Dr Mark Johansen - PhD Indiana
My primary research interests are the learning of categories and the associations between categories. Categories, e.g. cats, chairs, trees, games, etc., are a fundamental component of human cognition. Humans use categories constantly and mostly with little effort in their everyday lives, but this intuitive use gives little insight into where practical categories come from or how people learn them. I study category learning by have people learn new, unfamiliar categories and then evaluate what they have learned and how they have learned it by having them classify new cases in references to their newly learned categories. A key reason categories are useful is that they facilitate the prediction of hidden properties, for example, categorizing an animal as a cat means it’s likely that the animal has a heart, will purr when patted and might help eliminate a mouse infestation. So I also evaluate feature inference by having participants predict properties of new instances in relation to newly acquired categories.
I have recently become particularly interested in the use of virtual video-game environments to study learning. A problem with studying how people learn real-world categories is that it can be quite difficult to experimentally manipulate those categories in ways that allow precise hypothesis testing. Video games represent practical compromise in terms of providing a reasonably high level of realism while still allowing precise manipulation of the experimental.
The study of category learning has become a quite well formalized discipline where many mathematical models have been developed and evaluated. Much of my research is aimed generating data that will help to demonstrate the strengths and weaknesses of the present mathematical models of human learning as well as guide the creation of better models.
I teach on the research design and statistics modules at Levels 1 and2 (PS1015 and PS2006). While guiding students through the details of various formal statistical techniques (analysis of variance, correlation, regression, etc.), I spend a great deal of time emphasizing the intuitive aspects of statistics in the interpretation of experimental results: Human behaviour tends to be noisy and highly variable. Statistical analysis is an organizing tool that enables the researcher to distinguish real influences from random variability.
Selected publications (2014 onwards)
Full list of publications
I am collaborating with Marc Buehner (Psychology, Cardiff) on evaluating the influence of time in causal learning.
I am also working with Magda Osman (Psychology, Queen Mary University of London) on evaluating what the perception of coincidences indicates about mental learning mechanisms.
Lastly, Nathalie Fouquet (Psychology, Swansea University), David Shanks (Psychology, University College London) and I are researching attention to feature covariation as an emergent property of category instances.
Postgraduate research interests
In broad overview, I am interested in doing further empirical work which will help to demonstrate the strengths and weaknesses of the present mathematical models of human categorization and learning and will help to guide the creation of better mathematical models. Further, I believe that additional information can be gained from the modelling process if models are evaluated for their complexity and emphasis is placed on their parameter-free predictions. The inherent difficulty of this is heavily emphasized by the rarity in the literature of mathematical models that make accurate fixed-parameter predictions about psychological processes, but this is a goal worth pursuing. In more detail, I am interested in how category representations differ depending on how the categories are learned. I am currently comparing the category representation resulting from a standard classification learning task--here is an instance, what category is it in?--with the representation resulting from a feature inference learning task--here is an instance of this category with these features, what is the feature that is missing? I am exploring differences in representation by using mathematical models that embody these various kinds of representation.
In the long term, I want my research to be guided by a big-picture view of intelligent behaviour. There is an infinite number of ways that a person or any other intelligent system could generate abstract categories from the information received by their senses. Most of these arbitrary categories would be completely unadaptive because they would not mediate functional prediction/generalization or accurate control. A fundamental question then: How do humans select/learn adaptive categories?
If you are interested in applying for a PhD, or for further information regarding my postgraduate research, please contact me directly (contact details available on the 'Overview' page), or submit a formal application here.
B.S.C. in Biology with a Computing Emphasis (major and minor) and a major in Psychology at Andrews University (Berrien Springs, MI, USA), Summa Cum Laude. 1994. Research Supervisor: James L. Hayward.
Ph.D. Joint Degree in Cognitive Psychology and Cognitive Science with a Certificate in Mathematical Modeling at Indiana University, Bloomington, USA. 2002. Dissertation Advisor: John K. Kruschke.
2004-present: Lecturer, School of Psychology, Cardiff University, United Kingdom
2002-2003: Postdoctoral research fellow at the Center for Economic Learning and Social Evolution (ELSE), University College London, United Kingdom.