Sultan Mehmood, Shaheen Naseer and Daniel Chen

Mehmood, S., Naseer, S. and Chen, D. (2024) AI Education as State Capacity: Experimental Evidence from Pakistan. June. Available at: https://users.nber.org/~dlchen/papers/Transmitting_AI_Training.pdf

In an era increasingly defined by digital transformation and the burgeoning influence of AI in society, the potential of AI education in the public sector becomes critical. Assessing the consequences of knowledge and receptiveness of AI is essential for understanding the broader implications of this technological paradigm shift. In this paper, we randomize elite bureaucrats in Pakistan, a country of 240 million people, into AI educational workshops and find that it influences policymakers’ attitudes to adopt AI and their policymaking, such as funding allocations toward digitization. Cross-randomizing them into AI fairness activism reduced this willingness and funding. To capture downstream impacts from top public sector managers to the population, we utilized a digital democracy platform, where we observed that AI training enhanced citizen’s ratings of the civil servants’ efforts, particularly for land disputes, while AI fairness activism diminished them. Overall, our research shows that top government officials' human capital is malleable and affects their attitudes, policies, and even the broader population’s perception of public service delivery.