​How Robot 'Sex' Can Help Us Understand Gender in Mating Strategies

Alright, I’m going to be honest—the robot sex we’re about to talk about isn’t the sexy Jude Law from AI Artificial Intelligence type. Nope, it’s the type of robot sex that helps us learn about certain reproductive behaviors in animals.

One of the many hot topics explored in evolutionary biology is that of polymorphism, or the occurrence of different physical traits within one population of species. The idea that a population of one species can exhibit a variety of strategies and traits that can each lead to mating seems to oppose the idea behind natural selection: that the strongest trait will become dominant and cause the less successful traits to phase out and become extinct. In other words, if one trait is a clearly successful means of reproducing, why haven’t other traits simply phased out?

Dr. Stephen Elfwing of the Neural Computation Unit of Japan’s Okinawa Institute of Science and Technology (OIST) took it upon himself to explore this idea, creating an environment that incorporated robots and computer simulation to track behavior over the course of hundreds of generations.

Elfwing built robots lovingly referred to as cyber-rodents, programmed to mimic animal behavior based on food and mating. These hermaphroditic robots were tasked with the goal to survive by either foraging for a battery charger to increase their energy (food) and tracking partners to exchange genotypes (sex). One of the stipulations for proper “mating” was that both cyber-rodents had to be positioned so that their infrared ports faced each other (probably a realistic assumption for sex). Another stipulation was that the cyber-rodents had to have a certain amount of energy in order to reproduce.

The Economist describes the study and how over the course of hundreds of generations (in a computer program), two distinct mating behaviors emerged:

The first, dubbed “foragers”, were those who would never wait for the other robot to turn around so that mating could take place; they always headed straight for the charger. Foragers would mate only if they saw the face of a potential partner right away. “Trackers”, by contrast, assuming they had enough energy, would ignore the charger and wait for the other robot to turn around to mate.

After close to 1,000 generations, the population in those experiments where this polymorphism arose stabilised at about 25% foragers and 75% trackers. Moreover, populations with this mix were some of the “fittest” (ie, fastest-breeding) of the 70 experiments on cyber-rodents.

So the robots “figured out” the most efficient way to reproduce. BIG WHOOP CYBER-RODENTS. But the results don’t just end there. The cyber-rodents were not assigned sexes—any rodent could copulate with any other rodent, and yet the two distinct behaviors emerged. Elfwing points out parallels to sex differences in animals:

The most striking example of polymorphism when it comes to mating is the existence of males and females. Dr Elfwing believes that foragers and trackers may be the first step down that evolutionary path. For distinct male and female roles to evolve from a population that has no predefined mating preferences, the robots would have to consider the risks and costs of each role: females, for instance, have to bear the cost of giving birth to offspring.

It’s hard to imagine (and just a bit ridiculous) that as hermaphroditic robots, foraging and tracking could be assigned gender roles. But it’s also strange that after hundreds of generations, these behaviors could emerge without other mutations. So who knows? Maybe in a couple thousand more generations, some cyber-rodents will form foraging groups around the battery chargers. Or have surrogate foragers. Or perhaps even move onto deadlier, more competitive mating behaviors. Or maybe they’ll just nihilistically realize that they’re useless components in a simple reproductive simulation and stop reproducing and blog about it.