New framework ensures equity and inclusion in AI for health care

A recent study published in the Journal of Medical Internet Research introduced the EDAI framework, a comprehensive guideline designed to embed equity, diversity, and inclusion (EDI) principles throughout the artificial intelligence (AI) lifecycle. Led by Dr. Samira Abbasgholizadeh-Rahimi, PhD, the Canada Research Chair (Tier II) in AI and Advanced Digital Primary Health Care, the research addresses a significant gap in current AI development and implementation practices in health and oral health care, which often overlook critical EDI factors. With EDAI, AI developers, policymakers, and health care providers now have a roadmap to ensure AI systems are not only technologically sound but also socially responsible and accessible to all.

Through a 3-phase research approach, including a systematic literature review and two international workshops with over 60 experts and community representatives, the research team identified essential EDI indicators to weave into each stage of AI lifecycle, from data collection to deployment. Co-designed with input from diverse voices, this framework puts inclusion at the forefront, ensuring that AI in health and oral health care reflects a range of perspectives and serves everyone more equitably and responsibly.

The AI systems of today are often mirrors reflecting our societal biases rather than windows to a more equitable future. To use AI's power for societal good, we must ensure using frameworks like EDAI to integrate EDI into its lifecycle. Only then can we transform these powerful tools into bridges that connect and uplift everyone, not just the privileged few." 

Dr. Samira Abbasgholizadeh-Rahimi, PhD, the Canada Research Chair (Tier II) in AI and Advanced Digital Primary Health Care

The study funded by the Canadian Institutes of Health Research (CIHR) and the Research Funds of Quebec (FRQ) network i.e., Oral and Oral Health Research Network (RSBO), shows that embedding EDI principles into AI is about much more than just checking a box-;it's about tackling deeper biases within systems and organizations that can prevent AI from truly working for everyone. For example, the EDAI framework can be used by AI developers to design diagnostic tools that consider demographic and cultural diversity. Developers can ensure that datasets include diverse populations, enabling AI to provide accurate diagnoses across various demographics, and preventing biases that traditionally affected certain groups.

Similarly, when designing AI for health care management (like scheduling or resource allocation), using EDAI framework during design could ensure equitable health care by optimizing these systems to prioritize underrepresented or underserved communities. For instance, using EDAI, an AI-based patient scheduling system could be carefully developed and implemented with EDI principles in mind to identify underserved communities and marginalized groups facing accessibility challenges and facilitate access to care for these populations.

Along with offering practical steps and guidance, the EDAI framework sheds light on both the roadblocks and facilitators that can affect how EDI principles are incorporated, giving developers and policymakers the insight to tackle challenges and boost the framework's impact. This initiative is setting the stage for a new standard in AI development and implementation, redefining how AI can enhance health and oral health care for everyone, regardless of background or circumstances.