se-risk/players.py at main · code-lab-org/se-risk · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
import numpy as np
import scipy.stats as stats

from hunt.game import Player, Decision


class RiskAwareRDplayer(Player):
    def __init__(self, game, name, risk_aversion):
        super().__init__(game, name)
        self.risk_aversion = risk_aversion

    def get_utility(self, value):
        if self.risk_aversion == 0:
            return value
        else:
            return (1 - np.exp(-self.risk_aversion * value)) / self.risk_aversion

    def get_risk_dominance(self, design, independent_design):
        return np.log(
            (
                self.get_utility(
                    self.game.get_payoff(Decision(0, independent_design), Decision(0))
                )
                - self.get_utility(
                    self.game.get_payoff(Decision(1, design), Decision(0))
                )
            )
            / (
                self.get_utility(self.game.get_payoff(Decision(1, design), Decision(1)))
                - self.get_utility(
                    self.game.get_payoff(Decision(0, independent_design), Decision(1))
                )
            )
        )

    def get_decision(self):
        with np.errstate(invalid="ignore"):
            independent_value = [
                self.get_utility(self.game.get_payoff(Decision(0, design), Decision(0)))
                for design in range(len(self.game.designs))
            ]
            independent_design = np.nanargmax(independent_value)

            risk_dominance = np.array(
                [
                    self.get_risk_dominance(design, independent_design)
                    for design in range(len(self.game.designs))
                ]
            )

            collaborative_value = [
                self.get_utility(self.game.get_payoff(Decision(1, design), Decision(1)))
                for design in range(len(self.game.designs))
            ]
            collaborative_design = np.argmax(collaborative_value * (risk_dominance < 0))
            if np.nanmin(risk_dominance) < 0:
                return Decision(1, collaborative_design)
            else:
                return Decision(0, independent_design)


class RiskAwareRDplayer_averse1(RiskAwareRDplayer):
    def __init__(self, game):
        super().__init__(game, "RD_averse1", 0.1)


class RiskAwareRDplayer_averse2(RiskAwareRDplayer):
    def __init__(self, game):
        super().__init__(game, "RD_averse2", 0.2)


class RiskAwareRDplayer_seeker1(RiskAwareRDplayer):
    def __init__(self, game):
        super().__init__(game, "RD_seeker1", -0.03)


class RiskAwareRDplayer_seeker2(RiskAwareRDplayer):
    def __init__(self, game):
        super().__init__(game, "RD_seeker2", -0.065)


class RiskAwareRDplayer_neutral(RiskAwareRDplayer):
    def __init__(self, game):
        super().__init__(game, "RD_neutral", 0)


class RiskAwareEUplayer(Player):
    def __init__(self, game, name, risk_aversion, prior_collab=1, prior_no_collab=1):
        super().__init__(game, name)
        self.strategy_prior = np.array([prior_no_collab, prior_collab])
        self.risk_aversion = risk_aversion

    def get_utility(self, value):
        if self.risk_aversion == 0:
            return value
        else:
            return (1 - np.exp(-self.risk_aversion * value)) / self.risk_aversion

    def report_result(self, result):
        self.strategy_prior[result.their_decision.strategy] += 1

    def get_decision(self):
        p_collab = stats.beta(self.strategy_prior[1], self.strategy_prior[0]).mean()

        independent_expected_value = np.array(
            [
                self.get_utility(self.game.get_payoff(Decision(0, design), Decision(1)))
                * p_collab
                + self.get_utility(
                    self.game.get_payoff(Decision(0, design), Decision(0))
                )
                * (1 - p_collab)
                for design in range(len(self.game.designs))
            ]
        )
        independent_design = np.nanargmax(independent_expected_value)

        collab_expected_value = np.array(
            [
                self.get_utility(self.game.get_payoff(Decision(1, design), Decision(1)))
                * p_collab
                + self.get_utility(
                    self.game.get_payoff(Decision(1, design), Decision(0))
                )
                * (1 - p_collab)
                for design in range(len(self.game.designs))
            ]
        )
        collaborative_design = np.argmax(collab_expected_value)

        if np.all(
            collab_expected_value < independent_expected_value[independent_design]
        ):
            return Decision(0, independent_design)
        else:
            return Decision(1, collaborative_design)


class RiskAwareEUplayer_neutral(RiskAwareEUplayer):
    def __init__(self, game):
        super().__init__(game, "EU_neutral", 0)


class RiskAwareEUplayer_averse1(RiskAwareEUplayer):
    def __init__(self, game):
        super().__init__(game, "EU_averse1", 0.2)


class RiskAwareEUplayer_averse2(RiskAwareEUplayer):
    def __init__(self, game):
        super().__init__(game, "EU_averse2", 0.1)


class RiskAwareEUplayer_seeker1(RiskAwareEUplayer):
    def __init__(self, game):
        super().__init__(game, "EU_seeker1", -0.001)


class RiskAwareEUplayer_seeker2(RiskAwareEUplayer):
    def __init__(self, game):
        super().__init__(game, "EU_seeker2", -0.005)