Tackling algorithmic bias and promoting transparency in health datasets: the STANDING Together consensus...

Area of Science:

• Health Informatics
• Artificial Intelligence Ethics
• Biomedical Data Science

Background:

• Artificial intelligence (AI) in healthcare risks scaling existing health inequalities due to encoded biases.
• Biased data is a primary source of bias in AI health technologies, potentially exacerbating disparities.
• Lack of transparency regarding health dataset limitations hinders equitable AI development and deployment.

Purpose of the Study:

• To introduce the STANDING Together recommendations for addressing bias in AI health technologies.
• To promote transparency in health datasets and encourage proactive evaluation of their impact across diverse populations.
• To guide dataset curators and users in identifying and mitigating algorithmic biases that could worsen health inequalities.

Main Methods:

• Recommendations developed through a systematic review, stakeholder survey, Delphi approach, public consultation, and international interviews.
• Input gathered from over 350 representatives across 58 countries, including 194 Delphi participants from 25 countries.
• Consensus reached on 29 recommendations presented in two parts: Documentation of Health Datasets and Use of Health Datasets.

Main Results:

• 29 consensus recommendations provide guidance for dataset documentation and use to ensure AI health technology safety and effectiveness.
• Recommendations emphasize transparent communication of data limitations as a valuable asset, not a deficiency.
• Focus on proactive inquiry rather than a checklist approach to encourage critical evaluation of AI health technologies.

Conclusions:

• Adoption of STANDING Together recommendations can mitigate risks of AI perpetuating health inequalities.
• Promoting transparency in health datasets is crucial for developing safe, effective, and equitable AI health technologies.
• Encouraging stakeholders to acknowledge and address data limitations is key to realizing the societal benefits of AI in healthcare.