposes that AI firms increasing gender diversity among developers, senior managers, and members of the boards of technology firms in an effort to mitigate the risk of bias. Building such teams in firms developing or integrating AI will require creative and thoughtful recruiting and retention strategies. While improving gender balance may not alleviate cognitive biases that influence AI development and management teams, this Article concludes that integrating diverse may mitigate exposure to risks engendered by integrating complex AI models.

Keywords: Artificial Intelligence, AI, Algorithms, Algorithmic Bias, Corporate Governance, Boards of Directors

Suggested Citation

Johnson, Kristin N., Automating the Risk of Bias (November 14, 2019). George Washington Law Review, Vol. 87, No. 6, 2019, Tulane Public Law Research Paper No. 19-12, Available at SSRN: https://ssrn.com/abstract=3486723

Kristin N. Johnson (Contact Author)

Emory University - Law School ( email )

1301 Clifton Road, N.E.
Atlanta, GA USA 30306
United States

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

Paper statistics

Downloads
106
Abstract Views
585
rank
295,032
PlumX Metrics