Why on earth did you do that?!

Size: px
Start display at page:

Download "Why on earth did you do that?!"

Transcription

1 Why on earth did you do that?! A simple model of obfuscation-based choice Caspar Chorus Delft University of Technology Challenge the future

2 Background: models of decision-making Based on the notion that choices are based on motivations (preferences, desires, decision-rules) - Theory of Reasoned Action / Planned Behavior (Psych.) - Optimal Decision-Making (OR, Dec. Sciences) - Rational Choice Theory (Economics, Discrete Choice) - Belief-Desire-Intention framework (Multi-Agent Systems, AI) - In other words, motivations echo through in choices e.g. Revealed preference axiom Cornerstone of micro-econom(etr)ics 2

3 Research aim Current models of decision-making: Based on the notion that choices are based on motivations (preferences, desires, decision-rules) In other words, motivations echo through in choices This talk: decision-maker wishes to suppress the echo Exhibits choices, but aims to hide motivations from an onlooker Derivation of formal model of obfuscation-based decision-making + Illustration of selected properties Warning: work in (early) progress 3

4 Why obfuscate?! Human examples Protect privacy Hiding your preferences, motivations from other humans or from AI-powered systems (e.g. hide WtP for upgrade from KLM) Create moral wiggle room Faced with moral dilemma, unsure which moral principle to apply; choose so that it becomes difficult for others to judge you Create strategic ambiguity Avoid (legal, political, military) punishment by hiding motivations underlying actions; e.g. dual use nuclear technology 4

5 Why obfuscate?! AI examples Protect privacy Artificial agent / AI-system operating on behalf of human, e.g. in auction; seeks to preserve human s privacy Create moral wiggle room Autonomous, renegade AI with meta-intelligence wishes to avoid punishment from supervisor for, e.g. algorithmic discrimination Create strategic ambiguity Systems of AI-agents trying to hide strategic objections from each other in, e.g., military situations (autonomous weapons) 5

6 Relevance: Transportation Onlooker Obfuscator Human Artificial Human Moral decisions (e.g. taboo trade-offs) Keeping your Automated Vehicle on the moral high ground ( moral machine ) Artificial Recommender system (e.g. personal travel assistant) AV-AV interaction on multi-lane highways, crossroads 6

7 Relevance: Transportation (I) Onlooker Obfuscator Human Human Moral decisions (e.g. taboo trade-offs) Artificial Travel recommender Vehicle ownership tax Transport Policy 300 euro less tax system (e.g. personal travel Travel time assistant) 20 mins. less time Keeping your Automated Number of injured AV-AV interaction on 100 injured more Artificial Vehicle on the moral high ground ( moral machine ) Number of fatalities multi-lane highways, YOUR CHOICE crossroads 5 fatalities more I support I oppose 7

8 Relevance: Transportation (II) Onlooker Obfuscator Human Artificial Human Moral decisions by travelers and citizens (e.g. taboo trade-offs) Keeping your Automated Vehicle on the moral high Artificial Recommender system (e.g. personal travel assistant) AV-AV interaction on multi-lane highways, ground ( moral machine ) crossroads 8

9 Relevance: Transportation (III) Onlooker Obfuscator Human Artificial Human Moral decisions (e.g. taboo trade-offs) Keeping your Automated Vehicle on the moral high ground ( moral machine ) Artificial Recommender system (e.g. personal travel assistant) AV-AV interaction on multi-lane highways, crossroads 9

10 Relevance: Transportation Onlooker Obfuscator Human Artificial Human Moral decisions (e.g. taboo trade-offs) Keeping your Automated Vehicle on the moral high ground ( moral machine ) Artificial Recommender system (e.g. personal travel assistant) AV-AV interaction on multi-lane highways, crossroads 10

11 Relevance: variety of research fields Onlooker Obfuscator Human Artificial Human Artificial (moral) Psychology, Law, Artificial Intelligence, (geo-)politics, Expert Systems, (behavioural) Human-Computer econom(etr)ics Interaction Artificial Intelligence, Expert Systems, Multi-agent sytems Human-Computer Interaction 11

