Sergio in his office at LKL

I am a Spanish postdoc working full time on the MiGen project. The project tries to build a system which helps young learners with generalisation thinking - bridging the gap in mathematics between the stage in which we deal with numbers, and the stage in which we start dealing with variables. This transition is difficult for children of 11 to 13, so we are building a system which helps bridge the gaps between these two worlds - the world of the particular, the numbers; and the world of the general, the variables.

An important part of the system is a microworld in which the chidren can create shapes, play with them, and make relationships with them - saying things like 'this shape is twice as long as this one', or 'there are as many red shapes as green shapes.' These kind of general ideas are easier for kids to express in natural language than using algebra. To help them, we have developed a kind of iconic language that bridges the gap between numbers and variables.

Background intelligence

The system offers support to the learner. As a kid works with the system, she might get lost or stuck; we are putting intelligence into the system to help her. We are not trying to substitute for the teachers, but it is evident that a teacher in a classroom with 30 kids cannot support all of them all the time. Additionally, the system also aims at supporting the teacher - for example, it will be able to tell which students have more problems, or to give a general understanding of the classroom situation.

Finally, the system is also trying to encourage collaborative work between the learners. So imagine that you, as a kid, see how other kids have developed their general strategies; perhaps they have developed a different strategy to yours, but in the end, you get to the same expression or the same pattern. This makes the children reflect on what they have done - the other ways they could have done the same thing.

This is different from a typical tutoring system where you just match the answers against a list of correct answers. In this case we have the kids playing, creating shapes on the microworld, creating expressions that define the length of the shape, the colours, etc. And all this information that we can trace comes to the intelligent component quite unstructured. So right now we are trying to analyse all that data, and find patterns and similarities.

We plan to use different techniques from the field of Artificial Intelligence in Education. We have been making pilots using things like Markov models, sequence mining algorithms, and several other data mining techniques.

My main task in the project is the construction of the intelligent component - the part that studies the actions of the students, tries to find similarities, etc. I have also been involved in the creation of the microworld - the playground. The microworld is now complete to the point of being usable; so for the next two years, which is the rest of the project, I will work on the intelligent side.

My degree is in engineering; I am a telecommunications engineer. I did my degree in Spain, a five year degree. I worked in industry for a short while, then did a Ph.D. in Madrid, on IT applied to learning.

In my Ph.D. I was working mostly on sequencing of learning activites. Imagine that you have several pieces of information and you want to make a course with them for students. I developed a way a computer can easily reorder these pieces of information to show each learner a sequence that is optimal for her. So for example if you know a lot about topic A, but you need to learn more about topic B, instead of showing you the same course as eveyrbody else that has a lot of A and a lot of B, the system will skip most parts of topic A because you already know that, and put more emphasis on topic B. This is part of what my Ph.D. thesis was about - personalisation of the sequence in which the information is presented, not the content or the presentation.

My Ph.D. work was focused on college students. Working now with kids has been a dramatic (but enlightening) change.

Following ants

Sergio at the tomb of Fyodor Dostoevsky, in St. Petersbourg, Russia

In my Ph.D. work I used some swarm intelligence techniques to take a whole set of students and analyse what they do as a group, so help them find appropriate sequences of activities. For example, if everyone in a class follows a certain route (say, a combination of theory and exercises in some order), but I know I am above average and want to go deeper into the concepts I can follow a more difficult route. Or if I know I take more time to do things, I can follow the easier route. Finding these routes a priori is tricky, maybe impossible.  This is where swarm intelligence can be helpful.

An example of swarm intelligence is the so called ant-colony heuristic, which has been used in other fields like telecommunications and resource management; I adapted it to be used for sequencing learning activities. It is a way of taking account of the behaviour of real ants. What ants do for example when they try to get food for the nest — ants are very simple insects and do not have complex brains, but somehow they manage to find the shortest or easiest route from the nest to the food source. How do they do that?

What they do is, when they're walking randomly, they leave a pheromone trail. When they find a food source, they go back to the nest leaving a different pheromone trail. This trail is then followed by other ants, and as more and more ants follow the same route (without any abstract thinking) they make it shorter, because the pheromone trail is stronger at those points where ants have most recently followed. They follow the strongest scent, and in doing that, they incrementally make shorter and shorter paths. And every ant that follows that trail is reinforcing the trail.

This has several advantages. For example, as every child who has played with ants knows, if you break an ant path at some point, the ants seem to be lost. But as soon as one of them finds a connection between one part of the path and another, the others follow that new path, then it is reinforced by more pheromones, and they find a new optimum path between the food and the nest.

What is the relevance of this? It means that solutions found in this way ('the way ants do') can automatically adapt to future changes. In telecommunications engineering, a change can happen for example when an earthquake breaks undersea cables between Japan and the USA, and people still want to access their Gmail accounts from Tokyo. In learning, changes can happen when future learners have different backgrounds than former learners.

This is important because it means you don't have to engineer the best solution from the beginning. If you are an ant, you have a simple strategy which is adaptable to changes in the environment. This is why there are several researchers around the world trying to find ways of using this behaviour in various systems — for example to move information around the Internet. I tried to adapt it to learning sequences... but of course people are not ants, so it needed some changes.

My own path

Sergio and his friend Ismael in Matanzas, Cuba

I have been working at the LKL for one year, and I am very, very happy - with the work atmosphere, and the people in the project, which are great.

Some time ago, in Ireland, I worked in the IT Innovations Centre of Intel. It was really interesting for me. Intel makes money basically from microprocessors, so that is what everything is focused on. The group I was working in researched how to use IT to improve the production of microprocessors. My work was specifically for creating a knowledge platform that helped the workers in the factory to get better training on tasks they were doing. We used different technologies, like PDAs for making information available at all times.

At the University Cote d'Opal in France, some people were doing similar work using the ant-colony heuristic, applying it to things such as artificial vision and robotics. This was linked with Paraschool, a company that provides exercises on the web for French kids; they were trying to put some intelligence into the sequence of exercises, because at the beginning the company was giving the same exercises to everybody. So I went to Paris for four months to work with them. I was there five months actually - I spent August learning French, because I didn't speak a word of it before! Paris is a lovely city to live in, however.

I also spent five months in Greece. I worked with a professor at the University of Piraeus in e-learning standards specifications - to foster communication between e-learning systems, re-use of content, structures and strategies.

I still go back to Spain from time to time, and I love to travel generally - any place that is interesting from a human or historical point of view. Part of my holidays in 2008 I did a trip with my brother around the northeast of Europe - I went from Finland down to Poland, going through Russia and all the Baltic countries, by train and bus.

It was a bit exhausting but very interesting, especially in Russia because there are several big differences and things we didn't expect. We managed to get out of the country in the end, but it was not that easy! For starters, no one speaks English; either you speak Russian or you have a big problem - which was our case. Even simple things like trying to find a train or bus ticket to get out of the country can be difficult! The part of the trip in EU countries was much easier.

I also went to Cuba. I didn't go to the beach much, but spent my days going around the country, speaking to the people about how they see their country and their reality. From the outside it's a very special country and special historical situation, and I wanted to know how they saw themselves. I was there for two weeks, and spoke to a lot of people of different ages. It was very interesting - I enjoyed it a lot. And there I didn't have any problem with the language of course!