How and to what extent does anonymizing identifiers in recruitment data allow AI to increase fair and diverse hiring practices through the reduction of implicit human bias
Correcting Biases
Implicit Bias :
Identifying implicit bias
Theories
The Implicit Association Test (Greenwald et al. 1998)
Affect Misattribution Procedure (Payne et al.2005)
Go/No-go Association Task (Nosek & Banaji 2001)
Associationist View
Doxastic Model (Mandelbaum 2016)
In recruitment
Biased weighting of information
Evaluation criteria
Changing weighting of categories based on bias for candidate (Hodson et al., 2002;)
Unspecific criteria leaving gaps for bias (Uhlmann & Cohen, 2005)
click to edit
Biased Interpretation of information
Non-verbal cues (Hugenberg and Bodenhausen (2003),
Selection of hiring panel (Sommers & Norton, 2008)
click to edit
Behaviours (Griggs, 2018)
Job-related traits
Organizational level
Remove biasing category information
Inidividual level
Create unambiguous hiring criteria (Uhlmann & Cohen, 2005)
Highly structured interviews (Bragger et al., 2002)
click to edit
click to edit
AI in Recruitment
Prevalence (Upadhyay
and Khandelwal, 2018))
How does it work? (Raghavan et al. 2000)
Definition (Jarrahi, 2018))
Diversity in Hiring
Likelihood to replicate human bias
Gender Identity(Davison and Bruke 2000)
Ethnicity (Correll, et al. 2007)
Sexuality
Mental or Physical disabilities
In Canada BIPOC
Comprehensive definition of genders
Marketing job opening
Predicting organitzation fit
Screening applicants
Predicting likelihood to leave organization
Indicators used
Age
Gender
Race
Name
Education (in some cases)
Creating postings that use gender-neutral, non-biased language Rab-Kettler & Lehnervp, 2019)