A Theory of Decision Making Under Dynamic Context

Authors: Michael Shvartsman, Vaibhav Srivastava, Jonathan D. Cohen

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We simulate 100,000 trials for each model. Figure 1 shows results from the simulation of the flanker task, recovering the characteristic early below-chance performance in incongruent trials. This simulation supports the assertion that our theory generalizes the flanker model of [5], though we are not sure why our scale on timesteps appears different by about 5x in spite of using what we think are equivalent parameters. For the AX-CPT behavior, we compare qualitative patterns from our model to a heterogeneous dataset of humans performing this task (n=59) across 4 different manipulations with 200 trials per subject [24].
Researcher Affiliation Academia Michael Shvartsman Princeton Neuroscience Institute Princeton University Princeton, NJ, 08544 ms44@princeton.edu Vaibhav Srivastava Department of Mechanical and Aerospace Engineering Princeton University Princeton, NJ, 08544 vaibhavs@princeton.edu Jonathan D. Cohen Princeton Neuroscience Institute Princeton University Princeton, NJ, 08544 jdc@princeton.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes A library for simulating tasks that fit in our framework and code for generating all simulation figures in this paper can be found at https://github.com/mshvartsman/cddm.
Open Datasets Yes For the AX-CPT behavior, we compare qualitative patterns from our model to a heterogeneous dataset of humans performing this task (n=59) across 4 different manipulations with 200 trials per subject [24]. [24] O. Lositsky, R. C. Wilson, M. Shvartsman, and J. D. Cohen, A Drift Diffusion Model of Proactive and Reactive Control in a Context-Dependent Two-Alternative Forced Choice Task, in The Multi-disciplinary Conference on Reinforcement Learning and Decision Making, pp. 103 107, 2015.
Dataset Splits No The paper mentions comparing its model to a human dataset and using simulations, but it does not specify explicit training, validation, or testing splits for the data used in the experiments.
Hardware Specification No The paper describes simulations and modeling but does not provide any specific details about the hardware (e.g., GPU, CPU models, or memory) used to run the experiments.
Software Dependencies No The paper mentions a library for simulating tasks and code, but it does not specify any software dependencies with version numbers (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes The remainder of parameters are identical across both task simulations: σc = σg = 9, θ = 0.9, µc = µg = 0 for c0 and g0, and µc = µg = 1 for c1 and g1. To replicate the flanker results, we followed [5] by introducing a non-decision error parameter γ = 0.03: this is the probability of making a random response immediately at the first timestep. We simulated 100,000 trials for each model.