Varied Bot Behavior Without Randomness?

(I believe this applies to C++ as the underlying coding language for Unreal Engine 2.0, but I'm curious about this question in relation to any other programming language that could be applicable...)

In Unreal Tournament 2004, I have the ability to create a deathmatch among bots, and I can watch it as a spectator. Despite having an identical initial state to the system, and presumably identical bots, each such playthrough (same map, weapons, spawns, etc.) plays out very differently. This includes early movements before the first kill, respawn, or weapon pickup: Bots behave differently, right off the bat, each time.

1. Is this due to (pseudo) random seeds or ranges in the programming?

2. More importantly: Is there any other possible method or means for such variation? If not for this particular game, then in practice? Are there non-random behavioral programming methods that could account for variations in bot behavior despite an apparently-identical initial condition? (I'm being practical - not trying to start anything philosophical.)

(Note: Since this isn't about a specific programming issue or solution, if someone wants to punt this question into the correct stackexchange, I would be much obliged.)