12 Intermezzo: obfuscation vs deception Why not assume that agents try to mislead? (rather than assuming that they try to obfuscate) Protect privacy No need to mislead (agent is not malicious) Create moral wiggle room Agent does not know the right moral rule Strategic ambiguity Deceit is more costly when found out 12

13 Base Model - single rule 13

14 Notation Set contains actions Set contains rules (alternatives, options) (motivations) by matrix contains scores describing how an action performs on a given rule. +,0, : obliged (+), permitted (0), prohibited ( ) Strong rule: +, Weak rule: 0, Agent chooses one action, follows one rule 14

15 ,, example e.g. obliged by, permitted by, prohibited by,. e.g. obliges, prohibits,. 15

16 Agent beliefs about onlooker 1. Exists, watches agent 2. Observes,, ; has same perception as agent 3. Has uninformative priors about agent s rule: =1 4. Observes agent s choice, uses it to update beliefs about rules using Bayes rule: Posterior probability that agent uses rule ", conditional on observing action #. = Probability that action # is chosen if rule " is followed! Prior probability 16

17 Probability that action # is chosen if rule " is followed Strong rule =1 if # is obliged under " =0 if # is prohibited under " Weak rule =1 % if # is permitted under ", where % is #actions that are permitted under ". =0 if # is prohibited under " 17

18 Agent behavior 1. Rule follower agent follows his rule 2. Full obfuscator agent does not care about rule, only about obfuscating the onlooker 3. Hybrid agent willing to give up rule-compliance if obfuscation-gain big enough 4. Costless obfusc. agent obfuscates within boundary of rule-compliance 18

19 Behavior of a Rule-follower Strong rule =1 if # is obliged under " =0 if # is prohibited under " Weak rule =1 % if # is permitted under ", where % is #actions that are permitted under ". =0 if # is prohibited under " 19

20 Behavior of a Full Obfuscator The agent knows that the supervisor s updated beliefs after having witnessed him action # result in updated beliefs How to quantify onlooker s uncertainty given updated beliefs: Using the notion of information Entropy (Shannon, 1948): & = ' log! Agent s behavior characterized by: argmax!.. & 20

21 Full obfuscator example e.g. obliged by, permitted by, prohibited by,. e.g. obliges, prohibits,. 21

22 Full obfuscator example (II) Action-probabilities conditional on following a particular rule P a r =1; P a r 1 =0.5; P a r 3 =0; P a r 4 = = = = =1 3 6 =6 =0 P a 1 r =0; P a 1 r 1 =0.5; P a 1 r 3 =0.5; P a 1 r 4 =1 6 =0; 6 =6 = 1 4 ; 6 = 1 2 P a 3 r =0; P a 3 r 1 =0; P a 3 r 3 =0.5; P a 3 r 4 =0 22

23 Full obfuscator example (III) Rule-posteriors conditional on choosing action 1, 2, 3 P a r =1; P a r 1 =0.5; P a r 3 =0; P a r 4 = = = = =1 3 6 =6 =0 P a 1 r =0; P a 1 r 1 =0.5; P a 1 r 3 =0.5; P a 1 r 4 =1 6 =0; 6 =6 = 1 4 ; 6 = 1 2 P a 3 r =0; P a 3 r 1 =0; P a 3 r 3 =0.5; P a 3 r 4 =0 6 =6 =6 =0;6 =1 23

24 Obfuscator worked out example (III) Entropy resulting from choosing action 1: & = 2 3 log log 1 = Entropy resulting from choosing action 2: & 1 = 1 4 log log log 1 2 Entropy resulting from choosing action 3: & 3 = 1 log 1 =0 =0.452 Action chosen by a full obfuscator 24

25 Full obfuscator example Isn t Full obfuscation the same as choosing the action that complies with the most rules? NO. 25

26 = A B = 0 > +? A + B # rules with which the action complies Entropy associated with choosing the action

