# Inspired by Sondik's infinite-horizon paper
#

discount: 1
values: reward
states: interested bored
actions: tv radio nothing
observations: want-to-go dont-want-to-go

# transition probabilities

T: tv : interested      0.9 0.1
T: tv : bored           0.6 0.4

T: radio : interested   0.8 0.2
T: radio : bored        0.3 0.7

T: nothing : interested 0.5 0.5
T: nothing : bored      0.1 0.9

# observations

O: tv : interested      0.8 0.2
O: tv : bored           0.7 0.3

O: radio : interested   0.7 0.3
O: radio : bored        0.4 0.6

O: nothing : interested 0.9 0.1
O: nothing : bored      0.1 0.9

# cost structure

R: tv : * : *: * -10

R: radio : 1 : *: * -4

R: nothing : * : *: * 0