Interview with Prof. Mark Steedman

Cognitive Science in the School of Informatics, University of Edinburgh
2 Buccleuch Place, Edinburgh, EH8 9LW, Scotland, United Kingdom
Personal website

What do you think cognitive systems are?

I take cognitive systems to be programs that reason about the real world in ways that resemble the way humans and other animals do, at least in terms of their end results.

What is your area of research within Cognitive Systems?

Computational Linguistics and Artificial Intelligence.

Why did you become a researcher?

It beats working for a living.

How did you get into Cognitive System research?

I started as a psychologist, but was dissatisfied with the range of theory it offered. I moved into artificial intelligence because it offered algorithmic working models of things like planning and language processing.

Where did you study and what subjects did you study?

At the Universities of Sussex and Edinburgh. I studied Experimental Psychology and Artifical Intelligence. I then taught at the University of Warwick, University of Pennsylvania, and University of Edinburgh, where my students have taught me more than they know.

Can you describe briefly how you are doing what you do?

I work on quite a wide range of problems, most of which are recognized as open problems by linguists, psychologists, and philosophers. I try to solve them by novel, distinctively computational, means.

What are the techniques used in your research?

My students and I write computer programs that do things that can be evaluated as right or wrong. Sometimes we prove things.

Can you tell me why they are important?

Many problems in our parent disciplines are in fact quite irrelevant to the business of building and understanding artifical and natural cognitive systems. Building working programs helps to avoid
getting sidetracked by such problems. Computer science is also a rich source of novel ways of thinking about the other, relevant problems, because computation is inherently dynamic, and so are we.

What are the major implications of your work?

That language itself is simpler than most theories would have us believe, and that all its surface forms, whether spoken, written, or gestural, reflect a universal natural logic that is directly related to general-purpose on-line context-dependent reasoning about action and causation in natural cognitive systems.

Who will benefit from your research / techniques?

I don't know. It depends how far we get. If we can crack the problem of natural language understanding and practical reasoning by machine, it will change access to information utterly, for everyone.

What skills do you think are most important to a Cognitive Systems researcher?

Ability to both think formally and computationally, and willingness to apply theory to practical problems. An understanding of the power of statistical models, combined with enough understanding of earlier approaches to natural cognitive systems to understand what the real problems are.

What do you think is most satisfying about Cognitive Systems research?

All knowledge is produced and used by cognitive systems. So Cognitive Systems Research is very important.

What do you consider is the most challenging about being a Cognitive Systems researcher?

Keeping up with the new ideas in computer science, and understanding how they relate to the old questions about natural cognitive systems.

What do you think are the main challenges for the future?

The greatest obstacle to natural language understanding by machine in our feeble understanding of the semantics and its relation to experientially grounded knowledge and practical inference, about which we also know next to nothing. Most of this is entirely latent and can only be inferred indirectly from the nature of the languages which were hung on to it almost as an evolutionary afterthought, and the very difficult business of building cognitive systems that learn from interaction with the real world.

There are several discussions or debates associated with Cognitive Systems research.

Could you mention issues relating to your work?

I think the biggest issue facing the Cognitive Systems movement is whether this program will work without us building in a lot of structure attributable to evolution, and if not, whether we know how to do that without falling into all the old traps of logicist AI.

Can you outline the arguments of the opposing sides of the debate?

My own response to that problem is to assume that natural language can tell us much more than we have been able to realise about that prelinguistic substrate. However, to exploit this possibility requires a much more radical approach to the relation of linguistic form to conceptual structure than is usual in linguistic semantics. Essentially, the approach is to refuse to believe in any theoretical mechanism or architecture that makes the relation other than entirely transparent. What makes it difficult is that language is also immensely ambiguous and elliptical.

Thank you!

Text only
An euCognition Network Action
Department of Computer Science
University of Bath
Send us an email:
euCognition -
HomeFor PressFor KidsFor AcademicsAbout Us
Meet the Scientists
Explore the Debates
Study at University
Robot Gallery
Home > Meet the Scientists > Prof. Mark Steedman