Supplementary Data for: Bridging the Cognitive Gap in Task Mining: Identifying Physiological Metrics for Cognitive Load Estimation in Office Work Environments

This repository contains the supplementary datasets associated with the research paper titled "Bridging the Cognitive Gap in Task Mining: Identifying Physiological Metrics for Cognitive Load Estimation in Office Work Environments". Two complementary datasets are deposited, corresponding to the two axes along which the study orders the literature. The FullTable_Supplementary Material_Physiological_Metrics_CWL_Office_RQ1_RQ2.xlsx file covers the first: which physiological metrics the reviewed literature reports as estimators of Cognitive Load (CL), and how strongly each is supported. Contents. The Excel file consolidates 183 physiological metrics grouped into 11 physiological categories, extracted from the 54 documents included in a tertiary review of 88 retrieved records. It is the primary source used to generate Figures 2 to 7 and the detailed tables of Appendix A of the manuscript. The Table sheet contains one row per metric, with the following variables: Metric: The name of the physiological metric as reported in the literature (e.g., "Pupil Diameter (PD)"). Category: The physiological grouping the metric belongs to (e.g., "Eye Tracking"). Subcategory: The specialized subgroup within the category, where applicable (e.g., "Pupillometry"). ID: A unique identifier assigned to each metric (e.g., "EYE01"), used throughout the manuscript to cross-reference metrics between text, tables and figures. Sources: The number of reviewed documents reporting the metric. CL Relevance (1, 2, 3): A three-level label indicating the strength of the literature support for the metric as a CL estimator, where 3 corresponds to the highest level. The label follows a case-based procedure combining two premises: whether the number of reporting sources exceeds the mean for the metric's category, and whether at least two thirds of those sources report consistent uses and effects. Publication Year Median: The median publication year of the documents reporting the metric. A secondary block on the same sheet reports two statistics per category, Mean value of sources per metric and Median Year of Sources, followed by a numbered legend of the 54 reviewed documents with first author and publication year. The category mean is the threshold used by the first premise of the CL Relevance procedure, so this block is what makes the labels recomputable. Context and Usage. Researchers and practitioners can use this dataset to identify which physiological indicators of CL are best supported by the literature, and to reproduce or recompute the CL Relevance labels under alternative thresholds. Of the 183 metrics, 37 attain the highest label, 69 the intermediate level and 77 the lowest, and the dataset makes the distribution behind that classification inspectable metric by metric. The FullTable_Devices_RQ3.xlsx file covers the second axis of the study: with which devices the physiological metrics can be acquired, and which of those instruments a worker can use without interrupting the task. It complements the metric catalog deposited alongside it. Contents. The Excel file contains two sheets. The Table sheet consolidates 48 device types distributed across the same 11 physiological categories used in the metric catalog, drawn from 164 distinct manufacturer-and-model designations reported by 20 of the 54 reviewed documents. The Criteria sheet defines the three suitability criteria and the rules applied when grading. This dataset is the primary source used to generate Tables 3, 4 and 5 of the manuscript and the full device catalog of Appendix B. The Table sheet contains one row per device type, with the following variables: Category: The physiological grouping the device type belongs to. Consolidation was performed within categories, so a single commercial product that measures signals of two categories yields two device types. Subcategories / Metrics: The subcategories or metrics of the companion catalog that the device type can acquire. This column is what allows the two datasets to be cross-referenced. Sensor Device Type: The consolidated acquisition technology (e.g., "Smart ring (PPG)"), which is the unit of analysis of this catalog. Example models: The manufacturer-and-model designations reported in the reviewed documents for that device type. C1 (Contact and placement): Graded A, B or C according to the physical demands the device places on the worker, from contactless instrumentation to devices that prevent the activity from being performed normally. C2 (Operability): Graded A, B or C according to whether the worker can operate the device unaided, after simple preparation, or only with a trained technician. C3 (Provisioning): Graded A, B or C according to what the organization must supply, from hardware already present at the workstation to equipment requiring dedicated infrastructure or professional installation. Representative Image: A photograph of a representative product for each device type, embedded in the sheet. References: The reviewed documents reporting each model, listed in the same order as the models themselves. Grading conventions. The three criteria are reported independently rather than collapsed into a single score, since they answer different questions: two about the worker and one about the organization. Grades characterize the device type of a row rather than any individual model, so a type that groups models with differing demands is graded by the most restrictive case documented within it. Under this grading, 15 of the 48 device types receive no grade C in any criterion and are therefore deployable during ordinary office work. Context and Usage. Researchers and practitioners can use this dataset to determine which physiological signals are actually recordable in a given work setting, and at what cost to the worker and to the organization. Crossed with the metric catalog, it identifies which of the best-supported CL estimators are within reach of office instrumentation and which are not. Readers who disagree with a particular grade can recompute the suitability profile of any device type directly from this file without re-executing the extraction, which is why the three criteria are reported separately rather than aggregated.