Joint Modeling

For bridging implementational and algorithmic levels of analysis (Marr, 1982), Bayesian joint models (Turner et al., 2013) have been applied for investigating shared statistical constraints between neural activities and cognitive model parameters. However, the previous joint modeling approach has assumed linearity in its linking function, which might not always be ideal when dealing with complex brain dynamics. Moreover, joint models based on covariance estimation often sacrifice the temporal dynamics of cognitive processes. As a solution to these limitations, we propose a Gaussian process joint model (GPJM), a data-driven and nonparametric joint modeling framework based on hierarchical Gaussian process latent variable models (Lawrence & Moore, 2007). In the GPJM, latent Gaussian processes serve as a linking function and model temporal dynamics governing neural and behavioral observations. The GPJM can incorporate spatiotemporal covariance structures as its constraints and evaluate the relevance of each latent dimension to the process of data generation. To verify the utility of the GPJM, we tested the model performance with simulation and an application to real data. The simulation results showed that the GPJM estimates cognitive dynamics while exploiting spatiotemporal constraints. In an fMRI experiment based on a continuous motion-tracking task, the GPJM explained the neural and behavioral data appropriately and also revealed non-trivial underlying dynamics that generate the data. Cross-validation analyses demonstrated that the latent dynamics trained with complete neural data and partially observed behavioral data could predict test data reasonably. How the latent dynamics could be interpreted is an open question.

The flexibility to learn diverse tasks is a hallmark of human cognition. To improve our understanding of individual differences and dynamics of learning across tasks, we analyze the latent structure of learning trajectories from 36,297 individuals as they learned 51 different tasks on the Lumosity online cognitive training platform. Through a data-driven modeling approach using probabilistic dimensionality reduction, we investigate covariation across learning trajectories with few assumptions about learning curve form or relationships between tasks. Modeling results show significant covariation across tasks such that an entirely unobserved learning trajectory can be predicted by observing trajectories on other tasks. The latent learning factors from the model include a general ability factor that is expressed mostly at later stages of practice, and additional task-specific factors that carry information capable of accounting for manually defined task features and task domains such as attention, spatial processing, language and math.

Transfer of learning refers to how learning in one context influences performance in a different context. Because tasks are rarely performed in isolation, a well-versed theory of transfer is paramount to understanding learning. Yet, a thorough understanding of transfer has been frustratingly elusive, with some researchers arguing that meaningful transfer rarely occurs or attempts to detect transfer are futile. In spite of this pessimism, we explore a model-based account of transfer. Building on the laws of practice, we develop a scalable, quantitative framework to detect transfer (or lack thereof). We perform a simulation study to explore, under what conditions, can we detect transfer and the recoverability of the model. We then use our modeling framework to explore a large-scale gameplay dataset from Lumosity. Preliminary results suggest our model provides a reasonable account of the data and that the added complexity of transfer is justified.

Over the past decade, the role of attention in value-based decision making has become a central research topic in cognitive psychology and neuroscience. On the basis of eye-tracking data, extensions of prominent sequential sampling models of decision making have been proposed, including the attentional Drift Diffusion Model (aDDM; Krajbich et al., 2010, Nature Neuroscience). These models take the influence of attention on preference formation into account and can thus be tested not only on choices and response times (RT) but also on eye-tracking data. Importantly, however, parameter estimation of these models has relied exclusively on choices and RT so far. This limitation is particularly problematic in light of recent evidence that attention itself can be influenced by preference formation (Gluth et al., 2020, Nature Human Behaviour). The goal of the present work was to overcome this restriction and to include eye-tracking data when estimating parameters of sequential sampling models. Using the aDDM and extensions of it, a general joint modeling approach for predicting choices, RT, and eye movements is presented. This approach combines extensive model simulations with probability density approximation and differential evolution Markov chain Monte Carlo sampling to enable hierarchical Bayesian parameter estimation. With respect to eye-tracking data, the approach focusses on fixations and takes their latencies and locations into account. The proposed joint modeling technique is shown to provide a more sensitive comparison of different implementations of the aDDM. It promises to advance the research on computational modeling of attention and decision making substantially.

Humans are aversive to risk (irreducible uncertainty) and ambiguity (reducible uncertainty). However, strong ambiguity aversion does not necessarily imply strong risk aversion. Further, in real life it can be challenging to attribute uncertainty and one may treat ambiguity as risk. This can lead to biases in information sampling, i.e. premature stopping of collecting information that could reduce uncertainty. These biases in information sampling have also been linked to delusional thinking and hallucination disposition in both healthy individuals as well as in mental disorders like schizophrenia. Modelling allows to identify the processes and aberrances in decision-making. Here, we experimentally investigate these potentially aberrant attributions by using the draws to decision version of the beads task (Huq et al., 1988) and the risk and ambiguity lottery task (Levy et al., 2010). For each participant (N=77) we extracted their risk-, and ambiguity aversion using the hierarchical Bayesian modelling of Decision-Making tasks R-package (hBayesDM; Woo-Young et al., 2017), and used those parameters as predictors for explaining the draws to decision in the beads-task. Preliminary results indicate that a person’s risk aversion but not ambiguity aversion is related to draws to decision in the beads task. This displays both the usefulness and importance of modelling cognitive tasks to better understand and analyze the results from decision-making tasks, as well as its importance in order to better understand and disentangle the underlying mechanisms of everyday biases.