News and Events

Interview: 'Robots have a communication problem'

Prof. Emiel Krahmer knows all about the knowledge and communication skills of present-day robots and what they still need to learn. His expertise: translating knowledge into language. And no, not to worry, he is sure that robots are a long way still from beating humans in communication.

Photography: Gerdien Wolthaus Paauw

Photography: Gerdien Wolthaus Paauw

Interview with Professor of Computation, Communication, and Cognition, Emiel Krahmer, by Tineke Bennema

What is the focus of your robot research?

"First of all, we need to make a distinction between different types of robots. There are industrial robots, but also social ones. It is the latter type that people interact with, resulting in communication. We can build a real relationship with them. When robots look at something, for instance, people follow their gaze.

Right now, robots are not very good yet at communicating with us. They can answer some of our questions, but they often repeat themselves. And we get bored as a result. Also on a non-verbal level, their communication is far from perfect. They are not interesting enough for any kind of long-term relationship. This requires major improvement.

My research focuses on the question how robots can learn to communicate like humans. In practice, this boils down to figuring out how information can be translated into language. How a robot can write a sports report, for instance, or how robots can be deployed in health care.

People often ask me how long it will take before robots can communicate perfectly. I am not going to make any predictions there: technology tends to develop in ways you cannot imagine. Fifteen years ago, I would not have thought it would be possible to develop social robots like we are doing now, let alone robots made for love. At the same time, automatic conversation techniques are basically pretty much the same as they were years ago."

Where did you get this passion for and interest in robots?

"I studied Language and Artificial Intelligence at this university, focused on how computers communicate through language. As it turns out, it provides an interesting model of how humans function: you learn a lot about how human communication works. It is hard to capture human communication in algorithms. But it is also fascinating. It forces you to ask questions about the way we ourselves communicate."

What is the impact of your research?

"Europe and NWO (the Netherlands Organization for Scientific Research) are very much interested in robotics and room has been reserved for it in the European Horizon 2020 program, which includes the research we are doing. We are involved in research in the field of robot journalism, for the Dutch Persgroep (major publisher of Dutch newspapers and provider of internet news services) and for Sanoma Media (Dutch media company publishing magazines, for instance, and running TV channels). One of the things we are working on is how bots can write better articles, because, believe it or not, they already are writing articles. We focus on the question how articles can be tailored to specific target groups and to individual readers. We also collaborate with Schools for Journalism at Universities of Applied Sciences.

Next to these, we are working on applications in health care, focusing on communication obviously, more particularly on how health information can be communicated better and more personally. For instance, how doctors can best communicate to their patients different options for treatment. Another branch of our research concerns education, more specifically second languages for children. Within the L2TOR (‘el tutor’) project we are developing a small, friendly social robot that teaches kids in a playful way and that could be employed more frequently in education.

What we are running up against in the development of communicating robots is the challenge presented by creativity in communication. Robots can easily learn the basics, but they still quickly start repeating themselves. You can refine a great many of the robots’ competencies, for instance in the field of language, these are constantly being improved. You can have a bot write a novel, but it will do so according to a fixed pattern, and the result will be like a cheap Harlequin novel. It is as yet incapable of innovative and creative thinking. I think our expectations with regard to social robots are a little too high.

So no robot will ever be able to replace Professor Emiel Krahmer?

“Most certainly not! Robots will not be able to get rid of their communication problems for a long while yet, of that I am convinced!”