WEEK 7 - ECN306 Lecture 7 Deviations from Nash equilibrium(2)

Lecture 7: Deviations from Nash Equilibrium

Page 1: Introduction

  • Lecture Title: Deviations from Nash Equilibrium

  • Course: ECN306 Game Theory for Economists

Page 2: Today's iSheffield code

  • [Specific iSheffield code provided in class]

Page 3: Plan for Today

  • Focus Areas:

    • Discussion of the Nash equilibrium concept.

    • When does play converge to Nash equilibrium?

    • Reasons for behavioral differences from predictions.

    • Methods to improve accuracy of NE predictions.

    • Exploration of games of conflict.

    • Discussion of beauty contest games.

    • Examination of games lacking pure-strategy equilibria.

Page 4: Critical Discussion of the Nash Equilibrium Concept

  • Examination of Nash equilibrium in various contexts and implications.

Page 5: Normative and Descriptive Interpretation of NE

  • Normative Interpretation:

    • Represents how rational players should behave in a game.

    • If payoffs denote utility and all players are rational, behavior aligns with NE predictions.

    • Applicable in non-cooperative game theory; cooperative game theory involves different solution concepts.

  • Descriptive Interpretation:

    • Reflects actual player behavior in a game, which often diverges from NE predictions.

Page 6: Why Does Behaviour Differ from NE?

  1. Assumption of Selfishness:

    • NE assumes players are selfish expected utility maximizers.

    • This can be relaxed by modeling preferences considering altruism or spite.

  2. Cognitive Limitations:

    • NE assumes unlimited cognitive abilities; players may make mistakes or learn over time.

  3. Belief Accuracy:

    • NE assumes players have accurate beliefs; can be modeled to account for errors in belief.

Page 7: Reiteration of Behavioral Differences from NE

  • Re-emphasizes points from page 6 about selfishness, cognitive abilities, and belief accuracy.

Page 8: Preferences in Game Theory

  • Nash Equilibrium calculations usually assume players maximize monetary payoffs.

  • Real-world factors influencing utility include:

    • Inequality or concerns regarding others' earnings (altruism or spite).

    • Emotional responses (e.g., anger or guilt).

    • Social norms can also impact behaviors.

  • Recommendation: incorporate these factors into utility functions for realistic NE calculations.

Page 9: Illustration: Push-Pull Game

  • Structure of the Game:

    • Payoff matrix considering candy as the utility.

  • Equilibria Explanation:

    • Nash equilibrium under the candy assumption vs. alternative realistic utility models.

Page 10: Behavioral Differences from NE (Continued)

  • Repeats the reasons behavioral differences arise from NE from previous pages.

Page 11: Example: Traveller’s Dilemma

  • Scenario Description:

    • Lost bag reimbursement claim scenario outlining incentives to inflate claims.

  • Claims are independent and lead to penalties for the inflated higher claim.

Page 12: Best Response and NE in Traveler's Dilemma

  • Mathematical representation of claims and penalties leading to NE of (80, 80).

Page 13: Data from Traveller’s Dilemma Experiment

  • Claims were higher with a penalty of 5 vs. 80, illustrating NE predictions versus actual behavior.

Page 14: Laboratory Experiment Results

  • NE prediction insensitivity to penalty size demonstrated through experimental data.

Page 15: NE Sensitivity to Incentives

  • Discussing how NE predictions remain unchanged despite variable incentives.

Page 16: Quantal Response Equilibrium (QRE)

  • QRE incorporates player errors in decision-making by predicting best responses to beliefs about opponents' moves.

Page 17: Finding QRE in 2x2 Game

  • Explanation of sensitivity parameters and how to derive probabilities for players' strategies in equilibrium.

Page 18: QRE's Superior Data Explanation

  • QRE better explains observations in scenarios like the volunteer's dilemma as compared to NE.

Page 19: Case Study: Behavior in Contests

  • Experiment results show average effort exceeded NE predictions, reflected in investment distribution data.

Page 20: Reasons for Over-Expenditure in Contests

  • Factors influencing behavior include risk-seeking preferences, spiteful social preferences, and joy from winning—not purely monetary incentives.

Page 21: Cognitive Abilities and Convergence to NE

  • Discussion on players' abilities to converge to the NE through deliberation and learning from past actions.

Page 22: Learning and Feedback Mechanisms

  • Two main models:

    • Belief Learning: Based on observed past actions.

    • Reinforcement Learning: Players choose actions that historically yielded better payoffs.

Page 23: Non-equilibrium Models

  • Rational for using non-equilibrium models to simulate player behavior and learning dynamics rather than relying solely on NE predictions.

Page 24: Conditions for Convergence to NE

  • Identifying repeated interactions, effective feedback mechanisms, and simplified game structures as conducive to convergence.

Page 25: NE Convergence with Feedback

  • Emphasizes that players converge towards NE when given information about foregone payoffs.

Page 26: Reiteration of Reasons for Behavioral Differences from NE

  • Repeats previous points on behaviors diverging from NE regarding preferences, cognitive abilities, and beliefs.

Page 27: Another Game Example

  • Encouragement for students to log in for further resources related to the next game topic.

Page 28: Beauty Contest Game

  • Participants choose numbers between 0 and 100; the goal is to be closest to 2/3 of the average chosen number.

Page 29: Best Responses and NE of the Beauty Contest

  • Details that the best response is choosing 2/3 of the average, with potential NE outcomes discussed.

Page 30: Results in Beauty Contest Game

  • Emphasizes variable player experience influencing NE achievement, notably in trained versus untrained participants.

Page 31: Explanation of Beauty Contest Results

  • Level-k reasoning framework illustrating how players adjust their beliefs and responses based on perceived tendencies of other players.

Page 32: Belief Models for One-Shot Games

  • Discusses common belief modeling approaches in one-shot games and their implications for NE predictions.

Page 33: Games Without Pure Strategy Equilibria

  • Introduction to scenarios where traditional pure strategy equilibria do not exist.

Page 34: Penalty Kick Game Dynamics

  • Outlining strategic interactions between striker and goalkeeper in a penalty kick scenario, represented through probability-based payoffs.

Page 35: Lack of Pure Strategy Equilibrium in Penalty Kicks

  • Demonstrates mixed strategy equilibrium as the resolution in penalty kick scenarios.

Page 36: Idea Behind Mixed Strategies

  • Transition from discrete choices to developing continuous strategies using probabilities for player actions.

Page 37: Set Up of Expected Payoff Functions

  • Details calculating expected payoffs and determining conditions for mixed-strategy Nash equilibria.

Page 38: Best Responses for Striker

  • Mathematical conditions for the striker's best responses based on goalie probabilities.

Page 39: Best Responses for Goalkeeper

  • Determining optimal strategies for the goalkeeper based on striker's probabilities.

Page 40: Mixed-Strategy Nash Equilibrium Results

  • Summary of mixed-strategy equilibrium conclusions for the penalty kick scenario, highlighting critical probability thresholds.

Page 41: Next Time

  • Assignments for next class:

    • Worksheet 8

    • Engagement task 3

    • Reading: Chapter 5 focusing on NE discussions.