What happens to decisions under sustained stress?

engin-akyurt-Ii0dW4rU77g-unsplash | Cerebra

I was working with a group of Public Information Managers recently who are in the emergency response space, What a great group of people from across the Canterbury region, and across a range of organisations. Completely dedicated to their work.

In brief, I was talking about how cognitive performance degrades under stress. It was a bit of a skim across the highlights but did prompt me to write some more on an aspect we couldn’t really cover in any depth, so here you go.

As we get deeper into sustained stress, a natural disaster for them, for example, the decisions that appear the most confident can well be the ones that need the most scrutiny.

I wrote a few weeks back about acute fatigue, including the research showing that around 17 hours awake produces impairment comparable to being over the drink-drive limit. Acute fatigue is a relatively easy case. It’s visible, measurable and generally easy to roster for. Sustained stress is a different argument. Nobody might be dramatically impaired on any given day. Rather, everyone is slightly degraded, every day, for a sustained period, and the degradation compounds over time.

The research

A cornerstone study is that by Van Dongen and colleagues, published in Sleep in 2003. They restricted people to six hours of sleep a night for two weeks and measured performance daily. By the end of the study, cognitive performance had declined to a level comparable to going one to two full nights without sleep.

That’s the cumulative part. The more interesting part as I see it is what happened to self-assessment. Subjective sleepiness ratings, that is how sleepy people said they felt, rose for the first few days and then levelled off, while performance kept falling. People adapted to how the load felt, and then trusted that feeling, even as they performed increasingly poorly. They were, in the researcher’s terms, largely unaware of their own deficits. Naturally, the instrument you’d use to detect the problem, asking people how they’re going, is the instrument the problem itself disables first.

That can explain capacity, but not the confident decisions. Here’s where we need stress research. In 2009, Dias-Ferreira and colleagues published a study in Science showing that chronic stress reorganises the frontostriatal circuits that decision-making runs on. Frontostriatal circuits are neural circuits that connect the frontal lobes to the striatum. Their research showed that chronically stressed animals (chronic means longer term stress versus acute, or short-term stress) didn’t stop deciding. Instead, they shifted strategy, from goal-directed decisions that track whether an action still produces the outcome you want, to habitual ones that repeat whatever worked before.

The stressed brain just economises by running yesterday’s routines. The brain does love efficiency. Soares and colleagues found the same shift in humans under sustained, real-world stress, in 2012, with one notable addition.

Six weeks after the pressure lifted, the bias reversed.

Now let’s add one more bit.

Wiehler and colleagues (Current Biology, 2022) showed that even a single day of demanding cognitive work changes what the decision system prefers, shifting choices towards low-effort, short-term options. They proposed a build-up of glutamate in the lateral prefrontal cortex as the mechanism. For them, fatigue wasn’t something to push through, but a change in the brain machinery doing the choosing.

A different angle

Fast, confident, non-deliberative decisions are what expertise looks like. Gary Klein’s recognition-primed decision model, built from studying fireground commanders, showed that experienced decision-makers rarely compare options. They recognise the situation as a type they know, retrieve a course of action that has worked before and run it. And they’re usually right. I’m not arguing against that. (Also, Klein’s work is great.) Having worked with both, recognition is a strength of firefighters, and people in the emergency response space. Treating their speed as a symptom misses the intent here.

Klein’s model requires machinery to run. Brain machinery. And this is where load research comes into play. An expert doesn’t just retrieve an action. There’s more to it. An expert runs a quick mental simulation of it against this specific situation, and they keep monitoring whether events are unfolding in the way the recognised pattern predicts. If the pattern breaks, and expectations are violated somehow, that’s the alarm that the situation isn’t typical after all.

Both of those components, the simulation and the anomaly detection, are effortful work for the prefrontal cortex, and precisely the capacity that fatigue and stress research show is drained first. From the inside, degraded recognition feels just like expert recognition. We might just strip out checks and explain away surprises.

There’s a boundary problem too. Kahneman and Klein, in their 2009 adversarial collaboration, agreed that intuition can be trusted where the environment offers valid cues, and the expert has had the chance to learn them.

In an emergency response, the library of patterns is built on more acute events. Consequently, deeper into a response, with cascading consequences and cumulative depletion, the event has likely drifted outside the conditions someone may have learned on.

Put these pieces together

As sustained load continues, a plan might get defended because it’s the plan, not because we’ve recently tested it against the situation. The range of options may decrease, not because there are genuinely fewer, but because we consider a narrower window. Decisions come faster and feel cleaner, because they’re increasingly retrievals, rather than deliberations. Re-opening a settled question may be seen as unnecessary friction, rather than an evaluative opportunity in a changed context. What is unlikely to change, is the confidence. That’s what sustained load leaves intact. There’s comfort in habits.

Firefighters, and emergency managers, and others like them are extremely dedicated. We need more of them. When numbers are low, dedication becomes staying late, taking the extra shift, waving away relief so others might have it, and so on. These are the things that most reliably defeat the protections we might have in place.

Under sustained load, as we saw a few weeks ago, self-assessment is screwy. So resistance to standing down doesn’t tell you someone is coping; self-assessment research says it’s likely the opposite.

Thus, the answer can’t be self-assessment, nor can it be willpower. It has to be design.

Decisions by design

It could mean rotating decision-makers on a mandatory schedule. We could separate decision types because we know a routine operational call and a novel high-stakes judgment deteriorate differently. The novel ones need the freshest people, not the tired ones, even if they’re senior. Put fresh eyes on significant and complex decisions to help with the mental simulation that a tired brain won’t manage as well, and use senior brains for guidance. Klein’s premortem tool applies here too. Task someone with tracking expectancies, by which I mean whether events are unfolding the way the current plan predicts. If there’s drift, and a gap between expectations and reality, we’ve a signal that the pattern no longer fits. And someone re-opening a decision is a contribution if circumstances have changed and we haven’t seen it because, based on evidence, confidence in decisions is itself a risk.

For balance and transparency, I should highlight some limits. The Dias-Ferreira study was on rodents, and the human evidence, while consistent, is smaller and a bit messier. Van Dongen studied sleep restriction specifically, which overlaps with sustained load but is not identical, and there was an erratum on Van Dongen, later corrected, but which didn’t change the central findings. Wiehler is one day of load, and not three weeks, and the lab-to-field translation always requires care. And I’ve synthesised two Klein literatures for one conclusion.

That said, all of the evidence, and the pattern across it all, points to the same thing: sustained demand pushes people toward efficient, easy, often quick, habits, and away from flexibility, while leaving confidence untouched. The reversibility finding is the hopeful part, and also an element to add to people’s recovery. Capacity comes back when the load comes off. That makes recovery an operational resource to be managed.

Deeper into an emergency response, people may still be confident. The question we need to ask is whether someone is checking the decisions that feel most obviously right.

Photo by engin akyurt on Unsplash