Unsupervised categorization: Analytic and non-analytic processes
Lead Research Organisation:
UNIVERSITY OF EXETER
Department Name: Psychology
Abstract
Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.
Organisations
People |
ORCID iD |
Andy Wills (Principal Investigator) |
Publications
Beesley T
(2010)
Syntactic transfer in artificial grammar learning.
in Psychonomic bulletin & review
Haslam C
(2007)
Does maintenance of colour categories rely on language? Evidence to the contrary from a case of semantic dementia.
in Brain and language
Lea S
(2008)
Use of multiple dimensions in learned discriminations
in Comparative Cognition & Behavior Reviews
Lea SE
(2009)
A comparative analysis of the categorization of multidimensional stimuli: II. Strategic information search in humans (Homo sapiens) but not in pigeons (Columba livia).
in Journal of comparative psychology (Washington, D.C. : 1983)
Milton F
(2009)
Long-term persistence of sort strategy in free classification.
in Acta psychologica
Milton F
(2008)
Processes of overall similarity sorting in free classification.
in Journal of experimental psychology. Human perception and performance
Milton F
(2009)
The neural basis of overall similarity and single-dimension sorting.
in NeuroImage
Milton F
(2009)
Eye movements in overall similarity and single-dimension sorting
in Proceedings of the 31st Annual Conference of the Cognitive Science Society
Wills AJ
(2009)
A comparative analysis of the categorization of multidimensional stimuli: I. Unidimensional classification does not necessarily imply analytic processing; evidence from pigeons (Columba livia), squirrels (Sciurus carolinensis), and humans (Homo sapiens).
in Journal of comparative psychology (Washington, D.C. : 1983)
Description | Categorization is dividing the world into groups of things. Our ability to do this is so central to our mental life that it's easily overlooked. As an illustration of its importance, try to imagine a world with only proper nouns - you could have the concept of "Fido" but not of "dog", of "Gordon Brown" but not of "prime minister". How do we come to form the categories that we have? Psychologists have traditionally studied this question by showing people examples of novel categories and telling them, repeatedly and for every single item, which category the object belongs to. This method has been quite successful, but it appears unlikely that we receive such extensive, specific and reliable feedback in the real world. The research conducted during this grant used a different technique - unsupervised categorization - where people are asked to classify novel items in the way that seems most appropriate to them, without any feedback from the experimenter. This methodology seems to have more potential to tell us about the way in which people prefer to create categories. One of the things we already know about unsupervised categorization is that people tend to classify on the basis of a single attribute of the stimuli (e.g. colour) and ignore other attributes (e.g. size). This is perhaps a bit surprising, given that few natural categories (e.g. dogs, games) have this structure. One pre-existing explanation for the prevalence of single attribute sorting is that the way psychologists tend to run the experiments puts people into an "analytical" (problem solving) mindset. If people can be encouraged to be less analytical, previous research suggests they can be more likely to sort on the basis of overall similarity (which is the basis on which natural categories are usually assumed to be formed). Some of the results of the current grant re-affirm this hypothesis. However, our results also make a clear case that family resemblance sorting can also be the result of an analytic process. The evidence for this case comes from a number of converging findings. Specifically, we find that: 1. When people are given a demanding task to do at the same time as they are categorizing, they can become less likely to produce family resemblance categories. 2. When given instructions to act in a meticulous and careful way, people can be more likely to produce family resemblance categories than when they do not receive such instructions. 3. People with a comparatively small "working memory" capacity (measured by, for example, their ability to recall digits over short intervals) can be less likely to produce family resemblance categories than those with comparatively large working memories. 4. People rated as impulsive on standard tests can be less likely to produce family resemblance categories than those who are rated as reflective on those tests. The most likely conclusion is that family resemblance sorting can result from both analytic and non-analytic processes, although the conditions under with these two processes are most likely to produce family resemblance sorting differ. A question for future research is to elucidate more clearly the ways in which those conditions differ. |
Exploitation Route | By the end of the award, this work had already inspired a series of animal cognition and human neuroscience studies. Other uses of this work. are likely to be seen first in related work by other researchers of cognitive systems, including psychologists, neuroscientists, and those involved in intelligent systems research (e.g. computer scientists). In the medium term, there are a number of forseeable applications of an increased understanding of unsupervised learning, including an increased understanding of consumer choice behaviours, development of more effective educational systems, and the automated organization of very large data sets (e.g. data mining, search engines). |
Sectors | Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Education |
Description | EC FP6 |
Amount | € 1,300,000 (EUR) |
Funding ID | 516542 (NEST) |
Organisation | European Commission |
Sector | Public |
Country | European Union (EU) |
Start | 01/2005 |
End | 12/2007 |