27 Behavior of a Hybrid Agent cares about complying with his rule ("), but also wishes to prevent the onlooker from learning which rule that is Modeled e.g. using utility-maximization principles: C = 1 D + D & & FGH & FIJ & FGH If D=1 full obfuscator; if D=0 rule-follower NOTE: 0 < D willing to sacrifice rule-compliance 27

28 Costless obfuscation-behavior Agent wishes to prevent the onlooker from learning which rule governs his behavior, but not willing to sacrifice compliance argmax L M & L Where N denotes the set of alternatives permitted by the agent s rule 28

29 Costless obfuscation: Example Agent following 1 will prefer 1 over to obfuscate Agent following 3 will prefer 1 over 3 to obfuscate 29

30 Model extension: multi-rule 30

31 Notation Agent cares about multiple rules, to different degrees. Denote O the agent s utility associated with score (i.e., compliance or not of action # with rule ".) Simple specification, normalized: if # complies with ": O =P WeightP represents the relative importance of rule " to the agent (may be 0, negative). 31

32 Probability that action # is chosen given rule-weight vector Q One choice model could be Logit: -dimensional vector of P s Q = exp! O exp O!! This allows for an interpretation as a conventional choice model: Multi-attribute discrete choice experiment (DCE) Analyst wishes to infer tastes/weights for attributes By estimating a choice model based on observed choices Participant to DCE may wish to hide his tastes from analyst 32

33 Agent beliefs: multi-rule case 1. Exists, watches agent 2. Observes,, ; has same perception as agent 3. Has uninformative priors about agent s rule-weights: T Q 4. Observes agent s choice, uses it to update beliefs about weights using Bayes rule: Posterior beliefs about rule-weight vector, conditional on observing action #. T Q = Probability that action # is chosen given rule-weight vector Q T Q U Q T Q VQ W Prior beliefs 33

34 Rule-follower, Full obfuscator, Hybrid agent behavior Q = exp O X!!! exp O X Implicitly assumed in DCA Rule-follower max & = Y T Q log T Q W VQ Full obfuscator max C = 1 D Z [ \Z ]^_ Z ]`a \Z ]^_ +D b [ \b ]^_ b ]`a \b ]^_ Hybrid 34

35 Appendix: Active onlooker 35

36 Behavior of active supervisor Until now: Now: (implicit) assumption of passive supervisor: only exists in mind of the agent supervisor is able to determine the set (c) of actions from which the agent chooses. Select a set c of a given size d=e, which minimizes entropy ' & g c! k < ' & h g i h c h! c m Caveat: Entropy of action is contingent on set So, all choice-set compositions must be studied 36

37 Entropy of action is contingent on set example (single-rule) = = n = = n Same action, different choice set; different entropy 37

38 Single-rule, multi-rule onlooker o p [ q r s! & g! < & h gm k o p t qi r s h!! u! & g U g Q T Q VQ < & h g i U h g i Q T Q VQ v k h! u m v Note relation with experimental design for discrete choice analysis 38

39 Single rule case example w &, 1 =0.40 w & 1, 3 =0.38 y n, =B.B ' & g ' g!! = =0.4 39

40 (much) Work to be done 1. Theoretical: relax assumptions regarding mutual belief systems ( agent knows that onlooker knows that agent ) Game theory, Epistemic logic, Normative MAS 2. Empirical: to what extent and under what conditions does behaviour feature elements of obfuscation? Moral decisions, Negotiation support, Geopolitics 3. Econometrics: formulate within DCT, study identifiability of parameters (can D be identified, jointly with Ps?) 4. Simulations: norm formation in societies (inferring norms from actions) using Agent-based models 40

Why on earth did you do that?!

Why on earth did you do that?! Why on earth did you do that?! A formal model of obfuscation-based choice Caspar 5-4-218 Chorus http://behave.tbm.tudelft.nl/ Delft University of Technology Challenge the future Background: models of decision-making

More information

Belief and Desire: On Information and its Value

Belief and Desire: On Information and its Value Belief and Desire: On Information and its Value Ariel Caticha Department of Physics University at Albany SUNY ariel@albany.edu Info-Metrics Institute 04/26/2013 1 Part 1: Belief 2 What is information?

