Evolution Icon Evolution
Intelligent Design Icon Intelligent Design
Life Sciences Icon Life Sciences

To Regulate Foraging, Harvester Ants Use a (Designed) Feedback Control Algorithm

Eric Cassell
Photo: Harvester ants at the entrance to their nest, via Wikimedia Commons.

A recent study in the Journal of the Royal Society Interface reports on “A feedback control principle common to several biological and engineered systems.” The researchers, Jonathan Y. Suen and Saket Navlakha, show how harvester ants (Pogonomyrmex barbatus) use a feedback control algorithm to regulate foraging behavior. As Science Daily notes, the study determined that, “Ants and other natural systems use optimization algorithms similar to those used by engineered systems, including the Internet.”

The ants forage for seeds that are widely scattered and usually do not occur in concentrated patches. Foragers usually continue their search until they find a seed. The return rate of foragers corresponds to the availability of seeds: the more food is available, the less time foragers spend searching. When the ants successfully find food, they return to the nest in approximately one third of the search time compared to ants unable to find food. There are several aspects of this behavior that point to intelligent design.

Feedback Control

First, it is based on the general engineering concept of a feedback control system. Such systems use the output of a system to make adjustments to a control mechanism and maintain a desired setting. A common example is the temperature control of heating and air conditioning systems. An analogy in biology is homeostasis, which uses negative feedback, and is designed to maintain a constant body temperature.

Mathematical Algorithm

A second aspect of design is the algorithm used to implement the specific control mechanism. Suen and Navlaka describe the system as “multiplicative-increase multiplicative-decrease” (MIMD). The MIMD closed loop system is a hybrid combination of positive and negative feedback. Receiving positive feedback results in multiplying the response, while negative feedback results in reducing the response by a constant value. The purpose relates to the challenge of optimizing ant foraging. As the paper explains:

If foraging rates exceed the rate at which food becomes available, then many ants would return “empty-handed,” resulting in little or no net gain in colony resources. If foraging rates are lower than the food availability rate, then seeds would be left in the environment uncollected, meaning the seeds would either be lost to other colonies or be removed by wind and rain.

The authors found that positive feedback systems are “used to achieve multiple goals, including efficient allocation of available resources, the fair or competitive splitting of those resources, minimization of response latency, and the ability to detect feedback failures.” However, positive control feedback systems are susceptible to instability (think of the annoying screech when there is feedback into microphones in a sound system). Therefore, a challenge for MIMD systems is to minimize instability. 

In this application, when foraging times are short, the feedback is positive, resulting in a faster increase in the number of foragers. When foraging times are longer, the feedback is negative, resulting in a reduction in the number of foragers. A mathematical model of the behavior has confirmed that the control algorithm is largely optimized. (See Prabhakar et al., “The Regulation of Ant Colony Foraging Activity without Spatial Information,” PLOS Computational Biology, 2012.) As I describe in my recent book, Animal Algorithms, the harvester ant algorithm is just one example of behavior algorithms that ants and other social insects employ.

Suen and Navlakha point out that the mechanism is similar to that employed to regulate traffic on the Internet. In the latter context, there are billions of “agents” continuously transmitting data. Algorithms are employed to control and optimize traffic flow. The challenge for Internet operations is to maximize capacity and allow for relatively equal access for users. Obviously, Internet network control is designed by intelligent engineers. In contrast, the harvester ant behavior is carried out by individuals without any central control mechanism. 

Physical Sensors

A third feature indicating design is the physical mechanism used by the ants to determine how long returning foragers have been out. When ants forage for food, molecules called cuticular hydrocarbons change based on the amount of time spent foraging. This is due to the difference in temperature and humidity outside of the nest. As the ants return to the entrance of the nest, there are interactions between the returning and the outgoing ants via their antennae. These interactions enable detection of the hydrocarbons, which provide a mechanism to enable outgoing ants to determine the amount of time that returning ants spent foraging.

These three elements of harvester ant behavior (feedback control, mathematical algorithm, and physical sensors) present a severe challenge for the evolutionary paradigm. From a Darwinian perspective, they must have arisen through a combination of random mutations and natural selection. A much more plausible explanation is that they are evidence of intelligent design.