My research asks who has the standing to name something as a harm in the development and governance of computing systems, and what institutional conditions make that standing real. I work across Human-Computer Interaction and Responsible AI, treating AI as a sociotechnical system. My methods are qualitative and empirical, combining in-depth interviews and fieldwork with design-based research, and I work comparatively, following governance from the settings where it is authored to the settings where it arrives.
Compliance and responsible AI in industry
Responsible AI is usually framed as an ethical project, but inside organizations it increasingly runs on legal logic. Interviewing practitioners working under the EU AI Act, the NIST framework, and related regimes, I found that ethical toolkits are valued for documentation and audit readiness rather than for deliberation. Compliance does not simply support ethics. It decides which ethical concerns count as legitimate in the first place. This work appears at AIES 2026, and the dissertation develops a principle-to-practice matrix connecting ethical and legal requirements to concrete developer workflows.
Ethical responsibility without regulation
If compliance shapes ethics where regulation exists, what happens where it does not? My current fieldwork with AI practitioners in Pakistan examines how they decide what counts as a harm, who they answer to, and what substitutes for regulation when no regulator is watching.
Who gets to evaluate AI
Expert audits miss harms that affected people recognize. With collaborators at UW-Madison, I have studied how non-expert auditors reason about fairness in public-sector AI. We found that they act as normative agents who surface value tensions rather than as error detectors (CSCW 2026).
Recourse and design where institutions are thin
For close to a decade before my PhD, I did applied research in Pakistan, with roughly 200 interviews across projects in financial services, mobile, and health. I studied how victims of mobile fraud sought recourse when formal channels failed (CSCW 2021), and how informal savings groups ran on trust without banks, and what digitizing them would require (CSCW 2019). I contributed to a voice-based platform that let low-literacy users earn through crowdsourcing (CHI 2021), and to research on how religious values shape women's health decisions across Pakistan, Bangladesh, and Malaysia (CHI 2021). Across these projects the lesson was the same: when formal institutions cannot reach people, people do not stop seeking recourse. What fills that gap is the question I now bring to AI governance. Detailed case studies are on the Projects page.
Every profession that handles consequential decisions eventually builds the same infrastructure. Physicians answer for a diagnosis, lawyers for their conduct before a court, pilots for the aircraft and everyone in it, and in each case that accountability belongs to the person rather than to the employer. A pilot's certificate does not come from the airline, which is why a captain can ground a plane. AI has no equivalent. Responsibility dissolves into the organization, and my dissertation shows what follows: practitioners fall back on compliance because they have no standing of their own from which to object.
My next research program asks what professional accountability could look like for the people who build AI systems, and whether it would do what its advocates hope or simply relocate blame. I approach this comparatively, across settings with mature regulatory infrastructure and settings with almost none.
Peer-Reviewed Publications
Corey Jackson, Tallal Ahmad, Devansh Saxena. "Re-imagining Fairness in Machine Learning: A Framework for Building in Socio-cultural and contextual Awareness" Workshop Paper in Supporting User Engagement in Testing, Auditing, and Contesting AI at CSCW 2023. (Presentation)
Industry Reports