'OpenBloom': A Stigma-Sensitive LLM Design Probe for Navigating Reproductive Well-being Conversations with Young Adults

LLMs can support reproductive health conversations but risk oversimplifying values: The study found that while OpenBloom's outputs were perceived as "safe" and non-offensive, they "tend to paraphrase or rely on factual recall, which may lead to value dilution"—meaning the system may strip away important cultural and personal values that matter in sensitive health discussions.

Young adults are already turning to AI for reproductive health information: The research documents that young adults are increasingly using LLMs to seek sensitive health information, highlighting an urgent need for responsible AI design in this domain rather than leaving the space unaddressed.

Feminist design principles matter for health AI: The study applied Feminist HCI principles to develop OpenBloom, suggesting that intentionally centering diverse perspectives and values—rather than treating health information as purely neutral—is important for responsible reproductive health technologies.

"Safety" alone is insufficient for responsible AI: The finding that outputs were perceived as safe but potentially value-diluted reveals that avoiding offense is not enough; responsible AI systems must actively preserve and respect the cultural and personal values embedded in sensitive topics like reproductive well-being.

Contestability should be built into health AI systems: The researchers recommend future reproductive health technologies adopt "contestability and value-sensitive frameworks," meaning users should be able to question, challenge, and understand the values embedded in AI-generated health guidance.