"Girl, I'm so Serious": CARE, a Capability Framework for Reproductive Equity in Human-AI Interaction
A new framework exposes 'source opacity' and 'response rigidity' as key harms in AI health tools.
A team of researchers, including Alice Zhong, Phoebe Chen, and Anika Sharma, has published a new paper titled "'Girl, I'm so Serious': CARE, a Capability Framework for Reproductive Equity in Human-AI Interaction." The work introduces CARE (Capability Approach for Reproductive Equity), a novel framework designed to evaluate whether AI tools genuinely expand capabilities for sexual and reproductive health (SRH). The framework translates concepts from the Capability Approach—like functionings and conversion factors—into a practical 'Normative Design Lens' for building equitable AI and an 'Evaluation Lens' for auditing existing systems.
Applying their framework, the researchers evaluated a range of AI tools, including SRH-specific non-LLM chatbots, general-use large language models (LLMs), and search engine features. Their analysis identified two critical 'epistemic harms' prevalent in current systems: 'source opacity,' where the origins of health information are unclear, and 'response rigidity,' where AI fails to adapt to diverse user contexts and needs. The paper concludes by moving beyond diagnosis to offer actionable recommendations for AI design, participatory auditing strategies involving affected communities, and policy implications for regulating AI in other high-stakes domains where technology intersects with structural inequity.
- Introduces the CARE framework with a Normative Design Lens and Evaluation Lens for auditing AI equity.
- Identifies 'source opacity' and 'response rigidity' as two key epistemic harms in SRH chatbots and LLMs.
- Provides concrete design recommendations and participatory auditing strategies for developers and policymakers.
Why It Matters
Provides a critical tool for ensuring AI health tools don't perpetuate existing inequalities, moving beyond simple access to meaningful support.