Semi-structured interviews with 10 ART professionals/regulators and 10 patients in Australian clinics explored perceptions, concerns, and ethical considerations around machine learning (ML) use in embryo assessment and selection during assisted reproduction.
Both professionals and patients expressed mistrust of ML and stressed the need for human oversight. Key concerns included: accountability gaps in ML-assisted decisions, risk of discarding viable embryos, algorithmic bias, poorly trained models, and potential deskilling of embryologists. Most professionals preferred transparent ML models and called for full clinic disclosure of ML use.
Very small sample (n=20) from Australian clinics only, limiting generalisability. Qualitative design captures perceptions but not clinical outcomes. Regulators and professionals were grouped together, which may obscure differing perspectives.
Clinics using ML tools for embryo selection should proactively disclose this to patients and maintain clear human oversight and accountability structures. Embryologists should be aware of deskilling risks and advocate for transparent, well-validated algorithms.
