Want to dive deeper into political forecasting?
Join the ICEUR Community platform.

Community

Dr. Marek J. Druzdzel is a leading expert in Bayesian networks, probabilistic reasoning, and decision support systems, Professor at Białystok University of Technology and Professor Emeritus at the University of Pittsburgh.

For nearly four decades, his research has focused on decision-making under uncertainty, probabilistic inference, and intelligent decision support systems based on probability theory, statistics, and decision theory. A major part of his work has been devoted to Bayesian networks, ranging from their theoretical foundations and inference algorithms to their practical application in modelling complex real-world systems.

Prof. Druzdzel is the principal conceptual designer of GeNIe and SMILE, software tools for building Bayesian networks and graphical decision models. These tools have been developed over decades of academic research and are now used internationally in research, education, and applied decision modelling.

In 1995, he founded the Decision Systems Laboratory at the University of Pittsburgh, which he directed until 2019. The laboratory conducted research on decision analysis, probabilistic modelling, and intelligent decision support systems and played a central role in the development of GeNIe and SMILE.

In 2015, Prof. Druzdzel co-founded BayesFusion, continuing the development of GeNIe and SMILE and bringing Bayesian modelling and decision-support technology to a broad range of practical applications.

He received his PhD in Engineering and Public Policy from Carnegie Mellon University, where his research focused on probabilistic reasoning in decision support systems. His academic background also includes two Master’s degrees from Delft University of Technology and a habilitation from the Institute of Computer Science of the Polish Academy of Sciences.

His research and publications span Bayesian networks, causal reasoning, probabilistic inference, decision theory, knowledge engineering, and decision-making under uncertainty, with applications ranging from medical decision support and diagnosis to risk analysis and complex systems modelling.

Dr. Ludmilla Lobova is a political scientist, Scientific Director and Vice President of ICEUR-Vienna, specializing in political forecasting, international affairs, and the analysis of political processes under uncertainty.

A central focus of her work at ICEUR is the methodology of political forecasting. She examines the limitations of conventional expert analysis, the sources of systematic forecasting errors, and the ways in which probabilistic and Bayesian reasoning can provide a more rigorous framework for dealing with uncertainty. Rather than seeking a single “correct” prediction, this approach emphasizes assigning probabilities to possible outcomes, updating assessments as new information becomes available, and systematically learning from forecasting errors.

Her academic and analytical work also covers international relations, Russian foreign and security policy, political conflicts, and transformation processes across the post-Soviet space.

Dr. Lobova studied law, history, and political science and received her doctoral degree in 1995. In Austria, she was a Lise Meitner Fellow of the Austrian Science Fund (FWF), conducting research on Russian perceptions of Austrian neutrality and its significance for European security.

She has worked with the Ludwig Boltzmann Institute for Research on the Consequences of War, the Austrian State Archives, and the Institute for the Danube Region and Central Europe (IDM). She has also delivered lectures and seminars at the Diplomatic Academy of Vienna and the University of Vienna.

Dr. Lobova is the author of publications on Russian foreign and security policy, European security, political Islam, and ethno-political conflicts. Together with Prof. Hans-Georg Heinrich, she has also served as an editor of the academic series ICEUR Insight Studies.

At the ICEUR School of Political Forecasting, she combines decades of experience in political and regional analysis with Bayesian reasoning, scenario analysis, and probabilistic forecasting, emphasizing an approach in which the quality of a forecast depends not on an expert’s confidence, but on the ability to formulate testable assessments, account explicitly for uncertainty, and revise conclusions as new evidence emerges.

Эта вводная лекция открывает курс ICEUR, посвящённый политическому прогнозированию и анализу нового миропорядка, сформировавшегося в ходе войны в Украине. Лекция объясняет, почему привычные способы понимания мировой политики — линейные прогнозы, интуитивные оценки и экспертные «уверенные заявления» — всё чаще дают сбой. Мир вступил в фазу высокой неопределённости, где ошибки прогнозирования становятся системными, а не случайными.

Особое внимание уделяется различию между прогнозированием, предсказанием и пророчеством, а также причинам, по которым прогнозам не доверяют — от когнитивных искажений до неправильного понимания вероятности. На исторических примерах показывается, почему одни прогнозы оказывались успешными (распад СССР), а другие — провальными.

