There is evidence that birds form flocks to guard against predators (
link,
link) but other studies show no correlation (
link,
link).
It would be interesting to devise an experiment where a standard mathematical flocking model is used, with parameters such as flock size, flying speed, noise, and where a human user tries to catch birds of the flock using an interface such as a mouse/pointer directed model predator. The predator motion could be restricted to mimic a raptor, i.e. turning circle, acceleration, maximum speed.
Statistics of success rate (from the predator's point of view) versus parameters could then be measured.
This could all be automatic of course and driven by computer code, but the human interface gives an interesting way of engaging the public.
(One by product of such an experiment would be consideration of how different human subjects perform, with regard to age, gender etc).