We have been tossing this idea around for quite a while now, but only have some bits of pilot data to show for. Virtual ecology implies that part of the process is computer generated. In our case, the bees were real, but the flowers were computer generated. For lack of any realistic programs for generating flower shapes, Chris Evans and I used Richard Dawkins' Blind Watchmaker biomorph software for generating flower like shapes. We selected a flower-like biomorph and systematically degraded it to create a starting 'flower'. The starting 'parent' bred two offspring and the three were presented to bees, each offering the same reward of sugar water. The winning flower became the parent for the next generation, and generated two more offspring. An example of an 'evolutionary' series of flowers is this:
In this figure, the dark flower was the winner in each generation (row). In this series, individual bees were lured into the lab and made repeated choices among the three flowers of a generation. Another method of testing bees is to offer the flowers outdoors. Bees coming to a feeding station get to choose among the three flowers, and recruit hive mates, much like their natural foraging ecology. We count up the amount of sugar water drunk at each flower.
The testing methods with bees work well. We have since worked out methods for testing bees outdoors under more ecologically natural conditions. We need a more realistic 'genome' for generating flower shapes. We would also like to incorporate costs in producing flowers (e.g., cost per unit size).
This is a great project for a graduate student with the right skills.