Ed Fincham and Frank Karvelis
Tue 03 Oct 2017, 11:00 - 12:00
IF 4.31/4.33

If you have a question about this talk, please contact: Gareth Beedham (gbeedham)

Ed Fincham

Title: Knowledge construction within groups

Abstract: Socially enriched environments are increasingly pervasive of contemporary life. Accordingly, collaborative learning has attracted much attention and numerous studies have documented the benefits of effective collaboration. Collaboration touches on two separate but interrelated domains: the social, and the cognitive. As isolated topics, there is extensive literature on both. However, there has been limited research into the interaction of these two domains. Existing approaches do not adequately capture the complexity of this relationship and, in particular, provide little insight into how individual self-regulated learning relates to the collaborative construction of knowledge in groups, how roles emerge within these groups, and how group regulation relates to individual self-regulation. To capture these social and cognitive dynamics within online learning environments, we aim to augment existing methods from social network analysis (SNA) with behavioural information to investigate how collective regulation arises from the actions and self-regulating behaviour of individual agents.



Frank Karvelis

Title: Bayesian inference in Autistic and Schizotypy traits

Abstract: The Bayesian Brain hypothesis argues that the brain is an ‘inference machine’ – that is, it infers the causes of our often ambiguous and limited sensations (likelihood) by combining it with our model of the world (prior). Recently, it has been proposed that Bayesian inference framework is the relevant level of analysis for understanding the impairments underlying autism and schizophrenia. However,  whether it is the priors or the likelihoods that are primarily impaired has been subject to theoretical debates, while emerging experimental findings also present a mixed story. In addition, most experimental studies suffer from methodological problems – ironically, very few employ computational modelling and rely on behavioral data alone when testing a computational account. More methodologically sound work is needed.


We have conducted a study in motion perception by using a previously developed task which is designed to induce (implicit) prior expectations via more frequent presentation of certain motion directions, which in turn leads to characteristic biases in perception. We found that schizotypy traits had no effect on task performance and thus no effect on the acquired priors or sensory likelihoods. This speaks against the Bayesian hypothesis for schizophrenia. However, in the same sample but along autistic traits we found negative correlation with perceptual bias and variability. Modelling results revealed that such behavioral differences were underlied by more precise sensory likelihoods in autistic traits, while there was no effect on the acquired priors. This presents a challenge to most recent experimental studies, but is in line with some of the theoretical Bayesian accounts of autism.