Seminar: Privacy Is Not Enough: How Infrastructure Governs Machine Behavior in Decentralized Markets
- Date: Aug 4, 2026
- Time: 02:00 PM - 03:15 PM (Local Time Germany)
- Speaker: Davor Svetinovic (Khalifa University)
- Location: Max Planck Institute for Human Development, Lentzeallee 94, 14195 Berlin
- Room: Room 316 (CHM)
- Host: Center for Humans and Machines
- Topic: Discussion and debate formats, lectures
An autonomous agent can make a sound decision and still produce a poor outcome if the infrastructure delays, reorders, or excludes its action. Decentralized markets make this problem unusually observable because transactions, block-building decisions, and execution outcomes leave public traces. This talk uses that setting to examine how infrastructure governs machine behavior. Drawing on published work in transaction privacy, blockchain censorship, maximal extractable value measurement, and learning-based atomic arbitrage, I develop a layered account of trustworthy autonomy that links privacy, neutral access, and execution-aware decision making. These studies show why privacy alone does not guarantee agency and why evaluations of agent performance must include the conditions under which actions reach the market. I use ongoing research on market agents to consider bounded delegation: agents should estimate execution risk, abstain when uncertainty is too high, and expose the assumptions behind their actions. I close with questions about responsibility, institutional power, strategic adaptation, and the governance of machine behavior in digital environments.
Davor Svetinovic is an Associate Professor of Computer Science and Associate Chair for Graduate Studies at Khalifa University, and a Visiting Fellow at ADIA Lab. His research examines trustworthy AI and decentralized systems, with particular attention to how market infrastructure, privacy mechanisms, and institutional control shape machine behavior. His work includes BaseSAP, empirical studies of blockchain censorship and maximal extractable value, and learning-based methods for atomic arbitrage. He has published more than 100 papers, previously directed the Research Institute for Cryptoeconomics at WU Vienna, and has held visiting and affiliated research roles at MIT. He received his PhD in Computer Science from the University of Waterloo.
Davor Svetinovic is an Associate Professor of Computer Science and Associate Chair for Graduate Studies at Khalifa University, and a Visiting Fellow at ADIA Lab. His research examines trustworthy AI and decentralized systems, with particular attention to how market infrastructure, privacy mechanisms, and institutional control shape machine behavior. His work includes BaseSAP, empirical studies of blockchain censorship and maximal extractable value, and learning-based methods for atomic arbitrage. He has published more than 100 papers, previously directed the Research Institute for Cryptoeconomics at WU Vienna, and has held visiting and affiliated research roles at MIT. He received his PhD in Computer Science from the University of Waterloo.
Alternatively, join online:
Meeting ID: 376 925 894 710 637
Passcode: es6G9Ld7