Вторая часть лекции вводит слушателей в байесовское мышление как основу работы с неопределённостью. Вместо поиска «точных дат» и «окончательных ответов» предлагается подход, основанный на обновлении знаний, вероятностной логике и обучении на ошибках — с целью быть менее неправыми, а не абсолютно правыми.

Лекция также задаёт рамку всего курса: от теоретических оснований к практическому моделированию, сценарному анализу и работе с реальными политическими кейсами.

Prof. Hans-Georg Heinrich is a political scientist, political forecasting expert, and co-founder of ICEUR (Institute for the Comparative Study of European and Eurasian Regions) in Vienna.

His academic and professional career spans several decades of research into political systems, international relations, conflicts, and political transformation, with a particular focus on Eastern Europe, Russia, and the post-Soviet space. A central element of his work is the structured analysis of complex political processes and the development of methods for forecasting under conditions of uncertainty.

At ICEUR, Prof. Heinrich has played a key role in developing and teaching approaches to political forecasting based on probabilistic reasoning and Bayesian networks. His methodology combines expert political analysis with formal modelling techniques, transforming complex relationships between political, economic, and social factors into structured models and scenarios.
Prof. Heinrich previously held a professorship at the University of Vienna and has held visiting teaching and research positions at universities and academic institutions in Russia, Hungary, Poland, Iraq, Egypt, and other countries.

Alongside his academic career, he has extensive experience working in international environments, including assignments with the OSCE in Tbilisi, Chechnya, and Belgrade. This combination of academic research and field experience has informed his approach to analysing political instability, conflict, and complex international developments.

He is the author and editor of numerous academic publications on Soviet and Russian politics, Eastern Europe, political transformation, conflict, and international relations.

At ICEUR, Prof. Heinrich brings together decades of experience in comparative politics and international affairs with Bayesian modelling and structured political forecasting, applying these methods to the analysis of complex political scenarios and the systematic assessment of possible future developments.

Bayesian methods provide a practical framework for reasoning and decision-making in professional environments where information is incomplete, uncertain, or continuously changing. This lecture explores how Bayesian models can be applied across different fields to combine available data, expert knowledge, and assumptions about relationships between relevant factors.

Through examples from risk assessment, policy and political analysis, business and strategic decision-making, engineering, and scientific assessment, participants will examine how different professional questions can be translated into probabilistic models. Particular attention will be given to identifying relevant variables, representing dependencies, incorporating new evidence, and interpreting changes in probabilities as information becomes available.

The lecture will also consider how the same underlying Bayesian principles can be adapted to very different decision-making contexts, demonstrating how Bayesian networks can serve not only as forecasting tools, but also as a structured way to analyse uncertainty, compare scenarios, test assumptions, and support decisions under uncertainty.

Лекция будет вестись на английском языке

Байесовские методы предоставляют практический инструментарий для анализа и принятия решений в профессиональной среде, где информация может быть неполной, неопределенной или постоянно меняться. В этой лекции рассматривается, как байесовские модели могут применяться в различных областях, объединяя доступные данные, экспертные знания и предположения о взаимосвязях между значимыми факторами.

На примерах из оценки рисков, политического анализа, бизнеса и стратегического планирования, инженерии и научной экспертизы участники рассмотрят, как различные профессиональные задачи могут быть представлены в виде вероятностных моделей. Особое внимание будет уделено определению значимых переменных, моделированию зависимостей, включению новой информации и интерпретации изменений вероятностей по мере поступления новых данных.

Лекция также покажет, как одни и те же принципы байесовского подхода могут быть адаптированы к совершенно разным условиям принятия решений и как байесовские сети могут использоваться не только для прогнозирования, но и для анализа неопределенности, сравнения сценариев, проверки предположений и поддержки принятия решений в условиях неопределенности.

Лекция будет вестись на английском языке

Лекция посвящена типичным проблемам и ошибкам, возникающим при построении моделей, а также практическим способам их решения. Участники также познакомятся с QGeNIe, динамическими моделями и функциональными узлами и рассмотрят возможности их применения при разработке более сложных моделей.

This lecture deals with frequent modeling pitfalls and helpful workarounds. It will also cover QGenie, dynamic models and function nodes.
Hands-on Exercises.

his hands-on session focuses on the practical construction, analysis, and interpretation of Bayesian network models using GeNIe 5.0. Participants will work directly with the software, translating real-world problems into structured probabilistic models and exploring how assumptions, dependencies, and evidence affect model outcomes.

