Fairness in Business Process Analysis

 

Student: Timo Pohl

Title: Fairness in Business Process Analysis

Supervisor: Mahnaz Qafari, Alessandro Berti

1st Examiner: Prof. Dr. Wil van der Aalst

2nd Examiner: Prof. Dr. Martin Grohe

Summary:

Fairness is a concept that plays a vital role in supporting human cooperation and has been researched for over 60 years. There are multiple context-dependent meanings, interpretations, and even conflicting theoretical understandings exist. In the domain of process mining, fairness has hardly been considered. The goal of this thesis is to define concepts and measurements for assessing unfair treatment in processes, more specifically in event logs. Based on a comprehensive literature review, we identified concepts, measures, and approaches for assessing fairness applied in areas like machine learning and data science. Based on those concepts and measures, we identified and defined fairness-relevant process outcomes and a set of fairness measures suitable for identifying unfair treatment of cases, resources, and groups of cases in processes based on an event log. We implemented the defined concepts and measures in a fairness checker. The fairness checker automatically detects differences in the treatment of cases, resources, and groups of cases in event logs, and visually presents potential unfairness to the user. We evaluated our approach by applying the fairness checker to event logs with different levels of unfairness generated from four process models. Our evaluation suggests that our approach is able to identify different degrees of unfairness as well as potentially unfair treatments of individual cases, individual resources, and groups of cases in the event logs.