This winter, the Big Statistics group of the Epidemiology and Data Science department, Amsterdam University Medical Centers, will host a single-day workshop on explainable machine learning (xML), featuring renowned international speakers.

Date: February 4th 2026
Time: 9:30 – 16:30 (CEST)
Location: IJ-zaal of RDC ADORE building, Van der Boechorststraat 6B, 1007 MB Amsterdam
Registration deadline: January 15th 2026
Maximum capacity: 60 people
Cost: free
Description:
Complex machine learning models are increasingly applied to medical data because of their strong predictive performance and ability to capture interactions and nonlinear relationships. However, their complexity makes it difficult to gain insight in the application at hand. Explainable machine learning (xML) provides a framework of techniques that help researchers (1) understand how models arrive at their predictions and (2) identify which variables are important.

In health care, explainability is particularly relevant. Insight into why predictions are made may generate new biomedical understanding and may ultimately support clinical decision-making. However, the typically small sample sizes in medical studies pose challenges for both model fitting and interpretation.

This workshop is intended to researchers with a methodological/statistical background who are interested in applying state-of-the-art explainable ML methods to health-related data. The program features a mix of theory and applications, with a particular focus on Shapley values and their recent methodological developments. Other xML approaches will also be discussed.

Provisional programme:

09.00h – 09.30h Coffee + Welcome
09.30h – 10.05h Shapley values with Bayesian uncertainty quantification for rule ensembles (Marjolein Fokkema, Leiden University)
10.05h – 10.30h Asymmetric Shapley values to quantify the importance of genes in clinico-genomic applications (Jeroen Goedhart, Amsterdam UMC)
10.30h – 10.55h A principled approach for comparing Variable Importance (Angel Reyero Lobo, Institut de Mathématiques de Toulouse)
10.55h – 11.15h Break
11.15h – 11.50h A causal perspective on Shapley values (Ioan Gabriel Bucur, Radboud University)
11.50h – 12.25h Using Explainable Machine Learning for Assessing Treatment Effect Heterogeneity in Clinical Trials (Konstantinos Sechidis, Novartis)
12.25h – 13.55h Lunch (provided)
13.55h – 14.20h Functional Decomposition of Tree-Based Models (Marvin Wright, Leibniz Institute for Prevention Research and Epidemiology – BIPS)
13.20h – 14.45h Interpretable Machine Learning for Survival Analysis (Sophie Langbein, Leibniz Institute for Prevention Research and Epidemiology – BIPS)
14.45h – 15.30h Software demo: shapr – Conditional Shapley Value Explanation in R and Python (Martin Jullum, Norwegian Computing Center)
15.30h – 15.50h Break
15.50h – 16.25h Is my model perplexed for the right reason? Contrasting LLMs’ Benchmark Behavior with Behavior Specified via Token-Level Perplexity (Giovanni Cinà, Amsterdam UMC)
16.25h – 17.00h Kjersti Aas, Norwegian Computing Center

Specific topics include:
– Shapley values for causal interpretations
– xML methods for tree-based learners and survival analysis
– Computational aspects of Shapley values
– Applications in omics

Organizing Committee: Mark van de Wiel [mark.vdwiel@amsterdamumc.nl], Giorgio Spadaccini [g.spadaccini@amsterdamumc.nl], and Jeroen Goedhart [j.m.goedhart@amsterdamumc.nl]

Sponsors:

-Amsterdam Public Health research institute (APH)
-VVSOR Biometrics / IBS Dutch region (BMS-ANed)
-Dep. Epidemiology & Data Science of Amsterdam UMC

Registration:

Registrations are currently closed as the maximum capacity has been reached. If you want to be put on a waiting list, please fill out the form available at the following link: https://forms.office.com/e/AME0mhwPz2