The exercises will cover the key stages of model development: defining variables and states, creating dependencies between nodes, specifying probabilities, entering evidence, and analysing posterior probabilities. Participants will experiment with different scenarios and observe how new information propagates through a Bayesian network and changes its conclusions.

Particular attention will be given to the interpretation of model results and practical model refinement. By the end of the session, participants will have worked through the complete process of building and using a Bayesian network in GeNIe 5.0 as a tool for structured analysis and decision-making under uncertainty.

Практическое занятие посвящено построению, анализу и интерпретации моделей байесовских сетей с использованием GeNIe 5.0. Участники будут работать непосредственно с программой, преобразуя реальные задачи в структурированные вероятностные модели и исследуя, как исходные предположения, зависимости и новые данные влияют на результаты моделирования.

Упражнения охватывают основные этапы разработки модели: определение переменных и их состояний, создание зависимостей между узлами, задание вероятностей, ввод свидетельств и анализ апостериорных вероятностей. Участники смогут рассмотреть различные сценарии и проследить, как новая информация распространяется по байесовской сети и изменяет получаемые выводы.

Особое внимание будет уделено интерпретации результатов и практической доработке моделей.

К концу занятия участники пройдут полный цикл построения и использования байесовской сети в GeNIe 5.0 как инструмента структурированного анализа и принятия решений в условиях неопределенности.

Dr. Ullrika Sahlin is an Associate Professor at Lund University, Sweden, specializing in Bayesian modelling, risk analysis, uncertainty quantification, and decision-making under uncertainty.

A central focus of her research is how uncertainty can be represented, quantified, and communicated in scientific assessments and complex decision-making. Her work draws on Bayesian analysis, robust Bayesian methods, expert knowledge elicitation, and evidence synthesis to support reasoning and decision-making when information is incomplete, uncertain, or based on multiple sources.

A particularly important area of her research is the use of Bayesian networks and predictive models. She has studied how Bayesian networks can integrate empirical data, scientific models, and expert judgement while explicitly accounting for different sources of uncertainty. Her research on epistemic uncertainty in Bayesian networks demonstrates how these models can be used for risk assessment and for reasoning about uncertain future events.

Dr. Sahlin has extensive experience working at the interface between academic research and applied risk assessment. She has contributed to the work of the European Food Safety Authority (EFSA) on uncertainty analysis and has led EFSA-related projects focused on the development and application of Expert Knowledge Elicitation methods.

Her interdisciplinary academic background combines mathematical statistics with applied risk and ecological modelling. She holds a Master’s degree in Mathematical Statistics from Lund University, a Master’s degree in Forestry from the Swedish University of Agricultural Sciences, and a PhD in Ecology, where her research focused on risk assessment related to invasive species.

Her current research interests include robust Bayesian analysis, evidence synthesis, expert judgement, risk analysis, and decision-making under uncertainty. She also teaches doctoral-level courses in Bayesian analysis and decision theory, as well as risk, uncertainty, and decision-making.

This lecture examines the different types and sources of uncertainty encountered in Bayesian network (BN) modelling. It begins by distinguishing between BNs used to represent uncertainty about unique events and those describing sampling events, and then explores uncertainty arising from limited data through simple examples involving categorical outcomes and diagnostic testing.

These examples illustrate the effects of sparse evidence, the interpretation of probabilities for unique versus repeatable events, and the importance of Cromwell’s rule: probabilities should not be assigned a value of zero unless a proposition is logically impossible.

The lecture then considers uncertainty related to model structure, network structure, and expert judgement. Examples demonstrate how assumptions about dependencies between variables can affect inference, and how alternative network structures can be compared using likelihood-based methods commonly employed in machine learning. Approaches for reducing bias in expert elicitation are also discussed. Finally, the lecture presents a broader view of Bayesian networks, arguing that any Bayesian model can be represented as a BN, while the term is often used more narrowly for networks with categorical or discretised variables. A simple example of a BN with continuous variables is used to illustrate this perspective.

Dr. Trefor Bazett is a mathematician and lecturer in the Department of Mathematics and Statistics at the University of Victoria, Canada, with expertise in mathematical education and a strong focus on making advanced mathematical concepts accessible and intuitive.

A significant part of his educational work covers probability theory, conditional probability, Bayes’ theorem, and Bayesian inference — the mathematical foundations for reasoning under uncertainty and updating probabilistic assessments as new information becomes available. His teaching demonstrates how Bayesian reasoning provides a rigorous framework for moving from prior assumptions to updated conclusions based on new evidence.

