New research from the University of Sheffield is providing insight into the complex decision making processes of honey bees.
When you see a honeybee in your garden, it may seem like it is randomly hopping from flower to flower in search for nectar. In reality, it is undertaking a complex decision-making process to determine which flowers to investigate, and which to pass. For the honeybee, the stakes are high and every time a mistake is made, it has wasted valuable energy and exposed itself to potential dangers. To refine the decision-making process, honeybees rely on a brain the size of a sesame seed. None the less, they have evolved exceptional abilities, which are inspiring the next generation of AI and robotics.
Scientists at the University of Sheffield led by Dr. HaDi MaBouDi, in collaboration with Professor Andrew Barron from Macquarie University aimed to explore the strategies employed by honey bees to forage so efficiently and effectively, and the neural systems that enable them.
Researchers trained 20 honeybees to associate five different coloured artificial flowers with a certain likely hood of receiving either a sugar syrup solution (a reward), or bitter quinine (a punishment), based on their visit history. They then tested the bees by introducing them to a specially designed environment where they could track each bee's flight path and observe the time it took to decide on which flowers to visit or reject.
The team then conducted further tests to see how bees would react to more ambiguous stimuli. In the 'reduced evidence' test, researchers introduced flowers with colours that were blends of the previously rewarded or punished colours (a mix of blue and green), making the cues less evident. They also introduced a 'reduced reward likelihood' test, where the sugar syrup was offered less frequently than in the earlier experiments. This test evaluated the bees' resilience and flexibility in decision-making when the expected reward was less reliable. The response times and accuracy revealed that both the quality of evidence and the likelihood of reward played a significant role in the decision-making process-akin to that seen in primates.
Based upon these findings, the researchers then developed a computational model capable of replicating the decision-making patterns observed in the honeybees. Importantly, the model was designed to be biologically plausible, aligning with the bees' natural brain structure and functioning. This approach provided valuable insights into how a small brain, like that of a bee, can execute intricate decisions swiftly and effectively, which is of relevance to emerging research in robotics.
This study not only provides insight into the extraordinary capabilities of honeybees, but also hopes to aid the development of advanced autonomous systems. By mimicking the strategies of honey bees, AI can benefit from faster, more accurate decision-making, risk-averse behaviour, resource optimization, and better adaptability with limited data.