This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Effort-aware Fairness: Incorporating a Philosophy-informed, Human-centered Notion of Effort into Algorithmic Fairness Metrics
Created by
Haebom
Author
Tin Nguyen, Jiannan Xu, Zora Che, Phuong-Anh Nguyen-Le, Rushil Dandamudi, Donald Braman, Furong Huang, Hal Daum e III, Zubin Jelveh
Outline
This paper proposes the concept of Effort-Aware Fairness (EaF), pointing out that existing metrics, such as demographic equality, for assessing the fairness of artificial intelligence (AI) systems fail to account for the degree of effort exerted by individuals in the input feature space. EaF builds on the concept of "force," which considers the temporal trajectory and inertia of features. Beyond theoretical foundations, we present (1) pre-registered human experiments demonstrating that people consider the temporal trajectory of features more than their aggregate values in both stages of the individual fairness assessment process. Furthermore, we present a pipeline for calculating effort-aware individual/group fairness in criminal justice and personal finance. This research enables AI model auditors to detect and correct unfair decisions for individuals who remain disadvantaged despite significant effort to overcome systemic disadvantage.
Takeaways, Limitations
•
Takeaways:
◦
Emphasizes the importance of effort in assessing AI fairness and presents a new perspective that overcomes the limitations of existing indicators.
◦
Presenting theoretical foundations and empirical evidence for evaluating fairness considering effort.
◦
Developing a pipeline demonstrating practical applicability in criminal justice and personal finance.
◦
Improving the fairness of AI models and providing the possibility of correcting unfair decisions.
•
Limitations:
◦
Lack of clarity and potential for subjective interpretation in the definition and measurement of ‘effort’.
◦
The generalizability of the proposed pipeline and its applicability to various situations need to be verified.
◦
Further review of the number of participants and sample representativeness in human subjects trials is needed.
◦
Lack of detailed explanation of the quantification and calculation process of the concept of 'force'.
