The MPI-IS Director has received the prestigious award for his co-authored paper which laid the foundations of adaptive data analysis, revealing a deep link between differential privacy and statistical validity.
We are thrilled to announce that Moritz Hardt, Max Planck Institute for Intelligent Systems Director and Head of the Social Foundations of Computation Department, has been recognized with the 10-year Test of Time Award at the 57th ACM Symposium on Theory of Computing (STOC 2025) in Prague. The award celebrates enduring research that has significantly shaped the field over the past decade.
Hardt, who also serves on the faculty and board of IMPRS-IS and will be a keynote speaker at our upcoming 2025 IMPRS-IS Boot Camp, received the award for his landmark paper “Preserving Statistical Validity in Adaptive Data Analysis”, co-authored with Cynthia Dwork, Vitaly Feldman, Toniann Pitassi, Omer Reingold, and Aaron Roth. Originally published in 2015, the paper revealed that standard assumptions in statistical analysis fail in adaptive workflows—where each decision is influenced by previous data exploration.
The work introduced a formal model of adaptive data analysis and demonstrated that classical statistical tools break down when used in such iterative processes. Most significantly, it showed that differentially private methods maintain statistical validity in adaptive settings, bridging two previously separate areas: data privacy and statistical inference.
This insight has since inspired an entire body of follow-up research and continues to inform practical methods in data science and machine learning today. Moritz Hardt’s contributions exemplify the kind of foundational thinking that IMPRS-IS strives to nurture—combining deep theoretical insight with real-world relevance.
More information