More information

Correlated Equilibrium in Games with Incomplete Information

Correlated Equilibrium in Games with Incomplete Information Correlated Equilibrium in Games with Incomplete Information Dirk Bergemann and Stephen Morris Econometric Society Summer Meeting June 2012 Robust Predictions Agenda game theoretic predictions are very

More information

Learning Equilibrium as a Generalization of Learning to Optimize

Learning Equilibrium as a Generalization of Learning to Optimize Learning Equilibrium as a Generalization of Learning to Optimize Dov Monderer and Moshe Tennenholtz Faculty of Industrial Engineering and Management Technion Israel Institute of Technology Haifa 32000,

More information

Short Course: Multiagent Systems. Multiagent Systems. Lecture 1: Basics Agents Environments. Reinforcement Learning. This course is about:

Short Course: Multiagent Systems. Multiagent Systems. Lecture 1: Basics Agents Environments. Reinforcement Learning. This course is about: Short Course: Multiagent Systems Lecture 1: Basics Agents Environments Reinforcement Learning Multiagent Systems This course is about: Agents: Sensing, reasoning, acting Multiagent Systems: Distributed

More information

COMP3702/7702 Artificial Intelligence Week1: Introduction Russell & Norvig ch.1-2.3, Hanna Kurniawati

COMP3702/7702 Artificial Intelligence Week1: Introduction Russell & Norvig ch.1-2.3, Hanna Kurniawati COMP3702/7702 Artificial Intelligence Week1: Introduction Russell & Norvig ch.1-2.3, 3.1-3.3 Hanna Kurniawati Today } What is Artificial Intelligence? } Better know what it is first before committing the

More information

Game Theory. Professor Peter Cramton Economics 300

Game Theory. Professor Peter Cramton Economics 300 Game Theory Professor Peter Cramton Economics 300 Definition Game theory is the study of mathematical models of conflict and cooperation between intelligent and rational decision makers. Rational: each

More information

Notes on Mechanism Designy

Notes on Mechanism Designy Notes on Mechanism Designy ECON 20B - Game Theory Guillermo Ordoñez UCLA February 0, 2006 Mechanism Design. Informal discussion. Mechanisms are particular types of games of incomplete (or asymmetric) information

More information

Quantum Probability in Cognition. Ryan Weiss 11/28/2018

Quantum Probability in Cognition. Ryan Weiss 11/28/2018 Quantum Probability in Cognition Ryan Weiss 11/28/2018 Overview Introduction Classical vs Quantum Probability Brain Information Processing Decision Making Conclusion Introduction Quantum probability in

More information

Probabilistic Machine Learning. Industrial AI Lab.

Probabilistic Machine Learning. Industrial AI Lab. Probabilistic Machine Learning Industrial AI Lab. Probabilistic Linear Regression Outline Probabilistic Classification Probabilistic Clustering Probabilistic Dimension Reduction 2 Probabilistic Linear

More information

Predictive Processing in Planning:

Predictive Processing in Planning: Predictive Processing in Planning: Choice Behavior as Active Bayesian Inference Philipp Schwartenbeck Wellcome Trust Centre for Human Neuroimaging, UCL The Promise of Predictive Processing: A Critical

More information

Other-Regarding Preferences: Theory and Evidence

Other-Regarding Preferences: Theory and Evidence Other-Regarding Preferences: Theory and Evidence June 9, 2009 GENERAL OUTLINE Economic Rationality is Individual Optimization and Group Equilibrium Narrow version: Restrictive Assumptions about Objective

More information

Integrating State Constraints and Obligations in Situation Calculus

Integrating State Constraints and Obligations in Situation Calculus Integrating State Constraints and Obligations in Situation Calculus Robert Demolombe ONERA-Toulouse 2, Avenue Edouard Belin BP 4025, 31055 Toulouse Cedex 4, France. Robert.Demolombe@cert.fr Pilar Pozos

More information

CS 798: Multiagent Systems

CS 798: Multiagent Systems CS 798: Multiagent Systems and Utility Kate Larson Cheriton School of Computer Science University of Waterloo January 6, 2010 Outline 1 Self-Interested Agents 2 3 4 5 Self-Interested Agents We are interested

