University of Twente Student Theses


Recommendations on bias : detect, mitigate, repeat.

Belchev, B. (2022) Recommendations on bias : detect, mitigate, repeat.

[img] PDF
Abstract:In today's data-driven environment, the adoption of innovative algorithms to improve efficiency and effectiveness has expanded rapidly during the past decade. Despite the benefits they give, they also carry the shortcomings of their creators. The world has already seen these weaknesses in several instances where biased algorithms have sparked public outrage. This can occasionally have severe implications on the lives of individuals affected. A framework has been created to guide the attention of individuals and organizations developing and deploying these algorithms toward their ethical aspects and the sociotechnical system in which they will reside. The framework is intended to stimulate discussion on these ethical challenges, but it does not include recommendations for identifying and mitigating bias. Therefore, the purpose of this study was to identify and synthesize recommendations from the available literature on detecting and mitigating bias. Experiments were conducted with non-expert stakeholders to validate the recommendations for detecting and mitigating bias in algorithms and datasets. A total of 24 recommendations and sub-recommendations for identifying and reducing bias were developed, and the results of the experiments shown that stakeholders with limited expertise in the subject had a reasonable grasp of these recommendations and their applicability.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science BSc (56964)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page