Fig. 1
From: When optimization for governing human-environment tipping elements is neither sustainable nor safe
Conceptual model of a human-environment tipping element. a Agent-environment interface: based on the state information and received reward, the agent chooses an action a from its actions set to gain rewards. b The transition graph gives state transition probabilities and corresponding rewards for all triples of state s, action a, next state s′, i.e., in state s the agent takes action a and moves to state s′. c Risky and cautious policies including the resulting Markov chains as a transition graph
