Agentic AI · Financial technology · Institutions
The interesting part is what institutions have to live with.
I work on agentic AI and financial technology — how these systems actually behave once a firm, a market, or a regulator has to depend on them. Working papers, field notes, and the occasional prototype. Start with the work ↓
Publications
Agent-based modeling of blockchain designs for interbank, large-value payment systems. 24th International Conference on Autonomous Agents and Multiagent Systems, Detroit.
The open-source Python simulator behind the AAMAS work, built for modularity, accessibility, and data privacy in payment-system research. With Kenneth See and Hanzholah Shobri.
Text-mining forum and broadcast data to map the stereotypes attached to gangtaiqiang, a mediatized variety of Taiwanese Mandarin. With Chun-Yi Peng, in Open Linguistics 6(1).
How the ideological commitments behind a system shape the governance that gets built into it. Academy of Management Annual Meeting proceedings.
Working papers & projects
What happens to governance when prompt rules, retrieval stores, and model weights end up owned by different parties — and why observability and repair rights matter more than they look.
A simulated exchange run as a behavioral-finance lab: overconfidence, loss aversion, and the disposition effect observed across student trading contests in Singapore, Brazil, and the United States.
Shorter pieces on where AI actually belongs inside a firm, and what a founder can decide this quarter rather than admire in a demo.
About
Nicholas MacGregor Garcia is a Lecturer in Information Systems at the National University of Singapore and Deputy Director of the NUS FinTech Lab, where he leads industry relations. He teaches financial technology, AI governance, and blockchain, and holds a PhD in Information Systems from NYU Stern.
His research looks at how organizations actually govern and deploy AI and financial systems — from agent-based simulations of payment infrastructure to the question of who answers for a capable system when it gets something wrong. Most of the material here is live rather than finished. If something is useful or off, find me on GitHub.