Humanity’s evolutionary superpower is our behavioral flexibility. As we go through life, we learn how to navigate the world, building a store of knowledge, habits, and policies that have served us well in the situations we’ve encountered. But there will always be new scenarios that may require new solutions—something we’ve never done before or even thought of doing before. Such scenarios require “thinking outside of the box,” and when we need to do that, we draw on an unlikely resource: a little bit of randomness in the brain circuits that offer up options for action.
We have the cognitive capacity to adapt to complex, unpredictable scenarios across all kinds of new environments. But that flexibility—the almost infinite range of behaviors open to us at any moment—creates a problem. How do we narrow down the options and choose the best one?
It’s no use considering literally every possible thing we could do in every situation we encounter—we’d never finish trying to decide. What we need to do is narrow down the search space: come up with a few decent options and evaluate them for the best one. That is exactly what learning enables us to do.
As we grow and explore the world, we gradually accumulate knowledge and build up a model of how the world works. We learn about the properties and relations of objects in the world—especially what we can do with them, or what they can do to us. We learn about sequences of events that tend to follow each other. And, especially, we learn about the consequences of our own actions across all kinds of scenarios.
When we encounter a new situation, our behavior will therefore be more or less informed by all this prior knowledge, depending on how familiar the situation is. For very familiar ones, we may know very well what the best thing to do is. We don’t need to waste time and effort thinking about it because we’ve already done all that work—it just becomes habitual or automatic. That’s super efficient.
But for situations that are at least somewhat novel, we have to think a bit more about what to do. We need to draw on all that prior knowledge to come up with a few possible actions that we can then evaluate. At some level, these ideas just “pop into our minds”. But despite what some free-will skeptics argue, this doesn’t mean that you are not involved in generating them. They arise intuitively from largely subconscious processes that conduct a search of possible actions, a search that is informed or biased by the implicit model of the world that you have constructed from your past experiences.
Read More: How Listening to Silence Changes Our Brains
The pioneering psychologist William James incorporated these ideas into what he called a two-stage model of free will. Importantly, there’s a little bit of randomness—and yes, a bit of freedom—at play in this search process, which affects the options that actually spring to mind. But once these subconscious processes have generated a set of possible actions, they are submitted to systems of evaluation so that we can exercise our will in order to select one of them, based on how well we think they’ll turn out for us. As Robert Doyle summed up James’s position: “Our thoughts come to us freely; our actions go from us wilfully.” The ideas that spring to mind are effectively competing to “grab the wheel” and drive action. Our evaluation systems allow us to simulate the likely outcomes from the range of suggested actions, evaluate the predicted utility of those outcomes with respect to all our goals, and select one of them for execution, while inhibiting all the others. Both the suggestive and the evaluative systems are thus configured to enable us to do things for our own reasons.
But what happens when we don’t have any good reasons? When we really don’t know what to do? This could be the case, for example, when we encounter a truly novel scenario, when circumstances are changing and our world model is no longer reliable, or when our current behaviors are just not achieving our goals. In those circumstances, we may need to come up with some new ideas.
That’s when we can turn to specialized brain systems, ones that rely on the noisiness of networks of neurons. When we’re frustrated in our goals, regions in the front of our brains—the prefrontal and cingulate cortex—register this fact and send signals to regions in the brainstem that regulate our behavioral states. One of these regions is the locus coeruleus—the “blue spot”—deep in the brainstem. Neurons in this region project throughout the brain and release the neurotransmitter known as noradrenaline, signalling that our current model of the world is no longer working, increasing arousal and vigilance, and focusing attention on areas of volatility or uncertainty. In particular, this can be a signal that we need to do something different.
In the cerebral cortex, options for one possible action or another are encoded by patterns of neurons that are active together. In any given situation, rival ensembles compete within a larger set of neurons. Because neurons that fire together, wire more strongly together, patterns that are regularly active become well-worn—it becomes more likely for the population of neurons to settle into that pattern. They wear a rut, becoming habits of thought.
But those systems have to be flexible and responsive to changing conditions. It’s adaptive, in a changeable world, to not just lock into doing one particular thing, but to be a little bit variable—to at least have the scope to explore other possibilities every once in a while. These circuits are therefore a little bit noisy and poised in what is known as a critical state—they can shift from one pattern to another in response to new information, or sometimes just at random. The noradrenaline released from the locus coeruleus increases the noisiness of all the neurons in those cortical regions. This “resets the network,” shaking up the system and allowing it to resettle into possibly novel patterns.
In machine learning, this is known as raising the “temperature” of the system, a well-known method to shake systems out of locally optimal, but globally suboptimal states. In our own brains, it enables us to expand the search space for options of what to do—to think outside the box. As Nobel laureate Linus Pauling said, “The best way to have a good idea is to have lots of ideas”. But you still need a system to tell the good ones from the bad ones. These options are thus subjected to the normal systems of evaluation so we still choose what we eventually do.
This mechanism of creativity is similar to how evolution allows exploration of new forms: by generating variation at random and then subjecting it to selection. The same principle applies in the immune system in the generation of antibodies. In all these cases, a controlled source of randomness is used as a creative resource, in combination with a powerful selective filter. More error and trial, than the other way around. Thus, even when we choose to use this noisy idea generator to expand our possible options, our final actions are still very much up to us.
