Scaling function learning from individuals to groups

dc.contributor.authorChubala, Chrissy
dc.contributor.examiningcommitteeLeboe-McGowan, Jason (Psychology) Singer, Murray (Psychology) Bunt, Andrea (Computer Science) Thompson, Valerie (University of Saskatchewan)en_US
dc.contributor.supervisorJamieson, Randall (Psychology)en_US
dc.date.accessioned2017-09-06T17:30:57Z
dc.date.available2017-09-06T17:30:57Z
dc.date.issued2017
dc.degree.disciplinePsychologyen_US
dc.degree.levelDoctor of Philosophy (Ph.D.)en_US
dc.description.abstractScale invariance, the notion that scientific principles ought to hold over different scales of analysis, is a regularity in the physical and biological sciences but is underappreciated in psychology. Whereas the standard approach in psychology is to explain behaviours at different scales of analysis with different mechanisms, I argue that sophisticated behaviours at any scale are an emergent consequence of simple processes interacting with a structured environment. Changing the scale of analysis, whether temporal, physical, or otherwise, may alter the structure of the environment but need not imply changes to the mechanisms that interact with that environment. To illustrate the centrality of scale invariance and emergence to human cognition, I replicate signature findings from a function learning task after scaling up the unit of analysis, from individuals to groups. In a standard function learning task, individuals learn the relationship between two variables by trial and error, matching one variable (i.e., Y) to a target value of the other variable (i.e., X) and adjusting their responses according to feedback. In an analogous group function learning task, groups of non-communicating individuals learn the relationship between two variables by making individual-level decisions in response to group-level feedback. My experiments with this task demonstrate that groups, like individuals, can learn both simple and complex functions by trial and error, and can generalize their knowledge of a trained function to untrained target values in a transfer test. Groups are, moreover, resilient to disruption of their knowledge, a central feature of distributed representations in biological and artificial neural networks. The results recommend a principled approach to cognition, in which simple processes interact with the structure of the environment to produce sophisticated behaviours, and in which the patterns of behaviour produced at one scale of analysis are reproduced at other scales. Finally, the data show that, when constrained by a collective environment and common goals, individuals self-organize into unique decision roles that support group-level learning. An exploratory analysis of self-reported strategies, individual behaviours, and personality profiles demonstrates how complex social variables can help or hinder the emergence of learning at the level of the group.en_US
dc.description.noteOctober 2017en_US
dc.identifier.urihttp://hdl.handle.net/1993/32448
dc.language.isoengen_US
dc.rightsopen accessen_US
dc.subjectgroup cognitionen_US
dc.subjectfunction learningen_US
dc.subjectscale invarianceen_US
dc.subjectself-similarityen_US
dc.subjectgroup learningen_US
dc.subjectemergenceen_US
dc.titleScaling function learning from individuals to groupsen_US
dc.typedoctoral thesisen_US
Files
Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Chubala_Chrissy.pdf
Size:
1.57 MB
Format:
Adobe Portable Document Format
Description:
License bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
2.2 KB
Format:
Item-specific license agreed to upon submission
Description: