PhD Students
PhD Project Suggestion: Human Reasoning: Computational and Experimental Studies
Howard Bowman
From empirical studies carried out in cognitive psychology, we know quite a lot about human reasoning. In particular, we have data on what lines of logical reasoning humans find difficult (and typically get wrong) and what lines of reasoning they find easy ( and get right). Furthermore, a number of theories to explain this pattern of data have been proposed. Notable amongst these are Johnson-Laird's Mental Models and Oaksford and Chater's explanation in terms of
bayesian probabilities. In addition, there have been some computational implementations of these theories.
Despite these successes, there is still much to do and there are a number of potential directions for PhD research. One starting point would be to develop a
connectionist realisation of Oaksford and Chater's theories. This would respond to the observation that current theories of human reasoning
are typically developed within the symbolic tradition. However, the brain must somehow perform such reasoning in neural circuitry, which prompts the question of whether a neural network implementation of these theories could be developed.
These theories and models raise a number of empirical questions that could be
investigated through behavioural experimentation. It may even be possible to
develop questions that could be explored using the Centre for Cognitive
Neuroscience and Cognitive System's EEG and ERP recording facilities.
Finally, recent research has started to reveal how humans reason about temporal questions. However, comprehensive theories of the available data do not exist. A PhD student could work on extending mental models
and / or bayesian theories in this direction.