There is a deep connection between the way your brain and a swarm of bees arrives at a decision
Thursday, March 1, 2012 | 9
As goes a bee, so goes a neuron Image: Florin Tirlea
Our brains seem to work not by generating only “correct” actions and executing them in serial, but rather by representing many possibilities in parallel, and suppressing all but one. When this inhibitory action is lost, as happens in people with frontal lobe damage, these multiple possibilities become a burden, and can lead to so-called utilization behaviors. Such impaired individuals will indiscriminately reach for objects placed in front of them - a hairbrush or a hammer, for example - and use them even in inappropriate contexts.
In essence, despite our feeling that we are singular, unified agents, we are more like hive minds unto ourselves, our brains abuzz with multiple, often conflicting plans and interests that must be managed. To Dr. Thomas Seeley, a professor of neurobiology at Cornell University, the “hive mind” is more than just a metaphor. In a recent paper in Science, Seeley and his colleagues describe a potential deep parallel between how brains and bee swarms come to a decision. With no central planner or decider, both brains and bee hives can resolve their inner differences to commit to single courses of action.
To watch a group of bees is to see a frenzy of different interests coalesce into a single, clear thought. This is analogous to neurons in the brain, which must reach a consensus on how to achieve a behavioral goal by positioning the body in space. Bees in a hive must do something similar when deciding where to move the superorganism that is the swarm. Failing to move the swarm as a single, committed unit risks splitting up the hive and losing the queen. Similarly, making a poor move could expose the hive to predators or extreme temperatures.
Like many other decision-makers, the hive’s first order of business before making a springtime move is to consider the various possibilities. Toward this end, several groups of scouts are sent off to search for a suitable new hive. When the scouts return, they each advocate for preferred new sites - often different ones - by performing the famed “waggle dance,” a figure-eight series of movements that tells other bees the direction and distance to a potential new site. These dances recruit other uncommitted bees in the hive to also advocate for the advertised site.
For a while, many scientists thought that this strategy of steadily accumulating “votes” for a particular location was sufficient to explain the hive’s eventual decision. Others, including Seeley and his colleagues, were not satisfied. What happens in cases where similarly sized groups of bees are advocating for different locations? Wouldn’t this be a formula for deadlock?
Seeley suspected that the answer had to do with a head-butting move bees make. To explore this idea, he and his team first set up swarms on an island lacking natural nests, and gave scouts a choice between two identical artificial nesting boxes. Scouts that visited one site were marked with yellow paint, while scouts visiting the other site were marked with pink paint. By tagging these two different populations, Seeley and colleagues had in a sense labeled two competing ideas, which they could then watch unfold and interact back in the collective hive mind.
The researchers found that the yellow and pink-painted scouts displayed waggle dances advertising for their respective nests. In addition, however, the scouts were also seen to make brief buzzing head-butts to one another’s head and thorax. Dancing bees tended to receive head-butts toward the end of their dances, suggesting that the head butts were a signal to stop dancing. The most interesting finding came when looking at who was head-butting whom. Yellow-marked bees tended to receive these putative stop signals from pink-marked bees, and vice versa. In other words, the two different populations were mutually inhibiting one another - one proposal pitted against another.
The result of this arrangement is that it amplifies small differences between different populations of scouts, setting up a kind of winner-take-all scenario. Without inhibitory stop signals, the hive would be able to sustain multiple competing interests, as different groups of scouts accumulate more and more votes until the hive reaches some stable, but divided state. With stop signals, divided hive states are far less stable. A slight preponderance of one group of scouts will translate into greater inhibition of other groups of scouts, turning an initially small numerical advantage into a more sizable one. Over several iterations of this process, an initial slight majority is amplified into a consensus.
Ideally, a follow-up experiment would have eliminated the bees’ stop signals and studied the consequences on the hive’s decision process. Since this is nearly impossible to do, Seeley and his colleagues opted for a simulation based approach instead. In their models of collective bee activity, cross-inhibitory stop signals were essential for breaking decision deadlocks between two equally attractive nests. If the stop signals were indiscriminate, or absent altogether, the hive remained split, and never converged on a consensus.
Seeley and his team propose that cross-inhibition may be a general strategy for decision making, and indeed, their findings in bees recapitulate features of decision making and pattern formation in other systems. The remarkable unifying theme in all of these systems is how an aggregate swarm intelligence is built from just a few kinds of simple, local interactions between agents. Both neurons and bees are presumably unaware of how their impulses and signals transcend the individual, and lay the substrate for a grander, collective intelligence.