Summer School 2026

Courses

Registration is now closed, as all courses are full. Below is an overview of the courses offered during NORA Summer School 2026.

1. Causality and Machine Learning (full course)

Credits: 5

Venue: UiO, Georg Morgenstiernes hus, Blindernveien 31
Room: 152

Lecturer(s): Pekka Parviainen (University of Bergen), Johan Pensar (UiO), Fabio Massimo Zennaro (University of Bergen), Johan de Aguas (UiO)

Course Description: 
Current artificial intelligence (AI) and machine learning (ML) models work only on a probabilistic level evaluating associations between variables; causal reasoning could allow us to develop systems that are able to reason and that can be controlled in a more grounded, reliable and interpretable way. The course will introduce students to the theory of causality developed in the past decades, as well as to the methodologies devised to perform valid and sound causal inferences. It will start by presenting the graphical and probabilistic formalism used to represent causal systems; it will discuss principled ways to estimate causal effects relying on observational data and graphical structures; it will delve into the problem of discovering graphical structures from data; and, finally, it will review modern intersections between causality and machine learning. These topics will provide a solid understanding of the concepts related to causality, allowing the students to apply such ideas to their own domain and research. Students are expected to have good mathematical skills and knowledge in the fields of machine learning, probability and statistics.

2. The perception of human language in LLMs: Critical perspectives on language, bias, culture, politics and ethics in Large Language Models and the implementation of these (full course)

Credits: 5

Venue: UiO, Georg Morgenstiernes hus, Blindernveien 31
Room: 113

Lecturer(s): Jessica Pedersen Belisle Hansen (University of Østfold), Samia Touileb (University of Bergen), Leonora Onarheim Bergsjø (Østfold University College), Ida Jahr (University of Inland Norway) and Anja Salzmann (University of Bergen)

Course Description: 
Language is inherently human, a fundament for our thinking, our consciousness and the way we understand ourselves and the world around us. Language enables us to build social relationships, to express emotions, cultural identity and is a carrier of our collective memory. With the introduction of Large Language Models (LLMs) we are confronted with the fact that machines produce linguistic output that simulates human reasoning in a very convincing manner, that enables para-social interactions and relationships (between humans and machines) and the manufacturing of authenticity. LLMs are now being implemented in various contexts with the intention of providing end users with access to a multitude of content and services, e. g. modelling linguistic practices (e.g. for job applications and school essays), improvement of their personal human language (e.g. revision of texts), automized frontline service provision (such as public service chatbots) but also more advanced forms ) that are trained on a particular domain or selected datasets (e.g. virtual assistants/agents, research assistance).

This is an invitation to learn, reflect and discuss together about issues that arise in critical debates on the societal introduction and consequences of LLMs from transdisciplinary (linguistics and social interaction, language technology, critical theory, sociology, media sciences, ethics, cultural studies, histories of knowledge) perspectives and viewpoint. Looking at several examples, we will shed light on key issues and themes that arise from using LLMs in different contexts and look more closely at frameworks, guidelines and process-oriented approaches that can be applied to aid responsible development and implementation of large language models. This PhD course is interdisciplinary and invites students from a wide range of backgrounds to participate.

3. Deep neural networks for business and decision making (full course) 

Credits: 5

Venue: UiO, Georg Morgenstiernes hus, Blindernveien 31
Room: 144

Lecturer(s): Fred Espen Benth (BI Norwegian Business School)

Course Description: 
The main objectives of the course are to present a carefully selected portfolio of case studies for discussing and analysing different features and use of deep neural networks (DNNs) in business and decision making. Core use areas in the course include economics, energy and finance, where we present theoretical considerations through case studies. We focus on the following themes: the use of synthetic data in training DNN approximations of expectation functionals and partial differential equations; operator-learning along with structural information to construct efficient and explainable networks; probabilistic forecasting to quantify uncertainty in predictions using distributional DNNs; signature transform and classification. Throughout, we pay attention to the theoretical basis and practical limitations of modern use and developments of DNNs, and reflect on challenges such as explainability and energy use of modern AI.

By active learning and group-based discussions, the student gets familiar with the central methods and tools in modern use and developments of DNNs. The cases we work with includes high-dimensional partial differential equations modeling heat transport and financial options, electricity prices and risk diversification in investments.

Expected skills/background:

The course requires knowledge in calculus, linear algebra and probability theory. It is advantageous to have some familiarity with (partial) differential equations, function spaces (Hilbert spaces), Brownian motion and diffusion processes.

4. AI Adoption, Compliance and Societal Change (full course)

Credits: 5

Venue: UiO, Georg Morgenstiernes hus, Blindernveien 31
Room: 207

Lecturer(s): Halvdan Haugsbakken (Østfold University College)

Course Description: 
The PhD course AI Adoption, Compliance and Societal Change will be one of the five courses offered at NORA AI Research School Summer School 2026. Our course addresses the adoption and implementation of AI and relates them to questions and their consequences for organizations and society. An important question we address concerns ethical, regulatory, and sovereignty-related aspects. While much current AI research focuses on technical design or model performance, our course will instead address an important question: how is AI integrated into institutional and organizational structures, how do they challenge notions of compliance, control, and human understanding and learning, and how these processes contribute to broader societal transformations?

In our cross-disciplinary course, we will, among others, use a diversity of theoretical perspectives from sociology, science and technology studies (STS), institutional theory, innovation theory, ethical theory, and pedagogical theory and research. You, as a PhD student, will be equipped with analytical tools to understand the implications on human, organizational, and normative dimensions of AI. Pedagogical theory is used here not in a narrow educational sense, but as a framework for understanding learning processes, adaptation, and sensemaking in human-AI interaction and organizational change. The course combines lectures, seminars, workshops, and case-based discussions, and participants will actively connect course perspectives to their own PhD projects.

5. Biologically Inspired Learning in Spiking Neural Networks and Neuromorphic Systems (full course)

Credits: 5

Venue: UiO, Georg Morgenstiernes hus, Blindernveien 31
Room: 141

Lecturer(s): Hans Ekkehard Plesser (Norwegian University of Life Sciences) and Alexander Johannes Stasik (NMBU & SINTEF Digital)
External lecturers:
Agnes Korcsak-Gorzo, Research Centre Jülich
Julian Göltz, University of Bern / University of Heidelberg

Course Description: 
This course provides an in-depth exploration of spiking neural networks (SNNs) and the biologically inspired learning mechanisms that enable them to operate efficiently on neuromorphic hardware. The course bridges neuroscience-inspired computation with modern artificial intelligence, focusing on learning rules beyond traditional backpropagation. After an introduction to fundamental neuron models and spike-based encoding schemes, we will examine the challenges of learning in SNNs, including the non-differentiability of spikes and the locality constraints inherent to biological systems. We will then introduce you to modern learning approaches including surrogate gradient methods, event-based delay and weight training, generalized latent equilibria, time-to-first-spike learning, and eligibility propagation (e-prop). Hands-on sessions will allow participants to implement and train SNNs using frameworks such as Norse, snnTorch, and NEST, using datasets such as N-MNIST and DVS Gesture.

To benefit from this course, you should have a solid mathematical background in vector calculus and linear algebra and programming skills in Python.