The Sabine Krolak-Schwerdt lecture (SKS lecture) is an annual public lecture delivered by a distinguished researcher in the field of data science, AI or statistics.

It commemorates Sabine Krolak-Schwerdt, who served as the founding President of EuADS and sadly passed away in 2017.

SKS lecture 2026

This year will see a special edition as the SKS lecture 2026 will be jointly organised by

• the European Association for Data Science (EuADS), and
• the Global Economic Modeling Network (EcoMod).

It will be scheduled as the closing event of the EuADS Summer School and the opening address of the International Conference on Economic Modeling and Data Science.

We are pleased to announce that the SKS lecture 2026 will be delivered by

Senior Researcher at the National Institute for Research in Digital Science and Technology (Inria), France. A detailed bio can be found on the speaker’s webpage.

The lecture will be entitled

Federated Causal Inference for Policy Evaluation:

From Data to Informed Decisions

Abstract of the lecture. Federated causal inference is an emerging paradigm that integrates federated learning and causal inference to estimate causal effects from decentralised data sources, offering a powerful alternative to traditional meta-analysis. Federated learning enables multiple studies or institutions to collaboratively train models without sharing individual-level data, exchanging only model updates that are securely aggregated by a coordinating server. This decentralised architecture is particularly valuable in contexts where strong incentives exist to keep data on-site, such as compliance with data protection regulations, preservation of data ownership and governance, and prevention of unwanted knowledge transfer.

However, classical federated learning algorithms are primarily designed for predictive modeling rather than causal estimation. In this talk, we will present recent advances in federated causal inference, showing how causal effects can be estimated and evidence aggregated across multiple experimental and observational studies while preserving privacy and statistical efficiency. We will then address the problem of transportability and generalisation of causal evidence from experimental studies to target populations subject to distributional shifts. This perspective directly tackles one of the main limitations of experimental research: the lack of representativeness of study populations. While most existing approaches focus on average treatment effects expressed through risk differences, they overlook the rich diversity of causal estimands relevant for heterogeneous outcomes and decision contexts.

To overcome this limitation, we introduce a unified framework for transporting a broad class of causal effect measures under covariate shift. All these methods are emerging as a cornerstone for modern evidence synthesis, beyond the traditional boundaries of meta-analysis.

Practical information

The lecture will take place in Luxembourg on 8 July 2026 and will be followed by a reception. Admission to the lecture is free but registration is required.

More details about timing and venue, as well as a registration form, will be published by 16 March 2026.