Data and models for "An image-computable model of speeded decision-making"

Lost in Migration gameplay data and trained models for: Jaffe, P. I., Gustavo, X. S. R., Schafer, R. J., Bissett, P. G., Poldrack, R. A. An image-computable model of speeded decision-making. eLife 13, RP98351 (2024). This dataset can be used to reproduce all of the results of the manuscript, following the instructions in the code repository for the paper: https://github.com/pauljaffe/vam. The dataset includes the following components: gameplay_data.zip: Trial-level gameplay metadata for Lost in Migration. Lost in Migration is a variant of the flanker task offered as a part of the Lumosity cognitive training platform (Lumos Labs, Inc.). The .zip file includes a separate .csv file for each of the 75 Lumosity users (participants) that we trained models on. Each .csv file has one row per trial with the following fields/columns: "anon_id", numerical identifier for the Lumosity user; "nth_play", the nth gameplay of Lost in Migration for this user; "trial", the nth trial for the current gameplay; "xpos", the signed horizontal distance from the center of the target bird to the left edge of the game window (pixels, non-negative); "ypos", the signed vertical distance from the center of the target bird to the bottom edge of the game window (pixels, non-negative); "flanker_direction", (L/R/U/D); "response_direction", (L/R/U/D); "target_direction", (L/R/U/D); "response_time", (ms); "stimulus_layout", numerical code for the layout of the bird flock for the current trial (0: horizontal line, 1: vertical line, 2: cross, 3: <, 4: >, 5: v, 6: ^). vam_models.zip: Parameters for the 75 visual accumulator models (VAMs) analyzed in the manuscript. task_opt_models.zip: Parameters for the 75 task-optimized models analyzed in the manuscript. metadata.csv: Metadata for each Lumosity user that a VAM/task-optimized model was trained on. The .csv file has one row per user with the following fields/columns: "user_id", numerical identifier for the Lumosity user (same as "anon_id" in gameplay_data.zip); "gender", self-reported gender ('m', 'f', or null, indicating no response was given); "binned_age", age bucketed into decade-long bins (20-29, 30-39... 80-89). derivatives.zip: The RTs/choices generated by the trained models, organized into separate folders by model type (vam/task_opt/binned_rt) and user ID. Also includes a "summary_stats" folder with analysis products of the model activations and outputs. graphics.zip: Image files used to create the visual stimuli from the gameplay metadata. example_model_inputs.zip: The processed visual stimuli and gameplay data used as inputs to train one model (user ID 182). Note we provide instructions to recreate the stimuli and other model inputs for all models in the code repository.本数据集包含《迷失迁徙》（Lost in Migration）的游戏玩法数据与训练模型，相关研究论文为：Jaffe, P. I., Gustavo, X. S. R., Schafer, R. J., Bissett, P. G., Poldrack, R. A. 《基于图像可计算的快速决策模型》，发表于eLife 13, RP98351 (2024)。 本数据集可配合论文代码仓库中的操作指南复现该手稿的全部研究结果，代码仓库链接：https://github.com/pauljaffe/vam。 本数据集包含以下组成部分： gameplay_data.zip：包含《迷失迁徙》的试次级游戏元数据。《迷失迁徙》是Lumosity认知训练平台（卢莫斯实验室股份有限公司，Lumos Labs, Inc.）推出的侧抑制任务（flanker task）变体游戏。该压缩包为本次建模所用的75名Lumosity用户（参与者）各提供一份独立CSV文件。每份CSV文件以试次为行，包含以下字段/列："anon_id"，Lumosity用户的匿名数字标识符；"nth_play"，该用户游玩《迷失迁徙》的第n次；"trial"，当前游玩中的第n次试次；"xpos"，目标鸟类中心到游戏窗口左边缘的有符号水平距离（像素，非负值）；"ypos"，目标鸟类中心到游戏窗口下边缘的有符号垂直距离（像素，非负值）；"flanker_direction"，侧翼刺激方向（L/R/U/D，即左/右/上/下）；"response_direction"，被试反应方向（L/R/U/D）；"target_direction"，目标刺激方向（L/R/U/D）；"response_time"，反应时（毫秒）；"stimulus_layout"，当前试次鸟群布局的数字编码（0：水平线，1：垂直线，2：十字形，3：<形，4：>形，5：倒V形，6：正V形）。 vam_models.zip：包含本手稿中分析的75个视觉累积器模型（Visual Accumulator Model，VAM）的参数。 task_opt_models.zip：包含本手稿中分析的75个任务优化模型的参数。 metadata.csv：包含为其训练了VAM/任务优化模型的每位Lumosity用户的元数据。该CSV文件以用户为行，包含以下字段/列："user_id"，Lumosity用户的数字标识符（与gameplay_data.zip中的"anon_id"一致）；"gender"，自我报告的性别（'m'为男性，'f'为女性，或null表示未填写）；"binned_age"，按十年为单位划分的年龄组（20-29岁、30-39岁……80-89岁）。 derivatives.zip：包含训练模型生成的反应时与选择结果，按模型类型（VAM/任务优化/分箱反应时）与用户ID分为独立文件夹。此外还包含"summary_stats"文件夹，内含模型激活与输出的分析产物。 graphics.zip：包含用于根据游戏玩法元数据生成视觉刺激的图像文件。 example_model_inputs.zip：包含用于训练单台模型（用户ID 182）的预处理视觉刺激与游戏玩法数据。需说明：本数据集仅提供该示例输入，所有模型的刺激与其他输入的复现步骤详见代码仓库。