Dr. Bazett is the author of an educational series on Bayes’ theorem and Bayesian inference, published by Springer Nature. The series explores conditional probability, prior and posterior probabilities, different formulations of Bayes’ theorem, false positives, and the practical interpretation of probabilistic evidence.

Known for his ability to explain complex mathematical ideas clearly and intuitively, Dr. Bazett has created more than 500 educational mathematics videos reaching an international audience. His online mathematics courses and lectures have attracted hundreds of thousands of learners worldwide.

He received his PhD in Mathematics from the University of Toronto and has taught at the University of Toronto and the University of Cincinnati before joining the University of Victoria. His contributions to mathematics education have been recognized with major teaching distinctions, including the PIMS Education Prize (2024) and the University of Victoria Faculty of Science Award for Teaching Excellence (2025).

This lecture is an introduction to Bayesian Reasoning. We will discuss how Bayesian reasoning provides a framework for making and refining predictions in uncertain situations. We will begin by introducing the core ideas of prior and posterior beliefs and conditional probability.

We will then explore Bayes' Formula and it's consequences. Along the way we will see plenty of practical examples and explore how Bayesian reasoning can inform our understanding of the world.

Ilija Vorontsov is a software engineer and Team Lead at LOYTEC Electronics in Vienna, specializing in embedded systems, software development, and the design and verification of complex technical systems. His professional experience includes Zephyr RTOS, Bluetooth LE and Bluetooth Mesh, embedded Linux, hardware design, as well as software testing and verification.

He studied Computer Engineering at TU Wien, with a focus on automation and control theory, software validation and verification, dependable and distributed systems, FPGA and microcontroller programming, and SAT solving. His academic work includes real-time performance analysis on ARM Cortex microcontrollers and proof transformation for SAT solver verification — areas that require rigorous approaches to modelling, logical reasoning, and the analysis of complex interconnected systems.

This engineering and analytical background provides the foundation for his work with probabilistic models and Bayesian networks. His approach places particular emphasis on translating real-world problems into structured models, defining variables and dependencies, examining assumptions, and understanding how new evidence changes model outcomes.

Since 2025, Ilija has taught probabilistic and Bayesian modelling at ICEUR and is ICEUR’s leading specialist in the practical application of GeNIe 5.0. His teaching covers Bayes’ theorem, Bayesian networks, and the complete process of developing, analysing, and interpreting probabilistic models — from defining variables and their states to specifying dependencies, entering evidence, analysing posterior probabilities, and exploring alternative scenarios.

A central focus of his teaching is the transition from Bayesian reasoning as a theoretical concept to Bayesian networks as a practical analytical tool. Through hands-on work with GeNIe 5.0, he demonstrates how complex problems can be transformed into transparent probabilistic structures and how such models can be refined and used for systematic analysis and decision-making under uncertainty.

This lecture focuses on the development and interpretation of Bayesian network models, from translating a real-world problem into a network structure to understanding the probabilistic conclusions produced by the model.

Participants will learn how to identify relevant variables and their possible states, define dependencies between them, and specify conditional probabilities. The lecture will examine how evidence is introduced into a Bayesian network, how it propagates through the model, and how posterior probabilities change as new information becomes available.

Particular attention will be given to interpreting model results, exploring alternative scenarios, and understanding how assumptions about dependencies and probabilities influence conclusions. The session will provide participants with a practical foundation for developing Bayesian network models and using them for structured analysis, forecasting, and decision-making under uncertainty.

Лекция посвящена разработке и интерпретации моделей байесовских сетей — от преобразования реальной задачи в структуру сети до понимания вероятностных выводов, получаемых с помощью модели.

Участники научатся определять значимые переменные и их возможные состояния, устанавливать зависимости между ними и задавать условные вероятности. Будет рассмотрено, как новые данные вводятся в байесовскую сеть, как они распространяются по модели и как меняются апостериорные вероятности по мере поступления новой информации.

Особое внимание будет уделено интерпретации результатов моделирования, анализу альтернативных сценариев и пониманию того, как исходные предположения о зависимостях и вероятностях влияют на выводы. Лекция даст участникам практическую основу для разработки моделей байесовских сетей и их применения в структурированном анализе, прогнозировании и принятии решений в условиях неопределённости.