Spring 2026

See our homepage for future talks. Recordings of all past webinars are available on our YouTube channel.

• Tuesday, May 19, 2026:
  - Speaker: Naoki Egami (Massachusetts Institute of Technology)
  - Details: Zoom link, Meeting ID: 968 8371 7451, Passcode: 414559
  - Title: Conformal Policy Learning with Distribution-Free Safety Guarantees: Application to AI-Powered Interventions
  - Abstract: Generative AI is emerging as a new class of intervention in the social sciences, with applications designed to change attitudes and behaviors through scalable, personalized interactions. For example, conversational agents have been used to reduce political polarization and improve workplace productivity. At the same time, recent empirical studies highlight an important risk: while such interventions may benefit many individuals and tasks, they may also harm others. How, then, can AI interventions be deployed safely?
  In this paper, we develop a new statistical framework, conformal policy learning, to deliver pre-specified safety guarantees when deciding whether individuals should receive a new intervention or the status quo. For instance, a researcher may require that the probability that an individual is harmed by the chosen intervention is below 1%. Using tailored conformal hypothesis testing, our method provides finite-sample safety guarantees under the standard exchangeability assumption, without relying on any modeling assumptions. It also achieves asymptotically optimal power or welfare maximization when the conditional expectation functions of outcomes are correctly specified. Thus, our treatment assignment rule is guaranteed to be safe in finite samples while attaining optimality under standard modeling assumptions. In practice, our framework enables researchers to deploy AI safely by assigning AI interventions only to people and tasks that satisfy user-specified safety requirements, and by reverting to the status quo otherwise. This offers a middle ground between two undesirable extremes: unfiltered deployment that ignores AI risks and total avoidance due to safety concerns. We illustrate the method through extensive simulations and an experiment in which randomly assigned AI chatbots are used to reduce conspiracy beliefs. This is joint work with Ying Jin.
  - Discussant: Eli Ben-Michael (Carnegie Mellon University)
  [Slides][Video]

• Tuesday, May 12, 2026:  OCIS+INI joint webinar
  - Speaker: Fan Xia (University of California, San Francisco) & Gary Chan (University of Washington)
  - Time: This event starts at 8:30 am PT/ 11:30 am ET/ 4:30 pm London time/ 11:30 pm Beijing time
  - Zoom details: Link to join,  Meeting ID: 819 2387 7168​, Passcode: Newton1
  - Title: Robust and Efficient Semiparametric Inference for the Stepped Wedge Design
  - Abstract: Stepped wedge designs (SWDs) are increasingly used to evaluate longitudinal cluster-level interventions but pose substantial challenges for valid inference. Because crossover times are randomized, intervention effects are intrinsically confounded with secular time trends, while heterogeneity across clusters, complex correlation structures, baseline covariate imbalances, and small numbers of clusters further complicate inference. We propose a unified semiparametric framework for estimating possibly time-varying intervention effects in SWDs. Under a semiparametric model on treatment contrast, we develop a nonstandard semiparametric efficiency theory that accommodates correlated observations within clusters, varying cluster-period sizes, and weakly dependent treatment assignments. The resulting estimator is consistent and asymptotically normal even under misspecified covariance structure and control cluster-period means, and is efficient when both are correctly specified. To enable inference with few clusters, we exploit the permutation structure of treatment assignment to propose a standard error estimator that reflects finite-sample variability, with a leave-one-out correction to reduce plug-in bias. The framework also allows seamless incorporation of adjustment for imbalanced baseline precision variables through a design-based adjustment shown to be closely related to post-stratification, or a double adjustment that additionally incorporates an outcome-based component.  Simulations and application to a public health trial demonstrate the robustness and efficiency of the proposed method relative to standard approaches.