More information

Introduction to Game Theory

Introduction to Game Theory COMP323 Introduction to Computational Game Theory Introduction to Game Theory Paul G. Spirakis Department of Computer Science University of Liverpool Paul G. Spirakis (U. Liverpool) Introduction to Game

More information

Game-Theoretic Foundations for Norms

Game-Theoretic Foundations for Norms Game-Theoretic Foundations for Norms Guido Boella Dipartimento di Informatica Università di Torino-Italy E-mail: guido@di.unito.it Leendert van der Torre Department of Computer Science University of Luxembourg

More information

Information, Utility & Bounded Rationality

Information, Utility & Bounded Rationality Information, Utility & Bounded Rationality Pedro A. Ortega and Daniel A. Braun Department of Engineering, University of Cambridge Trumpington Street, Cambridge, CB2 PZ, UK {dab54,pao32}@cam.ac.uk Abstract.

More information

Costly Social Learning and Rational Inattention

Costly Social Learning and Rational Inattention Costly Social Learning and Rational Inattention Srijita Ghosh Dept. of Economics, NYU September 19, 2016 Abstract We consider a rationally inattentive agent with Shannon s relative entropy cost function.

More information

Rationality and Uncertainty

Rationality and Uncertainty Rationality and Uncertainty Based on papers by Itzhak Gilboa, Massimo Marinacci, Andy Postlewaite, and David Schmeidler Warwick Aug 23, 2013 Risk and Uncertainty Dual use of probability: empirical frequencies

More information

Communication with Self-Interested Experts Part II: Models of Cheap Talk

Communication with Self-Interested Experts Part II: Models of Cheap Talk Communication with Self-Interested Experts Part II: Models of Cheap Talk Margaret Meyer Nuffield College, Oxford 2013 Cheap Talk Models 1 / 27 Setting: Decision-maker (P) receives advice from an advisor

More information

Argumentation-Based Models of Agent Reasoning and Communication

Argumentation-Based Models of Agent Reasoning and Communication Argumentation-Based Models of Agent Reasoning and Communication Sanjay Modgil Department of Informatics, King s College London Outline Logic and Argumentation - Dung s Theory of Argumentation - The Added

More information

Great Expectations. Part I: On the Customizability of Generalized Expected Utility*

Great Expectations. Part I: On the Customizability of Generalized Expected Utility* Great Expectations. Part I: On the Customizability of Generalized Expected Utility* Francis C. Chu and Joseph Y. Halpern Department of Computer Science Cornell University Ithaca, NY 14853, U.S.A. Email:

More information

Modeling Bounded Rationality of Agents During Interactions

Modeling Bounded Rationality of Agents During Interactions Interactive Decision Theory and Game Theory: Papers from the 2 AAAI Workshop (WS--3) Modeling Bounded Rationality of Agents During Interactions Qing Guo and Piotr Gmytrasiewicz Department of Computer Science

More information

Mechanism Design for Argumentation-based Information-seeking and Inquiry

Mechanism Design for Argumentation-based Information-seeking and Inquiry Mechanism Design for Argumentation-based Information-seeking and Inquiry Xiuyi Fan and Francesca Toni Imperial College London, London, United Kingdom, {xf309,ft}@imperial.ac.uk Abstract. Formal argumentation-based

More information

Logic and Artificial Intelligence Lecture 13

Logic and Artificial Intelligence Lecture 13 Logic and Artificial Intelligence Lecture 13 Eric Pacuit Currently Visiting the Center for Formal Epistemology, CMU Center for Logic and Philosophy of Science Tilburg University ai.stanford.edu/ epacuit

More information

Reinforcement Learning

Reinforcement Learning Reinforcement Learning Model-Based Reinforcement Learning Model-based, PAC-MDP, sample complexity, exploration/exploitation, RMAX, E3, Bayes-optimal, Bayesian RL, model learning Vien Ngo MLR, University

More information

Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty

Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty Stochastics and uncertainty underlie all the processes of the Universe. N.N.Moiseev Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty by Iouldouz

