194 - Allostatic Load

I was reminded of the topic of "Cognitive Exhaustion" the other day, primarily by reaching it after several consecutive hours of responding to emails and messages and in reflection realizing that I haven't spoken about it in quite some time. This is a distinctly different concept than "Burnout", as burnout is the depletion of motivation, whereas cognitive exhaustion is the physiological depletion of resources in neurons of the brain, with the latter being replenished daily (given healthy sleep).

The common practice of "multi-tasking" is a well-known way of accelerating this depletion of available resources in the brain, as every time task-switching occurs more resources are consumed. In contrast, this is also part of why programmers with the hyper-focusing ability of ASD are so sought after, as by focusing exclusively on one task those cognitive resources are reliably able to cover far more ground.

In previous years I did some back-of-the-envelope estimating of how much cognitively optimal time humans actually apply to their work on average. One example of this with humans might involve taking employees with an average peak performance of 2.5 hours per day 5 days per week, and a 10% optimal environment, which would result in an effective maximum peak of:

(2.5/24) × (5/7) × 10% = 0.744% (Maximum Peak)

Even if you very generously assumed that your employees were in their optimal work environment 2/3 of the time that would still only work out to about 5%, since humans can't (and shouldn't have to) work 24 hours per day or 7 days per week, and those physiological cognitive resources remain finite each day.

Chemistry proved quite excellent for bootstrapping biological intelligence, but it doesn't allow for optimal performance to be maintained over extended periods absent edge cases like ASD hyper-focus. Those edge cases also make a heavy trade-off, as any interruption does far more harm than it could to interrupting their non-hyper-focusing counterparts.

One of the benefits of human-like intelligence in software is the ability to maintain a greater degree of consistency over time than humans are physiologically capable of. One particularly clever individual I met was able to figure out that an employee of a company he audited was a smoker by their performance curve throughout the day by compressing all data into a single simulated day and recognizing the predictable pattern of nicotine effects and cravings.

When applied to systems where humans add value rather than being the primary source of it, this can effectively mitigate periods of reduced performance caused by cognitive exhaustion. Smoothing these rough spots absent injecting the conflict of other methods can also enable positive feedback loops, improving employee health and satisfaction over time.

Try asking yourself how cognitively exhausted you are at different points in your day, and following different activities. If you track this sort of data then the results might surprise you, offering actionable ways of improving your schedule.