快猫短视频

Honeybees can聽calculate probability, but it seems they don鈥檛 use it the same way we tend to.

Andrew Barron at Macquarie University in Sydney, Australia, and his colleagues trained 20 honeybees to associate the colours of artificial flowers with the likelihood of obtaining sweet water. Over multiple sessions, they presented the bees with five colours in various聽combinations of two colours at a time. Each colour was ranked one to five. For each pair, only the higher-ranking colour dispensed sweetened water.

The researchers then tested the bees on a combination they hadn鈥檛 seen yet: the second and fourth-ranking colours, representing odds of getting sweetened water of 66 and 33 per cent, respectively.

You might think the best strategy for obtaining the treat would be to only visit flowers with the highest odds of delivering, but the bees did something different, says Barron. They matched the proportion of visits with the probability of getting sweet water, so for flowers with 66 per cent odds of sugar, they visited them roughly two-thirds of the time. This is known as probability matching.

While that yields less sugar overall in the experiment than visiting the colour with the higher odds every time, it is a strategy that works better in a bee鈥檚 natural environment, says Zack Ellerby at the University of Nottingham, UK. With many bees competing for the same flowers with limited nectar, and the possibility of odds changing over time, this can be the optimal strategy, he says.

鈥淓cologically, when information is uncertain and when gathering information has a cost, then probability matching is the best thing to do,鈥� says Barron. 鈥淭hat鈥檚 what the bees did.鈥�

Humans tend to go for the highest odds in situations where probabilities seem to be fixed, such as a casino. The same might be true for bees; if they thought the odds for each colour were permanent, they might adjust their strategy, says Richard Mann at the University of Leeds, UK.

鈥淭he bees didn鈥檛 know their 鈥榯raining鈥� session was finished and the test had started,鈥� he says. 鈥淔or all they knew, they were still in a learning phase and were still assessing probability, while using the knowledge they already had to be efficient.鈥