Game strategies make IT systems more robust

Troels Bjerre Lund is a theoretical computer scientist specialized in developing algorithms that – inspired by economic theory – makes digital systems more robust against strategic users.

ResearchComputer Science Department

What is your current research about?

Today, we rely on various IT systems that don’t always take into account that users do what is best for themselves, and not necessarily what is best for the system - in other words, users who behave strategically. Strategic behaviour can cause even simple systems to crash or to not function as intended, and the problem only gets worse as more and more of the "users" are actually computer programs, not people. My research is, among other things, about designing systems where no users benefit from diverting from the behaviour intended by the designers.

An example from the real world might be an IT system designed to create a work schedule for ticket inspectors aiming to – with as few resources as possible – ensure that riding without a ticket is not worthwhile. In this example, the work schedule is a system in which passengers behave strategically. A good schedule would be one that motivates passengers to buy tickets because they consider the probability of meeting an inspector as sufficiently high. The better algorithms we have for calculating good game strategies, the more complex systems we can analyse and optimize.

What have you discovered so far?

My expertise is calculating robust strategies that take into account the potential weaknesses of other participants. We have long known algorithms for calculating a strategy that ensures that I do as well as possible if my counterpart plays the optimal counter-strategy. But I have found several new ways of efficiently calculating strategies for how I can win if the counterpart does not perform his best, while at the same time protecting me if he plays optimally. Metaphorically, it is a solution in which you are carrying both a sword and a shield. Until now, we haven’t known how to calculate such a strategy – we have had to choose either shield or sword.

What do you find most exciting about your field?

With relatively simple mathematics, we can achieve strategies that reach beyond what we as humans can work out. A simple piece of mathematics can magically transform into a brilliant poker game where advanced techniques like bluffing and slow playing occur all by themselves. Intelligent behaviour can arise from relatively simple definitions.

Further information

Troels Bjerre Lund, Associate Professor, phone +45 7218 5142, email trbj@itu.dk

Vibeke Arildsen, Press Officer, phone 2555 0447, email viar@itu.dk