Risk Reward Dilemma

This class models a two-state social dilemma where a single agent chooses between risky or cautious actions. The actions taken by the agent determine the probability of transitioning between degraded and prosporus states . In each state, the agent receives different rewards, reflecting the consequences of its chosen action!

Implementation

---

source

RiskReward

 RiskReward (pc:float, pr:float, rs:float, rr:float, rd:float)

An MDP model for decision-making under uncertainty with two states (prosperous and degraded) and two actions (cautious and risky).

---

source

RiskReward.TransitionTensor

 RiskReward.TransitionTensor ()

Define the Transition Tensor for the MDP.

---

source

RiskReward.RewardTensor

 RiskReward.RewardTensor ()

Define the Reward Tensor for the MDP.

---

source

RiskReward.actions

 RiskReward.actions ()

Define the actions available in the MDP.

---

source

RiskReward.states

 RiskReward.states ()

Define the states of the MDP.

---

source

RiskReward.id

 RiskReward.id ()

Provide an identifier for the environment.

Example

env = RiskReward(pc=0.3,pr=.1,rs=0.6,rr=0.8,rd=0.001)
# pc, pr, rs, rr, rd

env.id()

'RiskReward_pc0.3_pr0.1_rs0.6_rr0.8_rd0.001'

env.TransitionTensor()

array([[[1. , 0. ],
        [0.7, 0.3]],

       [[0.1, 0.9],
        [0. , 1. ]]])

env.RewardTensor()[0]

array([[[0.6  , 0.   ],
        [0.8  , 0.001]],

       [[0.001, 0.001],
        [0.001, 0.001]]])

env.actions()

[['cautious', 'risky']]

env.states()

['prosperous', 'degraded']