Elicitation Interface for Human-AI Cooperation for Personalized Fairness Optimization

Presenter

Anirudh Sathiya Narayanan

Campus

UMass Amherst

Sponsor

Yair Zick, Department of Computer Science, UMass Amherst

Schedule

Session 2, 11:30 AM - 12:15 PM [Schedule by Time][Poster Grid for Time/Location]

Location

Poster Board A1, Campus Center Auditorium, Row 1 (A1-A20) [Poster Location Map]

Abstract

In a world where Artificial Intelligence is increasingly being used for critical applications such as Healthcare diagnostics, Financial Services and Facial Recognition, one can no longer ignore biases in training models, which can result in huge real-life implications.

Present fair machine learning (ML) methods and fairness allocation strategies are often designed to cater to particular fairness objectives, and are typically predetermined. Contrary to this approach, our current study, "Human-AI Cooperation for Personalized Fairness Optimization" attempts to extract the understanding of fairness from a social planner by employing a series of resource allocation questions. 

The main challenge lies in dealing with the limitations of humans. We cannot expect them to provide precise answers to complex abstract questions involving algebra, but we can assume that they are capable of making comparative judgments about ethical decisions based on real-life situations. Thus, we can ask simply multiple-choice questions and construct a partial order that we can add to the knowledge base.

We will focus on the design of the Query Generator and the Elicitation Interface, the two components that work to generate and obtain the user's preferences. We will also examine how this process can be reciprocated to obtain the fairness concept from a Large Language Model to enable comparison between different models.

Keywords

Algorithmic Fairness, Artificial Intelligence, Ethical Decision Making

Research Area

Algorithms, Combinatorics, Graph Theory, and Game Theory

SIMILAR ABSTRACTS (BY KEYWORD)