How to model the neurocognitive dynamics of decision making: A methodological primer with ACT-R

Abstract

Higher cognitive functions are the product of a dynamic interplay of perceptual, mnemonic, and other cognitive processes. Modeling the interplay of these processes and generating predictions about both behavioral and neural data can be achieved with cognitive architectures. However, such architectures are still used relatively rarely, likely because working with them comes with high entry-level barriers. To lower these barriers, we provide a methodological primer for modeling higher cognitive functions and their constituent cognitive subprocesses with arguably the most developed cognitive architecture today-ACT-R. We showcase a principled method of generating individual response time predictions, and demonstrate how neural data can be used to refine ACT-R models. To illustrate our approach, we develop a fully specified neurocognitive model of a prominent strategy for memory-based decisions-the take-the-best heuristic-modeling decision making as a dynamic interplay of perceptual, motor, and memory processes. This implementation allows us to predict the dynamics of behavior and the temporal and spatial patterns of brain activity. Moreover, we show that comparing the predictions for brain activity to empirical BOLD data allows us to differentiate competing ACT-R implementations of take the best.

Some Metatheoretical Reflections on Adaptive Decision Making and the Strategy Selection Problem

“…Notably, ACT‐R models can allow making very detailed quantitative predictions about the shape of decision time distributions (rather than just their means or medians), the temporal and spatial dynamics of brain activation, eye movements, and decisional outcomes. The richness of those predictions can allow testing models against each other that can otherwise hardly be told apart (Dimov et al, , ; Dimov, Khader, Marewski, & Pachur, ; Marewski & Mehlhorn, ) . All that said, as the methodological papers in this special issue demonstrate, there are yet important methodological challenges to be met (Brown et al, ; Jekel & Glöckner, , ; Rieskamp, ).…”

Individual Differences in Reward‐Based Learning Predict Fluid Reasoning Abilities

“…All the models included the same three network nodes, corresponding to the default ACT‐R locations for the retrieval buffer (left lateral PFC), visual buffer (left fusiform), and procedural module (left head of the caudate nucleus, part of the striatum). As is common when using the ACT‐R module‐to‐brain mappings (Anderson, 2005, 2007; Borst & Anderson, 2013; Borst, Taatgen, Stocco, et al, 2010; Dimov, Khader, Marewski, & Pachur, 2020), the location and size of these regions were maintained constant across participants. Because the coordinates of these modules were originally defined in Talairach space (Anderson, 2007; Anderson et al, 2008), new coordinates in MNI space were defined using the transformations proposed by Brett, Christoff, Cusack, and Lancaster (2001).…”

Process and content in decisions from memory.

“…Another insight is provided by memory models like search of associative memory (SAM; Raaijmakers & Shiffrin, 1981), context maintenance and retrieval (CMR; Polyn et al, 2009), and DECISIONS FROM MEMORY 5 more general cognitive architectures like adaptive control of thoughtrational (ACT-R; Anderson et al, 2004; for ACT-R models of decision-making, see also Dimov et al, 2019;Marewski & Mehlhorn, 2011;Fechner et al, 2018;Gonzalez et al, 2003). These theories propose that the items retrieved from memory are used to update a dynamic context representation (or alternatively a short term memory store, declarative memory or imaginal buffer, or working memory representation), which guides subsequent retrieval and interacts with other processes in the cognitive system.…”

“…Such actions include memory probes to obtain further information, logical rules for reasoning, and motor actions to indicate responses. In the domain of decision making, ACT-R based models have been used to implement a variety of decision strategies and evaluative processes that aggregate information in different ways, addressing different paradigms and settings that require high-level judgment and decision making (Dimov et al, 2019;Fechner et al, 2016;Link & Marewski, 2015;Link et al, 2016;Marewski & Mehlhorn, 2011;Marewski & Schooler, 2011;Schooler & Hertwig, 2005). At the core of these models is the use of intermediary cognitive buffers (either the retrieval buffer for the declarative memory module or the imaginal module) to represent dynamically evolving task-relevant information.…”

“…and various production rules operate on the content of these buffers to guide deliberation. With this approach it is possible to describe multiattribute decision heuristics like take-the-best, (Dimov et al, 2019;Marewski & Mehlhorn, 2011) and also specify their relationship with the decision environment and other task-relevant variables (Marewski & Schooler, 2011;Schooler & Hertwig, 2005). This work is also important in its use of ecological data (e.g., natural language statistics) to model the accessibility of items in memorythis allows for the application of the models to certain types of naturalistic decision problems involving real-world concepts (Marewski & Schooler, 2011;Link & Marewski, 2015;Link et al, 2016; also see Anderson &Schooler 1991 andGoldstein, 2002) Another important theoretical framework relevant to this paper is query theory (Johnson et al, 2007;Weber et al, 2007;Hardisty et al, 2010).…”