• Tuesday, May 05, 2026: OCIS+INI joint webinar
  - Speaker: Chan Park (University of Illinois Urbana-Champaign)
  - Time: This event starts at 8:30 am PT/ 11:30 am ET/ 4:30 pm London time/ 11:30 pm Beijing time
  - Zoom details: Link to join,  Meeting ID: 819 2387 7168​, Passcode: Newton1
  - Title: Distributional Balancing for Causal Inference: A Unified Framework via Characteristic Function Distance
  - Abstract: Weighting methods are essential tools for estimating causal effects in observational studies, with the goal of balancing pre-treatment covariates across treatment groups. Traditional approaches pursue this objective indirectly, for example, via inverse propensity score weighting or by matching a finite number of covariate moments, and therefore do not guarantee balance of the full joint covariate distributions. Recently, distributional balancing methods have emerged as robust, nonparametric alternatives that directly target alignment of entire covariate distributions, but they lack a unified framework, formal theoretical guarantees, and valid inferential procedures. We introduce a unified framework for nonparametric distributional balancing based on the characteristic function distance (CFD) and show that widely used discrepancy measures, including the maximum mean discrepancy and energy distance, arise as special cases. Our theoretical analysis establishes conditions under which the resulting CFD-based weighting estimator achieves root-N consistency. Since the standard bootstrap may fail for this estimator, we propose subsampling as a valid alternative for inference. We further extend our approach to an instrumental variable setting to address potential unmeasured confounding. Finally, we evaluate the performance of our method through simulation studies and a real-world application, where the proposed estimator performs well and exhibits results consistent with our theoretical predictions.
  [Paper][Slides][Video]

• Tuesday, Apr 28, 2026:  OCIS+INI joint webinar
  - Time: This event starts at 8:30 am PT/ 11:30 am ET/ 4:30 pm London time/ 11:30 pm Beijing time
  - Zoom details: Link to join,  Meeting ID: 819 2387 7168​, Passcode: Newton1
  - Speaker: Bryon Aragam  (UChicago)
  - Title: Beyond identifiability in causal representation learning
  - Abstract: Causal reasoning has long been recognized as a crucial skill needed to build intelligent systems. Whether or not current systems possess this skill is the subject of much debate: Recent years have witnessed a flurry of activity with both positive and negative results on this topic from both theoretical and empirical perspectives. This talk will highlight the challenges intrinsic to this endeavour, focusing on the difficulties in translating existing causal identifiability results into practical, finite-sample algorithms. We will focus on two concrete subproblems in causal representation learning, namely neighbourhood selection and factor modeling, and present recent progress towards resolving these challenges.
  [Video]

• Tuesday, Apr 21, 2026:  OCIS+INI joint webinar
  - Time: 8 am PT / 11 am ET / 4 pm London time / 11 pm Beijing time (EALIER than usual)
  - Zoom details: Link to join, Meeting ID: 838 9026 8353, Passcode: Newton1
  - Panel discussion (joint with Isaac Newton Institute)
  - Panelists:
  Els Goetghebeur (Universiteit Gent);
  Jessica L. Rohmann (Berlin Institute of Health at Charité);
  Jessica Young (Harvard University); 
  Jonathan Sterne (University of Bristol).

• Tuesday, Apr 14, 2026:
  - Speaker: Matteo Bonvini (Rutgers University)
  - Zoom details: Zoom link (webinar ID: 968 8371 7451, password: 414559)  
  - Title: Causal Estimation and Inference under Slow Nuisance Rates
  - Abstract: Estimators of causal functionals based on first-order influence functions have been extremely successful in modern causal inference. Grounded in semiparametric efficiency theory, their estimation error depends on nuisance functions only through a second-order remainder term, enabling the use of flexible machine learning methods. Inference is typically justified by assuming that this remainder is asymptotically negligible. However, in high-dimensional or low-smoothness settings, nuisance estimators may converge too slowly for this condition to hold. In this talk, we introduce three works that aim to move beyond first-order estimation and inference. In particular, we discuss (i) a higher-order estimator of continuous treatment effects (dose-response) that achieves the fastest known convergence rate in Hölder smoothness models, (ii) a second-order estimator of the average treatment effect that combines structure-agnostic estimation with smoothness and can be used for doubly-robust inference, and (iii) an inferential method for functionals based on artificial noise injection and filtering that enables valid inference when the second-order remainder is not asymptotically negligible.
  - Discussant: Lars van der Laan (University of Washington)
  [Paper 1, Paper 2, Paper 3][Slides][Discussion slides]

• Tuesday, Apr 07, 2026:
  - Speaker: Thomas Icard (Stanford University)
  - Zoom details: Zoom link (webinar ID: 968 8371 7451, password: 414559)  
  - Title: Causal Inference as a Logical Problem
  - Abstract: The goal of this talk will be to show how problems of causal inference can be usefully and precisely understood as logical problems. Adapting tools and concepts from mathematical and computational logic affords new perspectives, raises new questions, and sheds light on some practical and theoretical issues in causal inference. We illustrate with several examples, including some ways in which a logical lens can help clarify the empirical status of assumptions sufficient to bridge gaps between limited data and substantive causal conclusions.
  - Discussant: Jiji Zhang (Chinese University of Hong Kong)
  [Slides][Video]