New research shows why you don’t have to be perfect to get the job done

When neuroscientists think about a strategy an animal might use to perform a certain task—such as finding food, hunting prey, or navigating a maze—they often propose a single model that outlines the best way for the animal to accomplish the job.

However, in the real world, animals—and humans—may not use the optimal way, which can be resource intensive. Instead, they use a strategy that’s good enough to handle it, but requires a lot less brain power.

In new research appearing in Scientific advancesJanelia scientists set out to better understand the possible ways an animal could successfully solve a problem, rather than just the best strategy.

The work shows that there are a huge number of ways in which an animal can accomplish the simple task of finding food. It also lays out a theoretical framework for understanding these different strategies, how they relate to each other and how they solve the same problem differently.

Some of these less-than-perfect options for completing a task work nearly as well as the optimal strategy, but with much less effort, the researchers found, allowing the animals to use scarce resources to handle multiple tasks.

“Once you break free from perfection, you’ll be surprised how many ways there are to solve a problem,” says Tzuhsuan Ma, a postdoctoral fellow in Hermundstad’s lab who led the research.

The new framework could help researchers begin to examine these “good enough” strategies, including why different individuals may adapt different strategies, how these strategies may work together, and how the strategies generalize to other tasks. This could help explain how the brain enables behavior in the real world.

“Many of these strategies are ones that we wouldn’t even dream of as possible ways to solve this task, but they work well, so it’s entirely possible that animals can use them as well,” says group leader Janelia Ann Hermundstad. “They give us a new vocabulary for understanding behavior.”

A look beyond perfection

The research began three years ago when Ma became interested in the different strategies an animal could potentially use to accomplish a simple but common task: choosing between two options where the chance of reward changes over time.

The researchers were interested in exploring a group of strategies that fall between optimal and completely random solutions: “small programs” that are limited in resources but still get the job done. Each program specifies a different algorithm for guiding the animal’s actions based on past observations, allowing it to serve as a model of animal behavior.

As it turned out, there are many such programs — about a quarter of a million. To make sense of these strategies, the researchers first looked at a handful of the top performers. Surprisingly, they found that they do essentially the same thing as the optimal strategy, albeit using fewer resources.

“We were a little disappointed,” says Ma. “We spent all this time looking for these little programs, and they all follow the same calculations that scientists already knew how to derive mathematically without all this effort.”

But the researchers were motivated to keep looking—they had a strong intuition that there must be programs that are good but diverge from the optimal strategy. Once they looked beyond the top programs, they found what they were looking for: about 4,000 programs that fall into that “good enough” category. More importantly, more than 90% of them did something new.

They could have stopped there, but a question from fellow Janelian spurred them on: How could they find out what strategy the animal was using?

This question prompted the team to dig deep into the behavior of individual programs and develop a systematic approach to thinking about the entire collection of strategies. First, they developed a mathematical way to describe the interrelationships of programs through a network that connected different programs. Next, they looked at the behavior described by the strategies and designed an algorithm to reveal how one of these “good enough” programs might evolve from another.

They found that small changes to the optimal program can lead to large changes in behavior while maintaining performance. If some of these new behaviors are also useful in other tasks, it suggests that the same program might be good enough to solve a number of different problems.

“If you consider that an animal is not a specialist that is optimized to solve only one problem, but rather a generalist that solves many problems, this is a really new way to study it,” says Ma.

The new work provides a framework for researchers to begin thinking beyond individual optimal programs for animal behavior. Now the team is focused on investigating how generalizable the tiny programs are to other tasks and designing new experiments to determine which program an animal could use to perform a task in real time. They are also collaborating with other Janelio researchers to test their theoretical framework.

“Ultimately, a good understanding of animal behavior is a fundamental prerequisite for understanding how the brain solves different types of problems, including those that our best artificial systems solve only inefficiently, if at all,” says Hermundstad. “The key challenge is that animals can use very different strategies than we might initially assume, and this work helps us uncover this space of possibilities.”