News

Published 16 February 2010

Focus on research: AI-scientist Bert Bredeweg

Foto: Bob Bronshoff.

The scudding of the clouds across the sky, the pouring of a cup of coffee; one may be more complicated to describe than the other, but they are both examples of systems that change according to set rules. Bert Bredeweg, a scientist specialising in artificial intelligence developed an interactive computer program for describing these types of systems conceptually. His software works similarly to the way humans reason and, consequently, uses no quantitative data. The software is now being applied successfully in scientific research. Bredeweg is currently developing a version of the software for educational purposes. He was recently awarded a European grant of three million euros to continue his research.

Example of a system that changes according to set rules.

Ever since his doctoral research, Bredeweg has been working on interactive software, software that allows you to describe systems while at the same time supports learning about them. ‘I am interested in knowledge construction. As people, we learn by exchanging knowledge with others. In my work I try to analyse and understand this communicative interaction in order to automate the process with computer software.'

Complex networks

He developed his first user-friendly programs, or ‘reasoning frameworks' as he calls them, in collaboration with a group of domain experts. Ecologists needed a framework in which they could formulate sustainability issues along established lines ‘The framework yielded surprising insights. Certain assumptions turned out to be false, as was the case, for example, with research into the ecological value of a river where ecologists determined the quantity of dissolved oxygen in the water. It was simply assumed that turbulence and water plants generated additional oxygen in the water and that fish used the oxygen. However, the conceptual models showed that this process was more complicated than it appeared on the surface. The quantity of fish and water plants also influenced one another. Rotting plants reduced the quantity of oxygen in the water. The entire process turned out to be a complicated interaction between cooperating and opposing influences. By formalising the conceptual knowledge of the experts, a far more subtle image of the whole process emerged. It also forced the experts to align the meanings of a variety of terms. For example, ‘sustainable biomass' and ‘carrying capacity' were common terms, yet everyone had their own way of using them.

To put it simply, Bredeweg's software has made an important contribution to scientific research. And the experts were very positive about the modelling environment. ‘While you can't compare it to Microsoft Word in terms of numbers of users, in the research community the software is used worldwide.'

DynaLearn

After the successful introduction of an interactive model construction environment for use in science, Bredeweg was awarded a European grant of almost three million euros to create a version of the software for education: DynaLearn. ‘This program makes students write down their knowledge using a specific vocabulary and according to a set pattern, and this coaches them to articulate information clearly. Moreover, during the process the software generates questions that assist students in developing their models. The machine encourages students to learn to think logically. What the model produces must tally with what you expected it to produce.'

The DynaLearn interface.

Bredeweg uses the example of a bath being filled. Imagine you want to know what will happen if you turn on the tap. To do that you need a number of quantities, such as the capacity of the bath and the capacity of the tap, and you need to know the relationship between these quantities: if you turn on the tap, the quantity of water in the bath increases by a certain speed. While this is relatively simple, it gets more complicated if, for example, there is a hole in the bath through which water can escape. DynaLearn helps students capture a system like this in a model in which it is not the quantitative data that is important but the correct analysis of cause and effect, and other qualitative insights. For example, what happens if the size of the hole is increased, a smaller bath is used or the water flows faster out of the tap? Will the bath fill up more quickly or more slowly? Will the water stay at the same level or will the level drop?

Learning using virtual characters

With DynaLearn, Bredeweg hopes to improve science education by making it more concrete and more interesting. He sees a big future for DynaLearn ‘It really can make a difference. Students will have a tool to create models of a whole variety of systems: from models that describe the economy to meteorological models. And they are extra motivated by virtual characters: moving and speaking figures that instruct you on the basis of your own models. There is also a ‘teachable agent', a virtual fellow student that ‘wants to be taught'. By constructing models, students shape this virtual character's competence. They can then have the character do a quiz to test how well it has learned by, for example, comparing the quiz scores of their character with those of other students. There are also virtual teachers on hand. Moreover, students can compare their work with that of their fellow students and share knowledge in a Wikipedia type of environment. Students can choose the type of character they prefer to work with.'

Bredeweg hopes to have the first version of the software finished by July. ‘It's a challenge, but it should be possible. The eight partners contributing to the project are specialists in their own field. For example, the virtual characters are being designed in Germany. Tel Aviv University has a lot of knowledge about university education and technology, and understands the differences between the DynaLearn approach and the application of numerical simulations. The University of Brasília is responsible for the environmental aspects of the project. And we are the experts in the area of conceptual modelling, determining what data needs to be recorded in the information system, how the data should be structured and their mutual relationships.'

Observing through different glasses

What motivates Bredeweg to invest so much in education? ‘People benefit greatly from a scientific approach in their education as it allows them to deal better with the growing complexity of the world in which we live. The world population is increasing and we want to keep society intact. We need solutions to problems such as waste disposal, energy and food supplies. And to solve these problems we need people with a technical background.'

Another motivation is that science changed Bredeweg's view of the world and he believes it can have the same effect on others. ‘Science has enriched my life. Science works like an instrument that allows us to better observe and understand the world around us. Think of, for example, a pair of glasses or binoculars that lets you see more of the world or see it differently. With DynaLearn, we are handing the student that pair of glasses. It will teach them to examine the world with enhanced powers of discernment.

Author(s)


Source: Communicatie FNWI
Also refer to