Brain-sensing and customization to adjust your job difficulty
Everyone has days where they are overwhelmed by their job or bored to death. So, what if there was a machine that could link up with your computer to personalize the difficulty of your job to be just right? Then, your job would not be too difficult, but it also would not be so easy that you find yourself binge watching another Netflix show while working. Afergan et al. published a paper in 2014 that shows a computer can customize your workload based on how your brain is reacting.
Past research focused on using metrics like how fast a user completes a task or how many steps it took them to reach a goal. These metrics are not ideal as it does not give true insight into what is happening with the user. That is where the brain data comes in!
A small machine called fNIRS can record the brain’s activity in real time, making it perfect for adjusting the difficulty of work as you go. The machine is small enough to fit on a desk or cart and the sensors used are lightweight and do not require a gel to put them on the head. Even better, unlike other brain tools, fNIRS users do not need to remain motionless for the data to be clear enough for the computer to understand.
The goal is for users to achieve flow. Flow is defined by psychologist Mihaly Csikszentmihalyi as the balance between the user’s skills and challenges in a task that achieves complete immersion and focus. Think of a time when you were on a roll, really getting work done for a few hours without realizing the time disappeared, whether that was crafting, cooking, reading, or your day job. With the evolving research around brain based interventions, you could achieve that flow in virtually any computerized task.
To prove that fNIRS could help individuals achieve flow in real-world tasks, and not just in strictly controlled psychology experiments, Afergan et al. had participants plan out movement paths for military drones in a simulation. If the participant’s brain data indicated they were bored or underwhelmed, more drones would be added. If the data showed that they were anxious or overwhelmed, drones would be taken out of their command. This is how the task was made more difficult or easier depending on how the participant reacted.
Instead of using a “catch-all” or “one size fits all” machine learning model, the scientists customized the models for each individual. They had individuals complete psychological tasks that mimic high workload environments and low workload environments. Then, the recorded brain data was modeled with machine learning to develop a unique signal recognition to understand when that particular participant was feeling bored or stressed. While this would make it more difficult to share the technology within a household or office, it is more effective because of the individual customization.
The drone path designation task showed that participants who had an adaptive simulation made less errors (flying a drone into an obstacle) on average than participants with a non-adaptive simulation. This means that participants were likely to be more engaged and in that flow state with an adaptive simulation compared to a more traditional simulation. Not only did participants make less mistakes with the adaptive system, their mistakes lasted for less time. Therefore, brain data really can improve a person’s work flow and success by keeping them in an ideal state.
In the future, we can expect to see more brain based programs coming onto the market. The algorithms described in this study will lead the way in developing software to improve the efficiency of computerized tasks for all sorts of users based on if they are bored or stressed by their job. It will be a worthwhile investment for companies to look into implementing in their own offices.
Afergan, D., Peck, E. M., Solovey, E. T., Jenkins, A., Hincks, S. W., Brown, E. T., Chang, R., Jacob, R. J. (2014). Dynamic difficulty using brain metrics of workload. Proceedings of the 32nd Annual ACM Conference on Human Factors in Computing Systems — CHI 14. doi: 10.1145/2556288.2557230