Resources › Whitepaper › Towards establishing fairness in ai based candidate screening › Fairness in hiring
Fairness in hiring
In order to build a case for fair hiring, it is important to first explore what fairness means in that context. We found no accepted single definition of fairness in the = context of hiring. However, there is agreement that a fair screening process should not display bias towards an individual or a specific group, demonstrate equity and ensure that decisions are explainable. A detailed discussion related to evaluating fairness and bias in AI systems that make predictions about humans can be found in (Landers & Behrend, 2022).
Authors provide three lenses under which fairness and bias can be evaluated, namely 1) individual attitudes related to justice 2) legal, ethical and moral perspectives and 3) technical domain embedded meanings as defined by quantitative measures such as statistical tests. In (Mehrabi et al., 2021) authors present an in depth analysis of bias and fairness in machine learning where fairness is presented as “In the context of decision-making, fairness is the absence of any prejudice or favoritism toward an individual or a group based on their inherent or acquired characteristics.”, which also applies to hiring decisions.
The American Psychological Association (APA) highlights four ways of thinking about fairness in their “Principles for the Validation and Use of Personnel Selection Procedures”, which is specifically related to hiring methods, (see pages 22-23 in (American Psychological Association, 2018)).
- Fairness as requiring equal group outcomes (e.g., equal passing rates for subgroups of interest). This is a strict view of fairness and lack of equal group outcomes alone is not considered unfair. For example, it ignores the disproportionate representation of groups in the applicant pool for example
- Fairness in terms of the equitable treatment of all examinees during the selection process. Equitable treatment in terms of testing conditions, access to practice materials, performance feedback, including providing reasonable accommodation for test takers with disabilities.
- Fairness as requiring comparable access to the constructs measured by a selection procedure without being unduly advantaged or disadvantaged by other individual characteristics such as age, race, ethnicity, gender, socioeconomic status, cultural background, and disability. These characteristics should not restrict accessibility and affect the measurement of constructs.
- Fairness as lack of bias. Specifically predictive bias where candidates of all subgroups of interest are evaluated using a method that is not impacted by group membership. We will visit bias measurement in depth in section 5, but it is important to note that lack of bias alone does not make a process fair. As stated in (American Psychological Association, 2018), a selection system might exhibit no predictive bias by race or sex but still be viewed as unfair if equitable treatment (e.g., access to practice materials) was not provided to all examinees.
A key takeaway from the above is that mitigating bias plays a key role in building a fair recruitment process. Issues of equitable treatment, access, and scrutiny for possible bias when subgroup differences are observed are also important concerns in candidate selection. However, there is no agreement that the term “fairness” can be uniquely defined in terms of any of these issues (American Psychological Association, 2018).
NEWSLETTER
Join our community
Join our community and get weekly news, tips and insights on recruitment, ethical Ai, better interviewing, and more.
