Research Interests

Robot Ethics

I am interested in robot ethics in general, and particularly in the ethics of robot care for the vulnerable, both older people and children.   In my publications on the topic I have considered the risks and benefits of such robot care, arguing that it is important to consider the likely implications of robotic developments.  We want to benefit from the opportunities that robotics affords, whilst at the same time avoiding the downsides associated with some forms of technological innovations.  There are risks stemming from the so-called 'automation bias', and a tendency to believe robots to be capable of more than they are.  The research field associated with robot ethics is inherently an interdisciplinary one, and is undergoing a rapid development, as the importance of responsible innovation is increasingly recognised.

Human-robot interaction

There are many factors that affect our perceptions of and our interactions with robots.  These include the human tendency to be anthropomorphic, and the role of human emotions. I am interested in exploring the ways in which these and other factors affect our views of robots, and our responses to them. 

Swarm Robotics

Apparently cooperative behaviours can emerge as the result of the interactions between simple robots that are autonomous and that communicate locally rather than globally.  Swarm robotics is inspired by observations and understanding of the self-organisation that underlies the behaviour of social insects.  There is a growing awareness that social insects are not as simple and reactive as was once supposed, and I am interested in exploring the ways in which the simple communicative and representatin abilities of insects can be incorporated into swarm robotics.

Neural computing and Combining Estimators

In the past, I have worked on the topic of adaptive generalisation, and looked at ways in which knowledge and prior experience
can be incorporated into neural nets.  I have also looked at the question of catastrophic forgetting in neural nets, and the use of neural nets for cognitive modelling. I am interested in evaluating the contribution of connectionism on our understanding of cognition, and in more general questions about modelling cognition. 

I have published on the topic of combining artificial neural nets - here the basic idea is that rather than using a single net that will inevitably make some generalisation errors, a combination of nets are used that make different errors.  If combined appropriately, the errors made by one net can be compensated for by correct responses from other nets in the set.   There is now a series of workshops  on Multiple Classifier Systems that extends this approach to combining estimators in general.