Equity AI: towards a framework to address AI Bias

TEAM

Canada: Christo El Morr (PI), Rachel Gorman (Co-PI), Elham Dolatabadi (co-PI),  Laleh Seyyed-Kalantari (co-PI)
​Research Assistants: Sarah Taleghani, Sabine Fernandes.

Project

Artificial intelligence (AI) has emerged as a transformative technology, shaping numerous aspects of our daily lives. However, its deployment often lacks sufficient consideration of the broader social and ethical implications. Within AI, the sub-field of machine learning (ML) has gained prominence, particularly in organizational and healthcare contexts, where it is increasingly utilized to analyze historical data and construct mathematical models that:

1. Classify data into predefined categories, such as medical diagnoses.
2. Identify previously unknown patterns or clusters based on data similarities.
3. Predict the classification or cluster of new data, such as a patient's condition or a case of violence against women.

Despite its potential, AI risks amplifying social and racial biases when not thoughtfully designed and implemented, contributing to greater health disparities and reinforcing systemic inequalities. Addressing these issues requires urgent and comprehensive assessment of AI biases and the development of strategies to mitigate them, with the ultimate goal of improving health and social outcomes.
​
While the concept of bias is often discussed in the literature, there is a notable gap in addressing its broader social and health implications. Traditional positivist approaches to bias frequently overlook critical insights from disciplines that examine ethnic, gender, and intersectional biases. This underscores the urgent need to understand how AI systems can exacerbate and perpetuate social inequities. Furthermore, it is essential to incorporate concepts such as racism, sexism, ableism, and colonialism into the development of equitable AI frameworks.

This projet aims at 

1. Organizing a Two-Day Seminar
   The seminar will focus on mitigating AI biases and promoting equity in the social sciences and health sectors. A key outcome will be the creation of a framework designed to address AI biases effectively.
2. Establishing a Research Collaboratory
   The project will establish a multidisciplinary and transdisciplinary Research Collaboratory aimed at advancing the Equity AI agenda. This collaboratory will raise awareness about AI biases, foster critical discourse, and provide tools and strategies to address these biases.