Assume there exists a behavior of an organism one wants to model in a robot.
Assume further, there exist properties of the underlying structure, usually neurons, you hypothesize the behavior to be based on.

Know this: A solution can not be more interesting that the problem posed.
Again, differently: How interesting a solution is, is bounded by the problem.

To make a little robot avoid walls and to follow lights, mostly any property of a neuron will do. Those include delays, multiplicative synapses, spiking neurons, channel plasticity, moving dendrites, large networks, neuromodulators, different time scales, field potentials, and the list goes and goes. All afford equally interesting solutions.

Each not really very much, though.

Einstein said, and everyone knows: " make your solution as simple as possible but not simpler".

He could be talking about modeling of neural systems.