Research events
Overflowing the Black Mediterranean. Creations, cultures and identities across the Euro-African borders
from May 28, 2026 to May 29, 2026
Research events
Published on May 12, 2026– Updated on May 12, 2026
Spring School on Explainable AI: XAIDATA
Location: SHS Auditorium and room MZ03, Rue des Chênes Pourpres, 95000 Cergy
Summer school coordinated by Vassilis Christophides from ETIS research center.
AI systems increasingly influence decisions in science, policy, and daily life. Understanding the data and its relationship to trained models is essential for building safe, reliable, and compliant AI systems across diverse applications, as all model decisions are rooted in training data. To this end three complementary families of interpretability methods have been proposed to shed light on data-intensive automated decision systems: (a) Explainable AI focusing on feature attribution to understand which input features drive model decisions; (b) Data-Centric AI emphasizing data attribution to analyze how training examples shape model behavior; (c) Functional Interpretability examining component attribution to understand how internal model components contribute to outputs. Different interpretability methods are currently used by different tasks of modern pipelines required to build modern AI systems. The proposed ETIS Spring School on the Explainability of Data Intensive AI Systems aims to bring together researchers and students from data management, artificial intelligence, and responsible computing to explore how transparency and interpretability can be effectively integrated into data driven environments. The workshop will investigate how explainability can provide actionable insights in different learning settings as Recommendation Rankings, Time-to-Event predictions, Graph-based Classification or Regression, Retrieval Augmented Generation (RAG) pipelines, Causal Feature Selection and Queries over Inconsistent Data.
More info
AI systems increasingly influence decisions in science, policy, and daily life. Understanding the data and its relationship to trained models is essential for building safe, reliable, and compliant AI systems across diverse applications, as all model decisions are rooted in training data. To this end three complementary families of interpretability methods have been proposed to shed light on data-intensive automated decision systems: (a) Explainable AI focusing on feature attribution to understand which input features drive model decisions; (b) Data-Centric AI emphasizing data attribution to analyze how training examples shape model behavior; (c) Functional Interpretability examining component attribution to understand how internal model components contribute to outputs. Different interpretability methods are currently used by different tasks of modern pipelines required to build modern AI systems. The proposed ETIS Spring School on the Explainability of Data Intensive AI Systems aims to bring together researchers and students from data management, artificial intelligence, and responsible computing to explore how transparency and interpretability can be effectively integrated into data driven environments. The workshop will investigate how explainability can provide actionable insights in different learning settings as Recommendation Rankings, Time-to-Event predictions, Graph-based Classification or Regression, Retrieval Augmented Generation (RAG) pipelines, Causal Feature Selection and Queries over Inconsistent Data.
More info
