Ants may be tiny critters with tiny brains, but these social insects are capable of collectively organizing themselves into a highly efficient community to ensure the colony survives. And it seems that the social dynamics of how division of labor emerges in an ant colony is similar to how political polarization develops in human social networks, according to a recent paper in the Journal of the Royal Society Interface.
“Our findings suggest that division of labor and political polarization—two social phenomena not typically considered together—may actually be driven by the same process,” said co-author Chris Tokita, a graduate student in ecology and evolutionary biology at Princeton University. “Division of labor is seen as a benefit to societies, while political polarization usually isn’t, but we found that the same dynamics could theoretically give rise to them both.”
Tokita and his adviser/co-author, Corina Tarnita, were collaborating with a group at Rockefeller University that was using camera tracking to study ants—specifically, how division of labor emerges in very small groups (between 12-16 ants). Their job was to devise a model for a behavioral mechanism that would explain the patterns that the Rockefeller people had observed in their experiments. “Originally, we thought social interactions might play a part,” Tokita told Ars. “But it turns out we didn’t need to think about social interactions to capture their results.”
Tokita was familiar with the growing body of research in the social sciences involving opinion dynamics models—that is, how people’s opinions can change over time as they interact with and influence each other. And he noticed that the emergence of political polarization within such models was similar to how division of labor emerges among ant colonies.
He thought it should be possible to combine the response threshold model he’d developed for the ants’ social dynamics with the basic mechanism behind political polarization: a feedback loop between social influence and interaction bias. Social influence is the tendency of individuals to become similar to those they interact with, while interaction bias describes our tendency to interact with others who are already like us.
In Tokita’s original ant model, the ants choose their jobs within the colony based on which need meets a critical internal threshold. For example, if one ant has a lower threshold for hunger, it will be more likely to go forage for food, while another ant with a low threshold for concern about the colony’s larvae will devote more time to the nursery. Over time, each ant will have more interactions with other ants with thresholds similar to its, leading to the natural emergence of two groups: foragers and care providers.
This is usually a positive development, since it allows for the efficient functioning of the colony. However, Tokita and Tarnita found that if you add a strong feedback loop between social influence and interaction bias into the model, the two groups soon become so divided that they rarely interact at all, to the detriment of the colony as a whole.
According to Tokita, when only social influence is present, individuals interact randomly and become similar, so no division of labor naturally develops. When only interaction bias is present, individuals don’t differentiate, so you don’t get social factions. When both social forces are present, a strong feedback loop develops between them, resulting in both division of labor and polarized social networks. As both social influence and interaction bias increase, individual behavior becomes more specialized (biased) and individuals increasingly interact with those who are similar.
“We basically showed that there are critical tipping points where you expect individuals to diverge in their behavior, and that’s when there is a strong enough bias towards those [most similar] to you,” said Tokita. Interaction bias might still exist below that threshold, but it probably won’t be strong enough to produce the strong feedback loop that further reinforces the polarization.
According to Tokita, it is possible to reduce that strong divide simply by interacting a little bit more with those who are less like us, and/or letting our internal thresholds shift a little so we are a little less like our current “in” group. This essentially erases the differences. When that happens, “You don’t get division of labor, you don’t really get the polarized social network structure,” he said.
This is a phenomenon that has also been observed in colonies of honeybees in search of a new location for their hive. The bees send out scouts, who come back and report on prospective sites. Other bees then go out to check out the sites, come back to report, and so on, until the hive eventually reaches a consensus and all the bees relocate to the same location.
“If you had the kind of social interactions that causes the group to become very divided or polarized, you would never reach consensus,” said Tokita. A honeybee colony where two strong factions of bees are vying for different locations will end up splitting in two and may not survive.
Tokita et al.’s findings are consistent with those of a study last year by researchers at the Santa Fe Institute. That study concluded that social perception bias might best be viewed as an emergent property of our social networks, dependent solely on the relative sizes of the majority and minority groups as well as the extent to which like nodes connected to other like nodes. It also suggested that one potentially effective strategy to counter bias would be to diversify social networks. However, people often strenuously resist such diversification efforts, in part because the associated cognitive dissonance can be so extreme and uncomfortable.
The high degree of specialization within the sciences is another example of a naturally emerging division of labor that can be beneficial but turn detrimental if scientists in specific disciplines primarily interact only with those within their speciality, isolating themselves from new ideas in other fields. “I think there’s a lot of cases where there are really interesting ideas that are broadly applicable to other areas, but it’s hard to hear about them because we are sort of in silos,” Tokita said, adding that he appreciated the chance to bring ideas from sociology and political science into the realm of animal collective behavior.
Tokita cautions against reading too much into these findings. “It’s pointing to new research directions, but it can’t directly say anything about politics necessarily,” he said. “People aren’t ants, and ants aren’t people. We just wanted to bring the social dynamics, and the connections between these different collective behaviors, into a broader context.”