Fighting Algo-Discrimination: Relying on AI's Learning Skills and Embedding Fair Information Principles within Machines’ Heart-Design-Specifications
Year: 2018
Type: preprint
Abstract: After having built private networks of knowledge, firms are aware of users’ personal data and are capable of creating systems to sort people into groups. The potential of powerful and opaque algorithms to create discriminatory biases has been widely acknowledged, and there is no explanation provided with regard to their decision-making. People ignore ways, in which their information is created or modified, and there is a need for user-centric systems to implement trust and transparency principles into machines’ design specifications. This paper studies practices that firms conduct to algorithmically reach a perfect audience. The European regime is examined to prove the discriminatory nature of these practices and support that it cannot be justified by law. Taking into account machines’ potential, but also their ability to learn and their inability to forget, proposals are submitted to avoid discrimination. Finally, conclusions are drawn to support that data-scientists, the ones capable of making data speak, could play an important role in embedding fair information principles within machines’ life cycle.
Source: SSRN Electronic Journal
Institutions: Tilburg University, Ionian University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2961648374