Consider the possible attributes of an environment among several modalities (observability, determinism, discreteness, etc.). For each, consider the possible options (e. g.,
deterministic vs. stochastic, or continuous vs. discrete) and compare them. Would one
prove harder to design an intelligent agent for than the other? Why?
(For example, designing an agent for a partially observable environment is naturally more
difficult than for a fully observable environment because we cannot be certain of the state
and must use additional overhead to maintain some sort of belief state.)