Applying socio-anthropological approaches to explore the socio-cultural and ethical requirements for AI-assisted obstetric ultrasound in Kenya: reflections and lessons learned

Pregnancy is a time of profound hope and vulnerability, where the simple act of seeing a developing baby through an ultrasound scan can transform anxiety into reassurance. In many parts of the world, this technology is a routine part of prenatal care, offering a clear window into the health of the mother and child. However, in low-resource settings, access to skilled sonographers and expensive machines remains scarce, leaving many women without this critical safety net. Artificial intelligence, the field of computer science dedicated to creating systems capable of performing tasks that usually require human intelligence, has emerged as a potential solution. By training computers to read medical images, researchers hope to empower less-specialized health workers to provide high-quality care. Yet, a machine that works perfectly in a laboratory does not automatically succeed in a rural clinic. Technology does not exist in a vacuum; it lands in communities with their own deep-rooted beliefs, social structures, and ethical concerns. If a new tool ignores these human realities, it risks being rejected, no matter how advanced its code.

This is the central challenge explored by a team of researchers in Kenya, who set out to understand how a specific type of artificial intelligence—designed to assist with obstetric ultrasound—would be received by the people it aims to help. The project, known as AIMIX, seeks to develop a framework that allows less-specialized providers to perform scans with the help of AI, potentially bridging the gap in maternal healthcare. But before the technology could be built and tested, the team realized they needed to understand the human landscape first. They turned to socio-anthropological methods, a way of studying people by listening to their stories and observing their lives, to explore the social and cultural rules that govern how new medical tools are accepted. Their work, detailed in a recent reflection paper, reveals that the path to successful technology is paved not just with data, but with trust, clear communication, and a deep respect for local context.

The researchers conducted their study in two very different settings: the bustling, urban environment of a private hospital in Nairobi and the rural, subsistence-farming community of Rabai on the Kenyan coast. They spoke with eighty-four people in total, including pregnant women, their partners, doctors, nurses, community leaders, and even traditional birth attendants. Using a method called grounded theory, they did not start with a fixed list of questions to prove a specific point. Instead, they began with open-ended conversations, allowing the participants to guide the discussion and reveal what truly mattered to them. This approach let the researchers discover issues they had not anticipated, such as the specific fears about data privacy or the unique ways religious leaders interpreted the role of machines in human life.

One of the most significant findings was that the concept of artificial intelligence was largely abstract and confusing to many participants. In the rural areas, where digital technology is less common, the idea of a computer analyzing a baby's image felt distant and potentially threatening. The team learned that simply explaining the technology with technical terms was ineffective. Instead, they found success by using visual aids, such as a simple drawing showing a nurse performing a scan with a new device, to make the concept concrete. They also discovered that understanding was not a one-time event; they had to check in with participants, asking them to explain the study back to the researchers in their own words. This simple practice of verification ensured that everyone was on the same page before moving forward, preventing misunderstandings that could have derailed the entire project.

The study also highlighted the critical importance of who gets to speak about new technology. In many previous efforts to introduce medical innovations, the conversation was dominated by doctors and engineers, while the people who would actually use or receive the care were left on the sidelines. The Kenyan team deliberately included a wide range of voices, from community health promoters to religious leaders and traditional birth attendants. They found that these grassroots stakeholders held the keys to trust. For instance, when a religious leader expressed concern that the Catholic Church might oppose the technology, the team did not dismiss the worry. Instead, they sought out a priest who clarified that the Church supported the technology as long as it upheld human dignity. Similarly, when a community leader doubted whether traditional birth attendants would accept the new tools, the team interviewed the attendants themselves and found they were open to the idea if it meant saving lives. These interactions showed that successful implementation requires engaging the entire social ecosystem, not just the medical system.

Another crucial lesson emerged regarding how the research was conducted. The team initially planned to interview everyone in person, but they quickly realized that in the busy urban hospital, strict face-to-face meetings excluded some willing participants who could not spare the time. They adapted their approach, allowing for virtual interviews to ensure no one was left out. This flexibility extended to their recruitment strategies as well. In the rural area, they struggled to gather groups for discussion until they partnered with a trusted community health promoter and scheduled the meeting to coincide with a routine maternity clinic day. By weaving their research into the existing rhythms of daily life, rather than trying to impose a new schedule, they were able to build genuine participation.

The researchers also emphasized the value of continuous engagement. They did not treat the community as a source of data to be extracted and then forgotten. Instead, they held meetings to update participants on the progress of the study and explain how their input was shaping the development of the AI model. This ongoing dialogue built a foundation of trust that is essential for any long-term partnership. The team noted that while the technology itself is still in development and has not yet been tested in the real world, the insights gained from these conversations are already shaping how the tool is being designed. They argue that without this deep, human-centered understanding, even the most sophisticated algorithm could fail to find a place in the communities that need it most.

Ultimately, this work serves as a reminder that technology is not just a set of tools, but a social practice. The paper suggests that for artificial intelligence to truly benefit maternal health in low-resource settings, it must be built with a profound understanding of the people it serves. The researchers found that the path forward involves more than just better code; it requires a commitment to listening, adapting, and respecting the complex web of cultural and ethical values that define human life. By placing these human factors at the center of the design process, the team offers a blueprint for creating technology that is not only smart but also wise enough to fit into the world it hopes to improve.