More information

MODULE -4 BAYEIAN LEARNING

MODULE -4 BAYEIAN LEARNING MODULE -4 BAYEIAN LEARNING CONTENT Introduction Bayes theorem Bayes theorem and concept learning Maximum likelihood and Least Squared Error Hypothesis Maximum likelihood Hypotheses for predicting probabilities

More information

Probabilistic numerics for deep learning

Probabilistic numerics for deep learning Presenter: Shijia Wang Department of Engineering Science, University of Oxford rning (RLSS) Summer School, Montreal 2017 Outline 1 Introduction Probabilistic Numerics 2 Components Probabilistic modeling

More information

Lecture Notes: Self-enforcing agreements

Lecture Notes: Self-enforcing agreements Lecture Notes: Self-enforcing agreements Bård Harstad ECON 4910 March 2016 Bård Harstad (ECON 4910) Self-enforcing agreements March 2016 1 / 34 1. Motivation Many environmental problems are international

More information

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability Due: Thursday 10/15 in 283 Soda Drop Box by 11:59pm (no slip days) Policy: Can be solved in groups (acknowledge collaborators)

More information

Systems Engineering - Decision Analysis in Engineering Michael Havbro Faber Aalborg University, Denmark

Systems Engineering - Decision Analysis in Engineering Michael Havbro Faber Aalborg University, Denmark Risk and Safety Management Aalborg University, Esbjerg, 2017 Systems Engineering - Decision Analysis in Engineering Michael Havbro Faber Aalborg University, Denmark 1/50 M. H. Faber Systems Engineering,

More information

CS 188: Artificial Intelligence Spring Today

CS 188: Artificial Intelligence Spring Today CS 188: Artificial Intelligence Spring 2006 Lecture 9: Naïve Bayes 2/14/2006 Dan Klein UC Berkeley Many slides from either Stuart Russell or Andrew Moore Bayes rule Today Expectations and utilities Naïve

More information

An Example of Conflicts of Interest as Pandering Disincentives

An Example of Conflicts of Interest as Pandering Disincentives An Example of Conflicts of Interest as Pandering Disincentives Saori Chiba and Kaiwen Leong Current draft: January 205 Abstract Consider an uninformed decision maker (DM) who communicates with a partially

More information

Probabilistic Graphical Models

Probabilistic Graphical Models Probabilistic Graphical Models Introduction. Basic Probability and Bayes Volkan Cevher, Matthias Seeger Ecole Polytechnique Fédérale de Lausanne 26/9/2011 (EPFL) Graphical Models 26/9/2011 1 / 28 Outline

More information

Bayes Correlated Equilibrium and Comparing Information Structures

Bayes Correlated Equilibrium and Comparing Information Structures Bayes Correlated Equilibrium and Comparing Information Structures Dirk Bergemann and Stephen Morris Spring 2013: 521 B Introduction game theoretic predictions are very sensitive to "information structure"

More information

1 [15 points] Search Strategies

1 [15 points] Search Strategies Probabilistic Foundations of Artificial Intelligence Final Exam Date: 29 January 2013 Time limit: 120 minutes Number of pages: 12 You can use the back of the pages if you run out of space. strictly forbidden.

More information

Preliminary Statistics Lecture 2: Probability Theory (Outline) prelimsoas.webs.com

Preliminary Statistics Lecture 2: Probability Theory (Outline) prelimsoas.webs.com 1 School of Oriental and African Studies September 2015 Department of Economics Preliminary Statistics Lecture 2: Probability Theory (Outline) prelimsoas.webs.com Gujarati D. Basic Econometrics, Appendix

More information

On the errors introduced by the naive Bayes independence assumption

On the errors introduced by the naive Bayes independence assumption On the errors introduced by the naive Bayes independence assumption Author Matthijs de Wachter 3671100 Utrecht University Master Thesis Artificial Intelligence Supervisor Dr. Silja Renooij Department of

More information

Imprecise Probability

Imprecise Probability Imprecise Probability Alexander Karlsson University of Skövde School of Humanities and Informatics alexander.karlsson@his.se 6th October 2006 0 D W 0 L 0 Introduction The term imprecise probability refers

