Dr. Stephen Read
We present a single-agent neural network model, based on the biologically plausible neural network framework Emergent (O'Reilly et al., 2020), that operationalizes our theory of how individual differences in the neural systems underlying motivation interact with situational characteristics to give rise to within-subject personality dynamics (Read et al., 2010). We first manipulate key parameters of our neural network model to create "individuals" varying in their underlying motivational structure and dynamics. We then simulate the interaction of these individuals with varying situational configurations in virtual environments created with the video game engine Unity to provide a complex model explaining multifarious factors including: A) how situational configurations produce high within-subject variability in behavior; B) how certain situational configurations give rise to some personality factors more than others; C) the degree to which one’s personality structure as opposed to one’s environment plays a role in producing behaviors; and D) how physiological factors influence the presentation of the Big Five personality factors through behaviors.
This is an in-person presentation on July 20, 2023 (11:00 ~ 11:20 UTC).
Social psychology has not yet examined the attitude change process in a situation where someone is persuaded in multiple directions from different sources. To examine the process in such a situation, Nakamura and Miura (2019) conducted an experiment that manipulated the cognitive resources of the participants and showed the applicability of the heuristic-systematic model (HSM), which is known as a model of the attitude change process in unidirectional situations. In this study, we propose a cognitive model that can not only tests the applicability of the HSM in such a situation but also quantifies the quality of the stimuli and manipulations used in the experiment from the data. In addition, we fitted the data from Nakamura and Miura’s study (2019) to the cognitive model and estimated the parameters in a Bayesian way. As a result of the posterior predictive check and the model comparison by the Bayes factor, it was shown that the HSM is applicable in such a situation; however, some evidence against the HSM was also obtained from the posterior predictive check. Moreover, parameter estimates indicated that the quality of some stimuli and manipulations was not as intended by the experimenter.
This is an in-person presentation on July 20, 2023 (11:20 ~ 11:40 UTC).
The wisdom of the crowd effect is the finding where a group aggregation of the crowd is more accurate on average than any random individual from the crowd. The wisdom of the crowd can be applied to a variety of tasks, such as a ranking one where participants are told to rank a set of items according to some criterion. In past work that applies the wisdom of the crowd to ranking tasks, participants are typically asked to rank all items (Lee et al., 2014) although they have also been asked to rank a random subset of the items (Lee et al., 2022). We consider how aggregated rankings are affected when participants are allowed to choose which items they include in their ranking and which items they exclude. Previous work on binary-choice trivia questions found that wisdom of the crowd aggregates were more accurate when individual participants chose which questions they wanted to answer (Bennett et al., 2018). We develop a Thurstonian model for our novel subset ranking task and evaluate performance by comparing the model-generated ranking to the correct ranking. References: Bennett, S. T., Benjamin, A. S., Mistry, P. K., & Steyvers, M. (2018). Making a wiser crowd: Benefits of individual metacognitive control on crowd performance. Computational Brain & Behavior, 1, 90-99. Lee, M. D., Bradford, N., & Tejeda, H. (2022). Using thurstonian ranking models to find the wisdom of the crowd. Paper presented at meeting of the European Mathematical Psychology Group. Rovereto (TN), Italy. Lee, M. D., Steyvers, M., & Miller, B. (2014). A cognitive model for aggregating people's rankings. PloS one, 9(5), e96431.
This is an in-person presentation on July 20, 2023 (11:40 ~ 12:00 UTC).
Tipping points or phase transitions separate stable states in psycho-social systems. Examples are quitting smoking, radicalization, and dropping-out of school. Two knowledge gaps prevent our ability to predict and control these tipping points. First, we miss explanatory mathematical models of such non-linear processes. Second, we ignore the multilevel character of psycho-social transitions. I contend that important changes in many psycho-social systems are cascading transitions, where individual transitions trigger or are triggered by social transitions. The cascade of radicalization of individuals in the context of political polarization in societies is an example of such a multilevel process. In this talk I will present and discuss a mathematical approach to the study of cascading transitions. This approach comprises theory construction in the form of mathematical modelling and innovative empirical analyses. The basic cascading transition model is: dX_i / dt = -X_i^3 + a_i + ∑_jb_ij X_j +∑_jc_ij X_i X_j With specific choices of a, b and c we can apply this model in different contexts. Variants of this complicated model have been recently analysed in climate research (Dekker et al., 2018; Klose et al., 2020; von der Heydt et al., 2019), mostly for the case that c_ij=0. In a new ERC project, we will apply this model to: a) opinion change from individuals to populations and back, b) learning, where progression and drop-out are embedded in collective processes, c) addiction, where transitions to addiction or abstinence within individuals are part of cascading epidemiological changes of substance use in populations, d) muti-figure multistable perception. These individual projects will also be presented as posters.
This is an in-person presentation on July 20, 2023 (12:00 ~ 12:20 UTC).