Stevan Harnad

Interview with Prof. Stevan Harnad

Electronics and Computer Science, University of Southampton, UK
Cognitive Sciences, Université du Québec à Montréal, Canada
Personal website

What do you think cognitive systems are?

A grandiose name for computational tools to help research, teaching and thinking.

What is your area of research within Cognitive Systems?

I work on category learning, language origins and representation, and scientometrics (Open Access to research online).

Why did you become a researcher?

Shop-keeping was less appealing. (I am interested in how the mind works.)

How did you get into Cognitive System research?

I quit my Chair in the Department of Psychology at Southampton and Tony Hey saved my career by giving me a Chair in the Department of Electronics and Computer Science at Southampton.

Where did you study and what subjects did you study?

McGill University and Princeton University. Psychology, alas, instead of Computer Science...

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

My students conduct experiments on human subjects while they are learning (mostly visual) categories; we also measure their brain activity changes during the learning. We also do computer simulations of language evolution and neural net models of learning. We also do analyses of meaning representation in digital dictionaries and their relation to the mental lexicon in our heads. We also do scientometric analyses of research usage and impact (citations, downloads, direction, growth) and create software tools ( for creating Institutional Repositories in which researchers and institutions can make their research output openly accessible online. We also create scientometric search engines for analyzing research impact. We will be validating research metrics in the latest RAE 2008.

What are the techniques used in your research?

Psychology laboratory tools such as Matlab and SPSS, event-related postential analysis, neural nets, web crawling robots, graph theoretic analysis.

Can you tell me why they are important?

The question is too vague and general. Computational tools are needed to gather, analyze and model empirical data.

What are the major implications of your work?

That language evolved so we could acquire categories verbally; other species can only acquire them from direct sensorimotor experience. But verbal categories have to be initially grounded in sensorimotor category learning too. The core words of our dictionaries are the ones we learned early, from experience. Beyond dictionaries, encyclopedias and textbooks are our research journal articles. The power of language is extended to all of science and scholarship on the "Category Commons" -- the Open Access research web, for which we are creating tools and conducting analyses and evaluations.

Who will benefit from your research / techniques?

Most of it is basic research, leading to further basic research on how the mind and brain work. There are applications in education and training, as well as in scientific and scholarly research collaboration and evaluation.

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

They should know some biology, neuroscience and psychology, as well as having a strong grasp on computational and statistical skills.

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

It can eventually explain how the mind works, and meanwhile can help it work.

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

Passing the Turing Test (TT) (designing a system that can do everything the mind can do -- thereby explaining it).

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

Scaling up toward the passing the TT.

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

Could you mention issues relating to your work?

One of the debates is about "computationalism": Is thinking just a form of computation? Computation is the manipulation of symbols according to rules, as in a computer. The rules are "syntactic," which means they are based on the shape of the symbols, not the meaning of the symbols.

The logician who co-invented to computer, Alan Turing, has proposed the TT as the test of whether we have explained the mind: Build a system that can do anything and everything a human mind can do, so well that humans cannot tell it apart from a real human being, say, in a lifelong exchange of emails.

If a computer program alone could pass the TT, would it have a mind? The philosopher John Searle, in his famous "Chinese Room Argument," showed that it would not, because he, John Searle, could manipulate the symbols according to TT software's rules, and pass the TT in Chinese without
understanding Chinese. So, he argued, the TT and computation are not enough.

I suggested that the solution is a robotic TT instead of just a language/email TT, and dynamical processes like sensory input motor output transduction, plus hybrid analog and digital processes in
between, would be needed in order to ground the meaning of the symbols (words) used. I called this the "symbol grounding problem": the problem of grounding the meaning of meaningless symbol-shapes to the objects in the world that they denote, by giving a robot the capacity to learn to categorize them.

Another issue in cognitive systems, related to this one, is whether a computational system can be conscious. The reply is the same: A TT-passing robot might be, but a purely computational system would not.

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

The opposing side holds that computation can do anything, hence it can do anything a mind can do. Therefore computation alone should be enough. The sensorimotor components are merely needed for input/output connections between the computations and the world.

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. Stevan Harnad