Theory session
Overview
The theory session addresses the problem of anti-discrimination for AI in recruitment with a transdisciplinary approach, providing to attendees an integrating viewpoint on the topic presenting and discussing legal/ethical, cultural, and technical aspects required to prevent, detect, and mitigate discrimination risks in artificial intelligence-based human recommendation and algorithmic hiring.
In this course, AI is understood in a broad sense as any computational method used for tasks normally requiring human intelligence: machine learning, deep learning, logic-based, statistical approaches, and possibly others. Human recommendation is understood as categorizing, ranking, and recommending people.
Structure
Three core macro sections will be aimed at understanding, mitigating, and accounting for bias and discrimination.
The first macro section on bias understanding will describe sources of bias and discrimination, how they can manifest in data and decision models and will introduce both an ethical and legal perspective on discrimination with a particular focus on employment and recruitment.
The macro section on mitigation will discuss algorithmic strategies for fairness in ranking: pre-processing (modifying training data), in-processing (modifying algorithms), and post-processing (modifying outputs).
The last macro section will present methods for explaining AI decisions and models behavior and methodologies for auditing algorithms for biases.
Format
The course will be organised into 10 theoretical modules of about 1.5 hours per session, with a break in the middle. The topics of the Theory Sessions (TS) are listed in below:
- TS01 Introduction
- TS02 Ethics I: Ethics in Algorithmic Decision Making
- TS03 Ethics II: Discrimination
- TS04 Legal Perspective on Discrimination
- TS05 Job Search Portals
- TS06 Fair Information Access
- TS07 Fair Rankings I
- TS08 Fair Rankings II
- TS09 Explainability
- TS010 Auditing Online Job Search