Your Mistakes May Have Real Structure: New Brain Research

We all, at times, blame the mistakes we make while juggling multiple tasks on attention lapses or our mental capacities being stretched. Multitasking: art or skill or road to ruin, you decide. Cheng Xue, PhD, a neuroscientist and postdoctoral associate at The University of Chicago in Chicago, thought the same thing.“Until we actually did a close study on it, I did not realize that there’s a structure in these mistakes that I make,” he said.In a new study published in Nature Neuroscience, Xue, Marlene Cohen, PhD, and colleagues found that in moments of uncertainty, neural activity from separate tasks starts to overlap. The researchers call it “feature interference.” Their findings suggest that multitasking mistakes aren’t due to a shortage of brainpower but rather the crosstalk within the brain.Predictable MistakesThe research team had more than 200 people complete a cognitive multitasking simulation by showing them a series of images on a screen. The images contained striped circles that had two features that could change: where the circle appeared on the monitor (left or right) and the thickness of its stripes. The participants were shown one circle with stripes, followed by another with one of those features subtly changed. The researchers then asked which feature changed, then repeated the exercise with different changes.The thinking behind this strategy was that there could be groups of neurons dedicated to analyzing position, and another dedicated to analyzing stripes, and for optimal efficiency, the brain would keep those channels separate. If those channels stay separate, but mistakes happen because the brain is tired or distracted, the data would show those errors happen at random. However, if the decisions leak into each other, patterns would emerge that occur more often than random chance across participants — for example, they might see that people say a circle shifted right when the stripes were thinner.The team also trained two rhesus monkeys to perform similar tasks, and while they were performing them, recorded activity in their primary visual cortex and parietal cortex — part of the brain that helps with task switching. The recordings indicated that when monkeys consistently got answers right, separate groups of neurons were involved. But when monkeys got a series of wrong answers, the neuronal activity overlapped.Xue said they were surprised to find that the errors in both humans and monkeys weren’t random but followed a pattern. He compared it to a math test. “So the same mathematical problem, when you give it to two kids, maybe they got it wrong. But from their wrong answers, you can infer how they are doing it in their head,” he said.To better understand the wrong answers, the team trained an artificial neural network to behave like the people and monkeys. “We just take the monkeys’ and humans’ decisions, and we train the network to reproduce their actions. And so this becomes like a specimen of the brain for us that we can dissect and investigate and see what’s happening,” Xue said.The AI model, which the team tested on the data from people and monkeys, showed that under task uncertainty, information from the irrelevant task impacts the neural activity of the current task.The finding is counterintuitive, Xue said. Each task only requires a tiny portion of the brain’s capacity. “It’s just two features, and we have billions of neurons in our brain,” he said. It isn’t that our brain has limited resources, he noted, it’s the way that tasks bleed into each other.He gave an example of writing an article about AI at a desk with a PlayStation on it — the console might inadvertently become a metaphor in the writing.Mistakes Are InevitableEven though the researchers understand the mechanism, they’re not any better at avoiding it.When asked about feature interference in his own life, Xue said, “I mess up so often.” That’s why the mechanism seemed surprising to him: being stretched in multiple tasks feels like hitting a resource limit, not like two tasks mixing together.Tim Buschman, PhD, a neuroscientist at Princeton University in Princeton, New Jersey, who was not involved in the study, has spent his career studying why our minds struggle to juggle information. He, too, has experienced information overload. “Oh, it always happens. There’s no avoiding it,” he said.One place he actively tries to fight it is in how he communicates his research on why we can only hold a few things in working memory at a time. “We study working memory capacity, and every time you put a comma, that’s adding something to working memory capacity, and it’s just going to reduce somebody’s ability to understand what’s going on,” he said.This new study has Buschman questioning some of his own lapses, too — times he would have blamed attention or noise but now thinks may be feature interference. “Normally I would have chalked it up to something else, but now I’m wondering: is it actually this that’s underlying it?”Old Problem, New ExplanationResearchers have long observed limits on cognitive flexibility, Buschman said, and have offered many explanations. What is different about this one is that different task features, like position and stripe size, can be tangled in our brains instead of staying separate.“These things are always thought to be represented independently because that is the ideal way that you would want to represent information,” he said. He found it striking that they don’t stay separate, and that this separation drives interference. He was also impressed that the team backed it up several ways: with neural network models, causal stimulation, and neural recordings. And he said the task design let the team argue against alternatives like attention lapses or noise.Buschman still wonders, though, why the brain works this way. When the researchers trained the AI network to perform the task optimally, it didn’t show the effect, which means that monkeys and humans are likely doing something suboptimal.“There is usually a normative reason for why something like this exists,” Buschman said. The interference may be the brain’s trade-off across the range of tasks it has to manage. The authors suggest it could instead be a byproduct of how brain regions pass information to one another.What We Can Do About ItFor clinicians and researchers, mistakes can be treated as data. Cognitive flexibility declines across many psychiatric disorders and in aging, Xue said, and different conditions seem to affect it in distinct ways. If researchers could map those error signatures, it’s possible they could develop diagnostic tools, which could also serve to direct interventions.Buschman also noted that understanding the biases in less-than-ideal choices could help engineers build systems, like airplane cockpits, that present information in ways that guide and support optimal choices.For our everyday life, Xue said the brain needs time to shift from task to task. He recommended leaving buffer time between tasks when you can and switching to doing something simple like tidying a table after cooking and before writing, so that the multitasking involved in cooking doesn’t leave a footprint that carries into the next task. A few moments of meditation between tasks, he said, could also help.Buschman said to ask yourself what’s driving a decision and whether it’s a good one. He said what matters is “being aware that you can have these biases.”Also Buschman recommended using fewer commas to make written communication less likely to get caught up in the brain’s crosstalk. He tells colleagues to “put a period” on it.Buschman reported having no conflicts. Disclosure information for study authors is available in the original study publication.