What decision-making strategies should be considered ideal?

New Study Published in Nature Communications on When Continuous Learning Leads to an Optimal Decision-Making Strategy Reveals Unexpected Results

Photos: depositphoto

Unlike machines, animal and human behavior almost always includes an element of unpredictability. Numerous experiences show that our responses to the same challenge are sometimes faster, sometimes slower, sometimes correct and sometimes wrong.

In the field of neuroscience, this variability is often attributed to so-called “noise” – a “neural babble”, which is always present, and which influences the way brains process and respond to the information they receive.

A new study, the result of a collaboration between scientists at the Champalimaud Center in Portugal, from Harvard Medical School in the USA and the University of Geneva in Switzerland, it shows that this variability is often incorrectly interpreted as noise.

This idea may be a reflection of a behavioral strategy that has hitherto been neglected on the basis of erroneous assumptions about how the individual should behave. These results, published on June 2, in the scientific journal Nature Communications., question what “optimal behavior” actually means in the performance of certain tasks.


an unexpected strategy

“It all started with a simple experiment”, recalls Maria Inês Vicente, co-author of the study, who collected experimental data during her doctorate at the Champalimaud Centre, currently working at the University of Leiden.

“We started with two different odors and created several mixtures of the two. During the experiment, the different mixtures were presented to the rats, one at a time. In each trial, the rats had to answer which of the two odors was dominant. If he thought the answer was scent A, the rat had to approach the water cooler on the right side, and if he chose scent B, he would have to go to the left side.

Some mixtures had much more odor of one type than the other, making it easier to identify which one was the most salient odor. But in other blends, the difference was more subtle. If the rat answered correctly, it received a reward (water)”.

The researchers recorded how quickly the mice responded and whether the response was right or wrong. To their surprise, when they analyzed the data, they found that the rats' behavior did not follow a common decision-making rule.

“In this type of task, we tend to see a clear dependence between the difficulty and time of decision making: in the most difficult and subtle cases, animals (and humans) take longer to decide than in the easier cases”, says Mendonça . "But instead, what we saw was that our rats took, on average, the same amount of time to make both difficult and easy decisions."

"It was not trivial to find an explanation for this unexpected observation," adds Jan Drugowitsch, another co-author of this study.

“Finally, we were able to understand what was going on – and for that we built a mathematical model that unites different fields of research in decision-making. In a way, what we did was to replicate the behavior of rats in the 'brain of a machine' in order to discover which underlying variables are responsible for this surprising result”.

The model revealed an unexpected strategy. In each trial, what the rat was doing was readjusting its behavior according to the results of the previous trial.

If the mouse was correct on the previous attempt when the “question” was difficult, its choice would be skewed to the same side on the next attempt. And vice versa, an incorrect answer would lead to a change of judgment on the next attempt.

Why did animals adopt this specific strategy? “This strategy is consistent with a worldview in which the environment is continually changing, which prompts animals to update their decision-making approach, trial after trial. From the outside, its behavior appears to be extremely variable and therefore it would have been easy to misinterpret it as 'just' noise”, points out Drugowitsch.


The ideal behavior is in the eyes of the beholder

Why did the rats choose a different strategy than expected? The authors explain that there are several reasons, the first being the nature of the task. “There is not just one type of sensory discrimination task”, says Mendonça.

“In reality, there are several elements of the task that can trigger different decision-making strategies. If, for example, we had asked rats to locate a sound rather than discriminate odors, their strategy would be in line with our initial expectation. This is because for certain sensory modalities there is a clear “natural” separation between left and right in the brain, but this is not the case when classifying odors.”

Another reason the researchers observed is trust. “Like humans, rats seem to evaluate their own decisions and change their behavior accordingly. When we feel very confident, we end up making the right decision and there isn't much to learn.

But what happens when we are confident and end up discovering that we were actually wrong? In this case, we will end up changing our behavior drastically. This is exactly what we saw in our rats”, explains Zachary Mainen, one of the leaders of the group that developed this study, principal investigator at the Champalimaud Center.

According to the authors, another explanation for the choice of strategy by the rats is the “hard neuronal circuit” that involves the learning process.

“Ironically, if the rats weren't constantly re-adjusting their responses, based on the outcome of the last trial, they would do much better in their choices. What we were initially waiting for was that they would build an 'odor A – odor B' category and implement it,” says Alex Pouget, another of the group's leaders and co-author of the study, principal investigator at the University of Geneva. “Still, the rats' strategy makes perfect sense.”

As the authors explain, this observation does not mean that the rat is an ill-adapted animal, quite the contrary, they claim that the scientific community should reconsider what it defines as “ideal behavior”.

“Rats have evolved over millions of years to explore an ever-changing environment. Therefore, when evaluating the behavior of these animals, we must remember that it is not necessarily just performance itself. The ideal behavior will depend as much on the problem itself, as on the nature of who will solve it”, argues Pouget.

“We believe that our work is a good starting point for us to continue to explore how different branches of the study of decision making can interact. We also hope that other scientists will use and refine our models using them in new experiments. It would be fascinating and informative to see when, how and why this model of ours starts to fail. Making a mistake is an opportunity to learn something new, which is the result and message of our study”, concludes Mendonça.


Author Champalimaud Foundation
Science in the Regional Press – Ciência Viva