More information

CS343 Artificial Intelligence

CS343 Artificial Intelligence CS343 Artificial Intelligence Prof: Department of Computer Science The University of Texas at Austin Good Afternoon, Colleagues Good Afternoon, Colleagues Are there any questions? Logistics Problems with

More information

Logic and Artificial Intelligence Lecture 12

Logic and Artificial Intelligence Lecture 12 Logic and Artificial Intelligence Lecture 12 Eric Pacuit Currently Visiting the Center for Formal Epistemology, CMU Center for Logic and Philosophy of Science Tilburg University ai.stanford.edu/ epacuit

More information

Towards a General Theory of Non-Cooperative Computation

Towards a General Theory of Non-Cooperative Computation Towards a General Theory of Non-Cooperative Computation (Extended Abstract) Robert McGrew, Ryan Porter, and Yoav Shoham Stanford University {bmcgrew,rwporter,shoham}@cs.stanford.edu Abstract We generalize

More information

9/12/17. Types of learning. Modeling data. Supervised learning: Classification. Supervised learning: Regression. Unsupervised learning: Clustering

9/12/17. Types of learning. Modeling data. Supervised learning: Classification. Supervised learning: Regression. Unsupervised learning: Clustering Types of learning Modeling data Supervised: we know input and targets Goal is to learn a model that, given input data, accurately predicts target data Unsupervised: we know the input only and want to make

More information

Multi-Attribute Bayesian Optimization under Utility Uncertainty

Multi-Attribute Bayesian Optimization under Utility Uncertainty Multi-Attribute Bayesian Optimization under Utility Uncertainty Raul Astudillo Cornell University Ithaca, NY 14853 ra598@cornell.edu Peter I. Frazier Cornell University Ithaca, NY 14853 pf98@cornell.edu

More information

Basics of reinforcement learning

Basics of reinforcement learning Basics of reinforcement learning Lucian Buşoniu TMLSS, 20 July 2018 Main idea of reinforcement learning (RL) Learn a sequential decision policy to optimize the cumulative performance of an unknown system

More information

Models of Reputation with Bayesian Updating

Models of Reputation with Bayesian Updating Models of Reputation with Bayesian Updating Jia Chen 1 The Tariff Game (Downs and Rocke 1996) 1.1 Basic Setting Two states, A and B, are setting the tariffs for trade. The basic setting of the game resembles

More information

Understanding Aha! Moments

Understanding Aha! Moments Understanding Aha! Moments Hermish Mehta Department of Electrical Engineering & Computer Sciences University of California, Berkeley Berkeley, CA 94720 hermish@berkeley.edu Abstract In this paper, we explore

More information

Deliberation-aware Responder in Multi-Proposer Ultimatum Game

Deliberation-aware Responder in Multi-Proposer Ultimatum Game Deliberation-aware Responder in Multi-Proposer Ultimatum Game Marko Ruman, František Hůla, Miroslav Kárný, and Tatiana V. Guy Department of Adaptive Systems Institute of Information Theory and Automation

More information

Probabilistic Reasoning

Probabilistic Reasoning Course 16 :198 :520 : Introduction To Artificial Intelligence Lecture 7 Probabilistic Reasoning Abdeslam Boularias Monday, September 28, 2015 1 / 17 Outline We show how to reason and act under uncertainty.

More information

Mechanism Design for Computationally Limited Agents

Mechanism Design for Computationally Limited Agents Mechanism Design for Computationally Limited Agents Kate Larson August 2004 CMU-CS-04-152 School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Submitted in partial fulfillment of

More information

CS 188: Artificial Intelligence Fall 2008

CS 188: Artificial Intelligence Fall 2008 CS 188: Artificial Intelligence Fall 2008 Lecture 14: Bayes Nets 10/14/2008 Dan Klein UC Berkeley 1 1 Announcements Midterm 10/21! One page note sheet Review sessions Friday and Sunday (similar) OHs on

