A fully generic Markov chain can have a state space with arbitrary variables, so can incorporate some memory that way. But that ends up with a continuous state space which complicated things in ways I’m not certain of without going back to the literature.
An easier solution is if you can consider discrete states. For example, ongoing great power war (likely increases many risks) or pandemic (perhaps increases bio risk but decreases risk of some conflicts) might be states.
A fully generic Markov chain can have a state space with arbitrary variables, so can incorporate some memory that way. But that ends up with a continuous state space which complicated things in ways I’m not certain of without going back to the literature.
An easier solution is if you can consider discrete states. For example, ongoing great power war (likely increases many risks) or pandemic (perhaps increases bio risk but decreases risk of some conflicts) might be states.