Just Hiring! How to counter discrimination in AI-powered software for recruitment. A toolkit for software developers
Contains bibliogr. pp. 79-86
"In this Toolkit, we offer introductory information and technical recommendations on how to mitigate the risk of discrimination in software development of systems used for recruitment. This Toolkit addresses software developers and product managers working or interested in this field, and it intends to help them gain a better understanding of how they can actively contribute." (Page 4)
PART 1: WHY IS ALGORITHMIC DISCRIMINATION PROBLEMATIC?
1 Algorithmic discrimination in hiring is a lose-lose scenario for all parties involved, 9
2 Discrimination can occur at each stage of hiring processes. Sources of bias and other challenges, 12
3 Algorithmic discrimination in hiring is real - and significantly affects real people. Experiences of navigating in opaque hiring processes, 16
PART 2: WHAT DOES THE LAW SAY AND WHAT LEGAL LOOPHOLES DOES IT LEAVE?
4 EU Non-Discrimination law: important safeguards as a starting point, 24
5 AI Act: Providers of high-risk AI hiring tools bnear special responsibility, 28
6 GDPR: Every processing of personal data needs careful legal assessment, 34
7 Detecting discrimination in AI systems by using sensitive data - a necessity that poses a dilemma, 39
PART 3: HOW CAN YOU AS A SOFTWARE DEVELOPER COUNTER ALGORITHMIC DISCRIMINATION IN HIRING?
8 Stakeholder mapping makes processes clear and straightforward, 44
9 UI/UX Design: Show less personal information on interfaces - and show it at later stages, 47
10 XAI: Explaining algorithmic suggestions is key to better recruitment, 50
11 Do not rely on module-level tests only - consider real-world settings, 54
12 Auditing: Thorough documentation and analysis across the AI life cycle helps prevent bias, 56
13 Fairness-enhancing interventions in the ranking pipeline can reduce ranking bias, 61
14 Comprehensive evaluation and monitoring before and after deployment is key, 66
15 Overview: FINDHR tools to counter discrimination in AI-assisted hiring, 73
1 Algorithmic discrimination in hiring is a lose-lose scenario for all parties involved, 9
2 Discrimination can occur at each stage of hiring processes. Sources of bias and other challenges, 12
3 Algorithmic discrimination in hiring is real - and significantly affects real people. Experiences of navigating in opaque hiring processes, 16
PART 2: WHAT DOES THE LAW SAY AND WHAT LEGAL LOOPHOLES DOES IT LEAVE?
4 EU Non-Discrimination law: important safeguards as a starting point, 24
5 AI Act: Providers of high-risk AI hiring tools bnear special responsibility, 28
6 GDPR: Every processing of personal data needs careful legal assessment, 34
7 Detecting discrimination in AI systems by using sensitive data - a necessity that poses a dilemma, 39
PART 3: HOW CAN YOU AS A SOFTWARE DEVELOPER COUNTER ALGORITHMIC DISCRIMINATION IN HIRING?
8 Stakeholder mapping makes processes clear and straightforward, 44
9 UI/UX Design: Show less personal information on interfaces - and show it at later stages, 47
10 XAI: Explaining algorithmic suggestions is key to better recruitment, 50
11 Do not rely on module-level tests only - consider real-world settings, 54
12 Auditing: Thorough documentation and analysis across the AI life cycle helps prevent bias, 56
13 Fairness-enhancing interventions in the ranking pipeline can reduce ranking bias, 61
14 Comprehensive evaluation and monitoring before and after deployment is key, 66
15 Overview: FINDHR tools to counter discrimination in AI-assisted hiring, 73
