"AI in Global Health: The View from the Front Lines"
by Ismail, Azra; Kumar, Neha (2021)
There has been growing interest in the application of AI for Social Good, motivated by scarce and unequal resources globally. We focus on the case of AI in frontline health, a Social Good domain that is increasingly a topic of significant attention. We offer a thematic discourse analysis of scientific and grey literature to identify prominent applications of AI in frontline health, motivations driving this work, stakeholders involved, and levels of engagement with the local context. We then uncover design considerations for these systems, drawing from data from three years of ethnographic fieldwork with women frontline health workers and women from marginalized communities in Delhi (India). Finally, we outline an agenda for AI systems that target Social Good, drawing from literature on HCI4D, post-development critique, and transnational feminist theory. Our paper thus offers a critical and ethnographic perspective to inform the design of AI systems that target social impact.
In the push for more data, the role of the humans and the work they do are routinely marginalized, even as they provide critical linkages to make the data/AI infrastructures work. HCI and related disciplines have invested consistently in making work visible. This body of work is wide-ranging, from Suchman’s and Gray and Suri’s investigation of invisible work to Sambasivan and Smyth’s description of the “human infrastructure of ICTD” Though not integrated into the digital economy quite yet, there are similar risks to extracting (invisible) labor from the FHWs who are already overburdened on account of responsibilities touching diverse, overlapping aspects of their lives. Our analysis detailed the limited compensation they receive for their ever-growing list of responsibilities. Even as FHWs are viewed as actors and changemakers across their communities, they are vulnerable to being taken advantage of for the data collection and scale that AI systems target. We recommend that AI systems be designed to foreground user agency—by not forcing workfows and allowing users the autonomy to decide whether to use the suggestions made by an AI system. Care work has also remained consistently undervalued, additionally marginalizing FHWs’ invisible labor, even when they are frequently expected to provide the services that generally lie within the purview of trained (and far better paid) doctors. AI systems may reinforce gendered and racialized notions of devalued work. A key refex we must cultivate is to always ask whose social good we are attempting, and who (not what) we are centering. (p.13)
KeywordsArtificial Intelligence, Healthcare, Sustainability, Social Good, Care Work
ThemesAI and Medicine, Automation
Links to Reference
How to contribute.