## Overview

This course is aimed at students and professionals wanting to get an understanding of, and some hands-on experience on, using Bayesian networks for interdisciplinary (and even transdisciplinary) environmental research. The first version of this course will be given at the University of Helsinki DENVI doctorate programme in autumn 2019. This course comes with study groups and consultations with the teachers; however we believe that the study materials can be also used for individual study.

The teaching is based on flipped learning, which requires an active attitude from the students. The students are expected to read the study materials and discuss the study questions in addition to watching the videos. Also, it is advisable to study the materials in the presented order.

No specific technical skills are required; however fluent use of workstation computers is necessary.

The content of the course has been prepared by Dr Laura Uusitalo; the pedagogical expert is Dr Riikka Puntila-Dodd.

### After the course, the student can

- Explain what Bayesian Networks are and how they work
- Evaluate theoretical, scientific, and cognitive factors that need to be taken into account when designing an interdisciplinary BN model
- Design and build an interdisciplinary Bayesian Network model on their research question using a readily available software package
- Find and evaluate information sources to populate the model

### Proposed schedule of the course

For the first 4 weeks, the material will consists mostly of reading scientific articles and watching the teaching videos. This material is fully linked below.

On weeks 5-7, the students will build their own BN models on their own research problem. In the organized course, there will be peer support groups and consultation with the teachers. In the end, there is a course conference in which all participants will present their models, including problem framing, interdisciplinary aspects, model structure and why it was selected, data sources, and critical assessment. All students will peer-review two other presentations.

**Technical note!** The videos may not work on Chrome due to a protocol problem with the University service.
In that case, try another browser.

## Week 1: Basics

### Introduction to the course:

### Read these introductory & overview papers:

- Advantages and challenges of Bayesian networks in environmental modelling, Uusitalo 2007
- Bayesian networks in environmental modelling, Aguilera et al. 2011
- Good practice in Bayesian Network modelling, Chen & Pollino 2012

### Introductory lectures:

### Discuss:

- In which contexts would you use frequentist probabilities, in which Bayesian?
- What kind of research questions seem to be well-suited to be approached with Bayesian network methodology?

## Week 2: Interdisciplinary research

### Interdisciplinarity: definition and motivation

Article: Open your mind to interdisciplinary research

Blog: The 5 Significant Advantages of Interdisciplinary Research

### Interdisciplinary environmental research and problem framing

Read these research papers

- Ten tips for developing interdisciplinary socio-ecological researchers. Kelly et al. 2019
- Growing into Interdisciplinarity: How to Converge Biology, Economics, and Social Science in Fisheries Research? Haapasaari et al. 2012
- Interdisciplinary problem framing for sustainability: Challenges, a framework, case studies. Clark et al. 2017
- What do you see as the main benefits and pitfalls in interdisciplinarity regarding your own research interests?
- Give examples of what a research project could look like, if it were multidisciplinary, interdisciplinary, or transdisciplinary.
- Bayesian Networks, Darwiche 2010
- The Hugin tool for probabilistic graphical models, Madsen et al. 2005 - this paper is rather technical, don't worry if it is a bit too much
- Did you run into problems when working with Hugin? Were those problems technical or related to the definition, or both?
- Integration of biological, economic, and sociological knowledge by Bayesian belief networks: the interdisciplinary evaluation of potential management plans for Baltic salmon, Haapasaari et al. 2011
- Participatory development of a Bayesian network model for catchment-based water resource management by Chan et al. 2010
- Building Probabilistic Networks: Where Do the Numbers Come From? Guest Editors' Introduction by Drudzel & van der Gaag, 2000
- An overview of methods to evaluate uncertainty of deterministic models in decision support
- A guide to eliciting and using expert knowledge in Bayesian ecological models
- Description of the research problem
- Problem framing
- Creation of the model structure and defining the variables (this is often an iterative process). Does your model include decision variables? Does it include utility variables?
- Quantifying the model: Think about where the numbers would come to each part of the model, and quantify at least one part of the model (at least one, non-trivial conditional distribution and its parents)
- How to write an abstract for a conference
- 6 Tips for Giving a Fabulous Academic Presentation
- Useful guidelines for presentation peer review: 10 Strategies to Make Peer Review Meaningful for Students

### Lecture: Why are BNs good for interdisciplinary research?

### Discuss:

## Week 3: Building Bayesian networks

### Model theory

### Hands-on: Hugin software

### Hands-on exercises to get familiar with Hugin

Download this file and work through the examples: Introduction to Hugin with exercises (pdf)

Also, take a look on Hugin help pages, particularly the one titled Introduction to (Limited Memory) Influence Diagrams.

### Read these research papers

### Discuss:

## Week 4: How to build models

## Lectures on defining the variables

### To read: Example papers including description of models structure, variable definition, etc.

### To read: Papers on parameterising the models

### Lecture: Introduction to cognitive biases

This lecture aims to shed light on some of the problems associated with humans as evaluators of numeric values.

## Weeks 5-7: Building your own model

Now, you're ready to build your own model!

The exercise during this course is to start building your own Bayesian Network to study an interdisciplinary environmental research question. It will not be possible to make a full model during this time; not to mention that an interdisciplinary model requires an interdisciplinary team. However, you can walk through all the steps of the modelling process:

Now, present your model to other students in the course conference! Also, receive and give feedback to two other students.

### Some presentation guidelines:

## Acknowledgements

This course has partially been developed within the BONUS BLUEWEBS project, which has received funding from BONUS (Art 185), funded jointly by the EU and the Academy of Finland.