VVZ API is not affiliated with ETH Zurich. Data might be outdated or incorrect. Please view the official ETHZ Vorlesungsverzeichnis for binding information.
Machine Learning Applications and Society: Interpretability, Explanations and Trust
Last Updated: 2026-02-05 16:07:01
Abstract
Machine learning models are widely used in multiple sectors of society (e.g., healthcare, financial services, job-markets and judicial system). The research domain of interpretable machine learning (iML) aims at designing and testing methods that allow users to understand machine learning models and their outcomes, assessing and managing the risks stemming from their use.
Objective
The seminar familiarizes students with advanced and recent ideas from the interpretable machine learning (iML) literature and relevant applications from the human computer interaction (HCI) research domain. The students will have to be critically review, contextualize, and present original scientific papers; they will test the interpretable machine learning methods presented in the papers on selected datasets and critically review results. The students will learn how to 1) structure a scientific review of research papers from the iML and HCI literature, 2) implement selected interpretable machine learning methods in Python or R, 3) analyse their points of strengths and limitations, 4) prepare, structure and conduct a scientific presentation in English which covers the key findings of their reviews.
Content
The seminar will cover a number of cutting edge research papers which have emerged as important contributions in the interpretable machine learning research domain. The methods therein presented are becoming a standard in industry, where practitioners apply them to data science projects. The seminar is interdisciplinary: it comprises a theoretical and practical part. The theoretical part of the seminar is focused on the emergence of the concept of interpretability of machine learning models, together with its motivation, definition and impact on different sectors of society (e.g., healthcare and insurance). The practical part of the seminar is centred on the overview and analysis of post-hoc interpretability methods of machine learning models—i.e., explanations—such as counterfactual explanations, Local Interpretable Model-agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP), and the effects of the use of interpretability methods on users' trust. This is discussed and measured in empirical studies from very recent HCI literature. The research papers will be introduced and allocated in the first sessions of the seminar. During the semester, selected guest speakers from academia and industry will give presentations on topics of relevance for the seminar. The seminar is open to all MSc and PhD students with an interest in machine learning, interpretability of machine learning models, and users' trust in applications that use machine learning methodologies. As prerequisite it is required a good knowledge of machine learning, with a focus on supervised learning, together with a good experience in using Python or R for machine learning modelling.
General Information
- Language
- English
- Levels
- MSC
- Frequency
- Yearly recurring
Examination
- Type
- ungraded semester performance
Registration & Places
- Max Places
- 20
- Signup End
- 20.02.2022
Course Components
| Type | Title | Time & Place | Hours |
|---|---|---|---|
| lecture | Machine Learning Applications and Society: Interpretability, Explanations and Trust |
|
2 h weekly |