The problem of bias in hiring remains acute. Few businesses today deliberately discriminate in their hiring practices. Yet a recent meta-study (a comprehensive review of multiple studies) reveals that hiring discrimination against blacks and Latinos has not significantly improved over the past 25 years. As the Financial Times notes in a recent article, “Bias in the hiring process is prevalent and hard to eradicate.”
Often the issue is not intentional discrimination, but the adverse impact of the unconscious biases we all have.
The term “adverse impact” relates to employment practices that may on the surface seem fair and objective, but that actually have a disproportionately negative effect on a particular set of applicants or workers. In the United States, a company employing policies that are determined to have an adverse impact on protected groups may find itself in legal jeopardy.
IBM Sets Watson to Combat Unconscious Bias in Hiring
Recognizing that unconscious bias in hiring is a serious issue for many organizations, IBM has introduced its Watson Recruitment’s Adverse Impact Analysis capability to help combat the problem. According to Bob Schultz, General Manager, IBM Watson Talent Solutions, the purpose of this new artificial intelligence (AI) product is to help users identify “instances of bias related to age, gender, race, education, or previous employer by assessing an organization’s historical hiring data and highlighting potential unconscious biases.”
The Necessity of Bias-Free Training Data
Schultz notes that IBM honed this new AI tool’s capabilities by working with psychologists to create an unbiased AI scoring system. A critical issue in that effort was ensuring that the AI was trained with data that did not itself contain hidden biases. Aware that some commercial AI products have themselves been shown to be discriminatory (in favor of white males, for example) due to faulty training data, IBM researchers specifically focused on developing training datasets that are free of their own inherent biases. As part of that process, IBM developed a rating system that ranks the relative fairness of an AI platform and explains how it makes its decisions.
The Watson Recruitment’s Adverse Impact Analysis capability was introduced as part of IBM’s largest ever AI toolset release. Customers that have already begun using the product include BuzzFeed and H&R Block.