ICCM I

As models of cognition grow in complexity and number of parameters, Bayesian inference with standard methods can become intractable, especially when the data-generating model is of unknown analytic form. Recent advances in simulation-based inference using specialized neural network architectures circumvent many previous problems of approximate Bayesian computation. Moreover, due to the properties of these special neural network estimators, the effort of training the networks via simulations amortizes over subsequent evaluations which can re-use the same network for multiple datasets and across multiple researchers. However, these methods have been largely underutilized in cognitive science and psychology so far, even though they are well suited for tackling a wide variety of modeling problems. With this work, we provide a general introduction to amortized Bayesian parameter estimation and model comparison and demonstrate the applicability of the proposed methods on a well-known class of intractable response-time models.

We present a spiking neuron-based model of the Stroop task where the attention mechanism is entirely implemented with distributed representations. This is done by using the Neural Engineering Framework and the associated Semantic Pointer Architecture to implement a selective attention mechanism. The resulting system exhibits the Stroop effect, as well as the associated Facilitation and Interference effects. In contrast with previous models, these effects are not generated via a localist competition mechanism. Rather, these effects are a result of controlled unbinding of information from a combined distributed representation.

Different accounts have been developed to explain the mechanisms underlying value biases during perceptual decision-making, within the model framework of bounded accumulation. The starting point bias account suggests a shift in the starting point of evidence accumulation, in the direction of the more valuable alternative. The drift rate bias account suggests that the mean rate of accumulation is steepened for the more valuable alternative. While most studies have supported a starting point bias (SPB) approach, recent work (Afacan-Seref et al., 2018) suggests that drift rate biases (DRB) may also be applied in certain circumstances. Here, we used human EEG signatures of competitive motor preparation to construct a cognitive decision model that can explain the biasing mechanisms through which participants perform a value-biased orientation discrimination task under a strict deadline. Motor preparation dynamics showed signs of a value bias that emerged prior to evidence onset and increased steadily with time. Accordingly, we constructed a model that included an anticipatory dynamic urgency signal towards the High Value alternative. This model provided a better fit to behaviour than models with either a starting point or a drift rate bias but no anticipatory dynamics. These results point to a role for value-modulated, anticipatory motor preparation in fast-paced decision-making tasks, and suggest a unitary mechanism that can generate both static (starting point) and dynamic (drift rate) biases at the same time.

While much is known about how humans make decisions based on the recency, frequency, and similarity of past experiences, much less is known about how humans weigh the contextual features and the impact it has on decisions. The present study uses a novel method of introspecting into a cognitive model of human decision making in an abstract cyber security game to gain insight about the cognitive salience of the features. The results show that cognitive salience can provide valuable evidence about how and why individuals make their decisions. The implications of these results are discussed with regard to theory and application.

Neurophysiology and neuroanatomy limit the set of possible computations that can be performed in a brain circuit. Although detailed data on individual brain microcircuits is available in the literature, cognitive modellers seldom take these constraints into account. One reason for this is the intrinsic complexity of accounting for mechanisms when describing function. In this paper, we present multiple extensions to the Neural Engineering Framework that simplify the integration of low-level constraints such as Dale's principle and spatially constrained connectivity into high-level, functional models. We apply these techniques to a recent model of temporal representation in the Granule-Golgi microcircuit in the cerebellum, extending it towards higher degrees of biological plausibility. We perform a series of experiments to analyze the impact of these changes on a functional level. The results demonstrate that our chosen functional description can indeed be mapped onto the target microcircuit under biological constraints. Further, we gain insights into why these parameters are as observed by examining the effects of parameter changes. While the circuit discussed here only describes a small section of the brain, we hope that this work inspires similar attempts of bridging low-level biological detail and high-level function. To encourage the adoption of our methods, we published the software developed for building our model as an open-source library.