Our research

What makes problems difficult? What factors determine difficulty of problem solving for humans? This question is important - humans enjoy to solve problems only if they have just right level of difficulty. Too easy problems are boring. Too difficult problems are annoying. We want to provide people with problems of right difficulty, particularly in education, but also in games and free time activities (such as Sudoku, which is quite a large business).

However, it is quite difficult to establish difficulty of problems. We currently try to create computational models of human problem solving and use these models to predict difficulty. We focus on puzzles, since they are well-defined and it is easy to collect data (people are willing to solve puzzle in their free time, so we do not need to pay them).

We have so are doing experiments mainly with the following puzzles: Sudoku, Sokoban, Replacement puzzle, Nurikabe, Rush hour.

Problem solving tutor In e-commerce it is customary to use recommendation algorithms, which give user a recommendation of products, which the user may like. These recommendations are not based on any fixed criteria, but rather on collective behaviour of all users (the approach is also called collaborative filtering). We would like to use this approach in education, particularly in training of problem solving skills.

We are building a problem solving tutor - a system which lets you solve different problems (mainly puzzles) and gives you recommendations what to solve next. These recommendations are based on your previous behaviour and on behaviour of other users.

Our research is supported by GAÄŒR grant P202/10/0334 "Solving Difficult Well-Structured Problems: Human-Computer Collaboration". At the moment we are focusing particularly on two topics.