Psychology of a superorganism

We are adapting ideas from psychology to understand how an ant colony functions as a single cognitive entity. We are particularly interested in differences between the performance of the whole colony and the individuals composing it. Much of our work centers on nest-site selection by Temnothorax ants. We have discovered that individual ants are prone to irrational changes in site preference when presented with certain challenging decision problems. Similar irrationality is well known in many animals, including humans. However, when colonies received the same challenge, they showed rationally consistent preferences. This suggests that groups can improve the quality of their decision-making by sharing their cognitive burden across many individuals.

Information sharing through social networks

The choice of a nest site depends heavily on information sharing among the scout ants who organize the colony's move. These ants use a behavior called tandem running (video here) to recruit fellow scouts to candidate sites. Scouts fully commit to a site only when its population reaches a minimum value, or quorum. By using this rule they combine their own direct evaluation of the site with an indirect cue about the judgements of other ants. This balancing of personal and social information is a general feature of collective decision-making, and quorums may offer a widespread mechanism for balancing the advantages of shared decision-making with the dangers of "groupthink". We have adopted a network approach to information sharing among scouts, giving us a synoptic view of how recruitment communication spreads information through the colony. The results suggest the importance of heterogeneity among scouts, with some consistently playing an outsize role. This contrasts with earlier views that all scouts are much the same.

Collective choice of the best food source

Ants also make group decisions about food, and we have found that colonies of Temnothorax rugatulus can choose the richer of two sucrose solutions. We are studing how this collective choice emerges from the ants' individual behavior, in particular their recruitment communication via tandem runs (video). The positive feedback of recruitment is fundamental to all forms of collective decision-making, but differences among recruitment behaviors have major impacts on group performance. Tandem runs are quite different from the highly nonlinear recruitment of trail-laying ants, instead showing essential similarities to the waggle dance of the honey bee, despite superficial differences. We are currently exploring this similarity to see if it gives T. rugatulus the same combination of selectivity and flexibility seen in the bee.

Teamwork by ants and robots

When scouts of the desert ant Aphaenogaster cockerelli find an item too large for a single ant, they organize a team that swiftly carries it to their nest (video here), thus evading more aggressive competitors. This impressive ability is not seen in most ant species, and we are working with Vijay Kumar's group at the University of Pennsylvania to understand the underlying behavioral strategies. In addition to its biological interest, this behavior offers a promising source of bioinspired algorithms for collective robotics. To measure the forces exerted by each ant, we have devised artificial loads ringed with sensors and induced teams of ants to retrieve them (videos here). We are using the resulting data to model how the ants achieve coordinated transport without central control, with minimum communication among team members, and in the face of obstacle-laden terrain and unpredictable load characteristics.

Using robots to probe colony function

Our collaborators at the University of Pennsylvania have also devised a simple robotic ant that successfully recruits real Temnothorax rugatulus via tandem runs (video here). The artifical ant is a small magnetic dummy dabbed with the ants' own recruitment pheromone. A larger robot is connected to it by a magnetic tether and can steer the dummy on predetermined paths. We are using this tool to manipulate information exchange among a colony's members as it carries out a collective choice of nest site. We are also using it to determine how tandem run followers learn visual landmarks that enable them later to retrace the followed path on their own.

 

Automated analysis of collective behavior

A major limitation on the study of collective behavior is the rate at which data can be gathered. We are collaborating with James Rehg and Tucker Balch at Georgia Tech to create an automated solution that will continuously track the positions of multiple interacting individuals and infer their behavior. These software tools will be widely applicable to studies of animal behavior, but development will focus on the challenging problems offered by ants, where multiple interacting animals must be simultaneously tracked. As a first step, we are using a prototype tracker to gather extensive data on the encounter rates of Temnothorax rugatulus scouts as they assess candidate nest sites (video here). We will use this data to test a simple model for how encounter rate data can be used to assess population density, a key source of information for the coordination of social behavior in ants.

---