You could try using the quadtree culling algorithms (seen here: http://lab.polygonal.de/2007/09/09/quadtree-demonstration/) used in collision detection. Just treat each chicken as a circle that's as big as the distance at which it stops considering other chickens its neighbor.
I am admittedly bad with this sort of thing, so it might really be the R-tree in disguise.
On another note, it's good to see someone on HN get interested in game development. The road is long, hard, and littered with half-finished projects, but also the occasional finished one. The sense of accomplishment and fun factor is several magnitudes higher than coding 'normal' stuff.
I'll look into it briefly, and though I haven't finished evaluating all my options, I get the notion that quadtrees were used for static collision detection culling, or it was precomputed for a model.
In addition, I think different trees have different insertion and deletion complexities (don't know which is which yet). I need something, like the other commenter says, that doesn't penalize for adjusting the tree.
I am admittedly bad with this sort of thing, so it might really be the R-tree in disguise.
On another note, it's good to see someone on HN get interested in game development. The road is long, hard, and littered with half-finished projects, but also the occasional finished one. The sense of accomplishment and fun factor is several magnitudes higher than coding 'normal' stuff.