Abortion AI: Toward an equity-centered research agenda for AI and abortion

While the implications of AI for health care are frequently discussed, AI's relationship to abortion has received limited attention. This paper surveys research on AI and abortion to date, adapts AI ethics principles for abortion contexts, and proposes a research agenda for abortion and AI.

We conducted an exploratory literature review, identifying peer-reviewed and grey literature relevant to AI, abortion and/or reproductive health and rights by iteratively searching academic databases. We identified pioneering articles on the implications of AI for reproductive health and two branches of empirical work: AI tools tailored for abortion, and assessments of ChatGPT's responses to questions about abortion. We compared abortion AI tools included in the category 'FemTech' with ChatGPT vis-à-vis their ethical implications for abortion access. Building on this analysis, we propose an Abortion AI research agenda.

The rapid development of AI holds both promise and challenges in sensitive and stigmatized areas like abortion. While existing AI tools allow humans with clinical expertise to review user-facing content for accuracy, concerns have been raised about abortion misinformation propagated by ChatGPT. Likewise, while ChatGPT's data use raises concerns about privacy and consent, precautions taken by specialized abortion AI tools highlight opportunities for harnessing AI while mitigating privacy risks in an era when abortion seekers' digital data has been used in criminal cases against them.

Existing AI applications designed for abortion support privilege contextual appropriateness over technological novelty, avoiding more sophisticated forms of AI when they do not serve patient needs or undermine bodily autonomy. Such choices help mitigate risks of AI misinformation and privacy breach.

With commitments to privacy, accuracy, and meaningful consent, AI can be harnessed to advance abortion access and reproductive autonomy and should be explored with the same rigor as it is in other fields. Community-engaged methodologies are well-positioned to explore appropriateness of AI to community needs.