Social Dilemma

Class for a symmetric two-agent stateless social dilemma environment

Typical examples are the Prisoner’s Dilemma, Stag Hunt game, and the game of chicken/snowdrift/hawk-dove.

 SocialDilemma (R:float, T:float, S:float, P:float)

Symmetric 2-agent 2-action Social Dilemma Matrix Game.

	Type	Details
R	float	reward of mutual cooperation
T	float	temptation of unilateral defection
S	float	sucker’s payoff of unilateral cooperation
P	float	punsihment of mutual defection

 SocialDilemma.TransitionTensor ()

Get the Transition Tensor.

 SocialDilemma.RewardTensor ()

Get the Reward Tensor R[i,s,a1,…,aN,s’].

 SocialDilemma.actions ()

The action sets

 SocialDilemma.states ()

The states set

 SocialDilemma.id ()

Returns id string of environment

env = SocialDilemma(R=1, T=2, S=-1, P=0)

env.id()

'SocialDilemma_2_1_0_-1'

env

SocialDilemma_2_1_0_-1

Reward matrix of agent 0:

env.RewardTensor()[0,0,:,:,0]

array([[ 1., -1.],
       [ 2.,  0.]])

Reward matrix of agent 1:

env.RewardTensor()[1,0,:,:,0]

array([[ 1.,  2.],
       [-1.,  0.]])

env.TransitionTensor()

array([[[[1.],
         [1.]],

        [[1.],
         [1.]]]])

env.actions()

[['c', 'd'], ['c', 'd']]

env.states()

['.']