In our modern world, technology advances at an unprecedented pace. As the tools available to us get more sophisticated, we find ourselves needing to reinvent ourselves, learn new knowledge, and switch contexts more frequently. All of these aspects can contribute to the cognitive load we are experiencing (but also feeling?) daily, which puts our health at risk. Thankfully, the challenges we face when assessing our cognitive load can be mitigated. The new and rare study paradigm presented here was balanced across cognitive load levels, the time recorded in controlled laboratory and uncontrolled real-life environments, and incorporated both subjective labels as well as multi-modal physiological signals from twenty-four knowledge workers, totalling over 300 hours of data. Join our collective journey towards maintaining health and download the dataset now! To learn more about the background of this study, continue reading.

Have you ever stopped to ask yourself how high your cognitive load (think of mental workload) was last week?

Could you confidently say that it has been higher that week than the same week last year? Usually, we are too busy to reflect. Rarely, if ever, do we stop our current tasks and start to think about how we work. By infrequently caring about our mental state, we risk a part so vital to our lives: our mental health. Yet our mental health is directly connected with our physical health. Hence, by rarely caring about our mental states, we risk our health, and in the ultimate consequence our lives. Shouldn’t we be internally motivated to prioritize maintaining our health rather than merely managing the diseases developed by neglecting our mental health?

Cognitive load can be assessed through three avenues.

An individual can be a) asked to subjectively and constantly rate their mental state, b) evaluated by experts that constantly monitor this individual, or c) objectively monitored using technology. However, each option has pitfalls. As such, a) individuals can be faced with recall bias or answer dishonestly, b) the amount of experts worldwide does not suffice to monitor all the individuals on this earth, and c) technology might be used to exploit individuals. While many foundational research efforts focus on investigating the mental states of individuals in controlled laboratory environments and thereby advance the field tremendously, an important aspect needs to be addressed: Life is not confined to laboratories.

More life happens outside than inside of laboratories.

Acknowledging this reality, and building on technological advances in measurement technologies, it became evident that a new study paradigm was needed that incorporated the best of both worlds: the firm control over tasks and high-quality data of traditional laboratory environments, combined with the flexibility to capture meaningful data from relevant real-life tasks. In turn, a new study paradigm with multi-modal recordings in both laboratory and real-life situations was developed. Twenty-four participants were equipped with wearable sensors capable of collecting electroencephalography (EEG), gyroscope (GYRO), accelerometer (ACC), photoplethysmography (PPG), skin temperature (TEMP), and electrodermal activity (EDA) to collect this rare and therefore influential dataset. Furthermore, the sensors utilised, a four-channel EEG headband and a wristband, allowed for a rather unobtrusive data collection, and most participants recorded data for eight hours.

The early-stage insights gained so far are extensive and multifaceted.

This unique and comprehensive dataset can be analysed for numerous research questions. As such, researchers can investigate whether different modalities influence each other, analyze the time effects of cognitive load, evaluate which signal quality indices are effective in both controlled and uncontrolled environments, or even develop a wearable personalised mental state management application. Additionally, the dataset provides an opportunity to explore new algorithms, such as merging multi-modality or addressing data science questions. The potential for this dataset is vast, and the usage options are not limited to the examples mentioned.

The data quality is good, so engage with this dataset, and join our collective efforts towards a healthier world.

Besides describing the research methodology to the level of enabling replicability and reproducibility, the paper published in Scientific Data (https://www.nature.com/articles/s41597-024-03738-7) details participants' demographics, tasks performed, data synchronisation processes, and data quality analyses including signal-to-noise ratios and signal-quality-indices. Statistical and foundational machine learning analyses further demonstrate the technical robustness of the dataset. This dataset is more than just the multi-modal signals and labels provided; it builds the foundation for innovative research that was previously impossible due to the lack of such records across diverse environments. The data was made publicly available via Zenodo (https://doi.org/10.5281/zenodo.10371068). The source code used to build the experiments with PsychoPy (https://www.psychopy.org/), as well as load, process, and analyse the data using Python, was made publicly available (https://github.com/HPI-CH/UNIVERSE) for everyone to build on top of.

Numerous possibilities for exploration of this dataset exist, so get involved!

This dataset is a foundation for innovative research. We invite you to study and learn with this dataset, and to join our collective journey towards maintaining health! Download “A Dataset on Unobtrusive Measurement of Cognitive Load and Physiological Signals (EEG, PPG, EDA) in Uncontrolled Environments” today from Zenodo, expand our codebase, and join us in pushing for a more compassionate and healthier world!