TY - GEN
T1 - A Habermasian Approach to Fair Processes in AI Algorithms
AU - Xivuri, Khensani
AU - Twinomurinzi, Hossana
N1 - Publisher Copyright:
© 2022, Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - The traditional emphasis of fairness in AI algorithms tends towards developing fair standards, even though the field of AI and its subfields advance rapidly and creatively, meaning that any AI fair standards could similarly become obsolete rapidly. This paper argues rather for an emphasis on fair processes that are adaptable to AI’s continuous creations and innovations. Specifically, we adapt Jurgen Habermas’ critical theory of communication, the lifeworld and meaning to develop a process framework for AI algorithmic fairness. The framework engages logical-semantic, procedural and performative rules that can be applied to avoid power imbalances and domination by any entity or individual before, during and after AI algorithm development. The framework is applied to the recent case of the biased Twitter image cropping algorithm, which focused on white faces and women’s chests but cropped out black faces.
AB - The traditional emphasis of fairness in AI algorithms tends towards developing fair standards, even though the field of AI and its subfields advance rapidly and creatively, meaning that any AI fair standards could similarly become obsolete rapidly. This paper argues rather for an emphasis on fair processes that are adaptable to AI’s continuous creations and innovations. Specifically, we adapt Jurgen Habermas’ critical theory of communication, the lifeworld and meaning to develop a process framework for AI algorithmic fairness. The framework engages logical-semantic, procedural and performative rules that can be applied to avoid power imbalances and domination by any entity or individual before, during and after AI algorithm development. The framework is applied to the recent case of the biased Twitter image cropping algorithm, which focused on white faces and women’s chests but cropped out black faces.
KW - AI machine learning
KW - Active learning algorithms
KW - Bias
KW - Ethics
KW - Fairness
KW - Jürgen Habermas
KW - Theory of communicative action
UR - http://www.scopus.com/inward/record.url?scp=85125276339&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-95070-5_22
DO - 10.1007/978-3-030-95070-5_22
M3 - Conference contribution
AN - SCOPUS:85125276339
SN - 9783030950699
T3 - Communications in Computer and Information Science
SP - 335
EP - 343
BT - Artificial Intelligence Research - 2nd Southern African Conference, SACAIR 2021, Proceedings
A2 - Jembere, Edgar
A2 - Gerber, Aurona J.
A2 - Viriri, Serestina
A2 - Pillay, Anban
PB - Springer Science and Business Media Deutschland GmbH
T2 - 2nd Southern African Conference on Artificial Intelligence Research, SACAIR 2021
Y2 - 6 December 2021 through 10 December 2021
ER -