Stressed out? The Sum of the Parts May Help...

Clenched fists, beating heart, quickened breathing. We know how that feels. We experience it quite frequently, especially when the deadline is too close for comfort. Stress may have debatable benefits, but it is no doubt an awkward stalker, peering over our shoulders. That personally makes me squirm and it becomes difficult to focus. Besides, the more silent and stealthy stress is, the more deadly it can be.

This week, as we discussed stress in Psychology, we were challenged to come up with devices, tools or solutions to help cope with stress. After all, stress is a daily challenge for many, and deserves a more solid answer than ‘take a break’ or ‘breathe deeply’.

With my newfound exploration in Deep Learning and analyzing data to find patterns, it seemed possible that anything with data can be rummaged through, and that predictions can be made if an approximate trend exists. One of the issues with stress is its sinister silence. With our minds preoccupied with work, our eyes gazing frequently at the ticking time, knees bouncing, fingers frantically typing away, I find it extremely easy to not realize my steadily increasing heart beat, or that my mind wants to succumb to ‘fight-or-flight’. To address this issue, I realized that we in fact do have access to data to make a prediction. Theoretically, a deep learning model could be trained on historical data, such as heart rate, blood pressure, perspiration, body temperature, all of which can be measured through fitbits or Apple watches, and predict with some certainty whether that moment in time represents a stressful situation or not.

More specifically, suppose a rising heart rate coupled with rising perspiration and blood pressure are indicative of a stressful situation, the machine learning model can be fed hundreds of thousands of data entries, with different combinations of heart rates, blood pressures, and perspiration rates and whether or not the situation was stressful. The model then learns what is considered stressful and what is not. Since these thresholds vary between individuals, the machine needs to be calibrated to each person. However, once the model is able to predict how stressful the fitbit-wearer is feeling, the individual can be alerted that they are indeed stressed out physically. The model can also be trained such that it can predict how stressed the individual is feeling, and intelligently figure out the proper method of communicating that.

With this technology, it is possible to know whether the individual is feeling stressed or not, however, communication plays an extremely important role. Telling someone who is stressed-out that they are stressed-out is ineffective, and can worsen the situation as they will be acutely aware of their stressful situation. This is similar to countdowns that are chanted, making the pressure heightened to beat the clock.

In 2019, a new wearable gadget called AlterEgo was created by a team of MIT researchers, with Arnav Kapur as the team lead. The wearable gadget used artificial intelligence to translate thoughts the individual wanted to say, though did not, and could send a message back through bone conduction. The result was a direct communication between the mind and the internet with no speech nor sound involved. This was essentially to help people who struggle with speech-communication, whether it be because of loss of sight, hearing or speech. Not only was the technology useful for helping people in need, but represented the opportunity to communicate mind to mind, instantaneously.

This ‘mind-speaking’ technology could be utilized to help calm down stressed-out students. When other people remind individuals that they should calm down, or that they should not be agitated, or even try providing positive messages to motivate them, especially in stressful situations, emotions cloud judgement and lead to ineffective communication. However, by directly communicating with the mind, communication can seem as though it is occurring within the mind. The boundary between what is an original thought and what was a message becomes fuzzy, allowing the stress-predictor to communicate positive messages, or provide a calming effect within the mind without worsening the situation.

Evidently, there are serious ethical implications with trying to manipulate or trick the mind into being more positive or less stressed since the same technology could be used for the reverse which would certainly lead to more stress, and long-term consequences. Technology, such as AlterEgo, require more safety-nets before it can be released to the public, but those safety requirements could hinder the modes of addressing the issue.

Another technical limitation would be that the physical signs of stress can be induced because of other reasons, such as, exercise or strong emotions (frustration, arguments, fear). In addition to understanding what the individual is doing and whether or not the stress alerts are false-positives, calibration is a challenge since some people might have conditions where their heart rates increase and decrease much faster than normal, or if they perspire different amounts under stress. The calibration would require monitoring the individual for a certain amount of time, and adjusting the instrumentation to personalize the stress thresholds.

In this suggestion, there are several components of Artificial Intelligence. The first is analyzing biological data to figure out if the situation is stressful or not. Then, the cause of this stress must be somehow communicated to the system, or figured out through intelligent and unknown means. Intelligently figuring out the ‘right-thing-to-say’ for communication purposes again requires machine learning. Then, translating an action into words and perhaps proper tone and sentiment requires Natural Language Processing. There are even more components that can be automated to become intelligent, but truly, this is a mix of existing technologies. We have the means to collect biological data from a user, therefore allowing us to monitor the situation closely. As I have mentioned, AlterEgo allows mind-mind or mind-device communication instantaneously without speech. Understanding what is the appropriate action to take and encoding it into speech is what Siri and Alexa currently do, using NLP to understand a user’s intention, automating the task, then providing feedback to the user. If we can combine these individually existing parts, that is a stress-predictor.

This week, while continuing my progress in Linear Algebra, we were learning about matrix multiplication. There are several ways to visualize it, one being ‘blocking’. When matrices can grow tremendously large, it becomes difficult to handle so many rows and columns.

Therefore, by breaking the bigger matrix into smaller partitions, the result matrix can be built one piece at a time. Similarly, by building on existing technologies, and constructing a potential solution from the sum of the individual parts, a stress-predictor could become the next crutch to help us eventually adopt healthier habits, such as focusing on breathing or taking a walk outside.