cryptogon.com

“Against weak competitors, a strategic approach is probably useful, because if you figure out the approach of the opponent, you can actually do better than when you’re just random. You can even out-compete them,” Karpova said.

But random exploration may be more appropriate to a new situation full of mixed signals, where experience offers no guide. Some part of the brain has to decide how to decide.

Researchers set out to find how and where such choices are made. They created three computer opponents with increasingly tough strategies to punish patterns in how rats sought selected holes to poke their noses in, looking for food. One was a bit lenient; the second, a bit less forgiving; and the third, strict — it was programmed to learn from the rat and thwart its strategy.

“When animals initially played against the weak competitors, they appeared to maintain a strategic approach, just like people have observed before with primates,” Karpova said. “But when we put them up against a very sophisticated competitor … the animals switched to the random mode.”

And they got stuck there, for hundreds of trials.

Researchers weren’t done. They created a new brain-buster. It gave higher rewards for a certain pattern of pokes (such as right-right-left) but also mild reward rates (16%) for non-patterned responses.

The “strategic” rodents that had played against competitors one and two beat the odds at the hidden pattern task. Their reward rates, about 20%, indicated that sticking to strategic thinking helped them uncover the hidden pattern. Those that had first played the sternest computer taskmaster earned just a 14.5% reward rate, below even the random rate.