Table 3 General recommendations to support six strategic priorities for successful use of artificial intelligence by public health organizations
Strategic Priority | General Recommendations |
|---|---|
Governance | |
- Clarify data leadership roles and responsibilities - Review current organizational governance - Understand and operationalize higher-level governance   ○ Involve subject-matter experts in AI, data management and information technology   ○ Develop a mechanism for community and public engagement - Establish transparent oversight and accountability | |
Infrastructure | |
- Assess infrastructural and analytic needs - Increase data access - Improve data interoperability - Increase availability of advanced analytic infrastructure and tools   ○ Consider investment in distributed data platforms and cloud computing | |
Workforce | |
- Identify and forecast desired skills and competencies, and review existing skills and capacity - Upskill existing staff   ○ Increase data literacy across the organization, with a focus on bias and equity considerations - Recruit new staff with desired skills - Engage with trainees; consider development of trainee fellowship programs - Foster multidisciplinary collaboration | |
Partnerships | |
- Identify areas where partnerships may be helpful (e.g., gain expertise, obtain or share access to data or infrastructure, engage a wider variety of perspectives) - Consider partnerships with:   ○ Local, provincial/state, federal government   ○ Educational institutions   ○ Private sector | |
Good AI Practices | |
- Default to transparent data and analytic processes and following reproducible and open science principles whenever possible - Ensure access to and use of practical guidelines for AI development, evaluation, and implementation | |
Equity and Bias | |
- Carefully evaluate and assess potential sources of bias throughout development (including sub-group validation) and implementation - Carefully consider potential biases that may exist in the underlying data used to train models - Consider use of an existing ethical AI framework - Foster diverse AI teams - Engage with the community | |