More information

A Probabilistic Relational Model for Characterizing Situations in Dynamic Multi-Agent Systems

A Probabilistic Relational Model for Characterizing Situations in Dynamic Multi-Agent Systems A Probabilistic Relational Model for Characterizing Situations in Dynamic Multi-Agent Systems Daniel Meyer-Delius 1, Christian Plagemann 1, Georg von Wichert 2, Wendelin Feiten 2, Gisbert Lawitzky 2, and

More information

THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE

THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE RACHEL CUMMINGS, FEDERICO ECHENIQUE, AND ADAM WIERMAN Abstract. This paper initiates the study of the testable implications of choice data in settings

More information

COMP3702/7702 Artificial Intelligence Lecture 11: Introduction to Machine Learning and Reinforcement Learning. Hanna Kurniawati

COMP3702/7702 Artificial Intelligence Lecture 11: Introduction to Machine Learning and Reinforcement Learning. Hanna Kurniawati COMP3702/7702 Artificial Intelligence Lecture 11: Introduction to Machine Learning and Reinforcement Learning Hanna Kurniawati Today } What is machine learning? } Where is it used? } Types of machine learning

More information

Bounded Rationality, Strategy Simplification, and Equilibrium

Bounded Rationality, Strategy Simplification, and Equilibrium Bounded Rationality, Strategy Simplification, and Equilibrium UPV/EHU & Ikerbasque Donostia, Spain BCAM Workshop on Interactions, September 2014 Bounded Rationality Frequently raised criticism of game

More information

Endogenous Information Choice

Endogenous Information Choice Endogenous Information Choice Lecture 7 February 11, 2015 An optimizing trader will process those prices of most importance to his decision problem most frequently and carefully, those of less importance

More information

Learning in Zero-Sum Team Markov Games using Factored Value Functions

Learning in Zero-Sum Team Markov Games using Factored Value Functions Learning in Zero-Sum Team Markov Games using Factored Value Functions Michail G. Lagoudakis Department of Computer Science Duke University Durham, NC 27708 mgl@cs.duke.edu Ronald Parr Department of Computer

More information

COMP90051 Statistical Machine Learning

COMP90051 Statistical Machine Learning COMP90051 Statistical Machine Learning Semester 2, 2017 Lecturer: Trevor Cohn 17. Bayesian inference; Bayesian regression Training == optimisation (?) Stages of learning & inference: Formulate model Regression

More information

On the Consistency among Prior, Posteriors, and Information Sets

On the Consistency among Prior, Posteriors, and Information Sets On the Consistency among Prior, Posteriors, and Information Sets Satoshi Fukuda September 23, 2018 Abstract This paper studies implications of the consistency conditions among prior, posteriors, and information

More information

The Lady Tasting Tea. How to deal with multiple testing. Need to explore many models. More Predictive Modeling

The Lady Tasting Tea. How to deal with multiple testing. Need to explore many models. More Predictive Modeling The Lady Tasting Tea More Predictive Modeling R. A. Fisher & the Lady B. Muriel Bristol claimed she prefers tea added to milk rather than milk added to tea Fisher was skeptical that she could distinguish

More information

Relative Benefit Equilibrating Bargaining Solution and the Ordinal Interpretation of Gauthier s Arbitration Scheme

Relative Benefit Equilibrating Bargaining Solution and the Ordinal Interpretation of Gauthier s Arbitration Scheme Relative Benefit Equilibrating Bargaining Solution and the Ordinal Interpretation of Gauthier s Arbitration Scheme Mantas Radzvilas July 2017 Abstract In 1986 David Gauthier proposed an arbitration scheme

More information

CS 4649/7649 Robot Intelligence: Planning

CS 4649/7649 Robot Intelligence: Planning CS 4649/7649 Robot Intelligence: Planning Probability Primer Sungmoon Joo School of Interactive Computing College of Computing Georgia Institute of Technology S. Joo (sungmoon.joo@cc.gatech.edu) 1 *Slides

More information

Structure learning in human causal induction

Structure learning in human causal induction Structure learning in human causal induction Joshua B. Tenenbaum & Thomas L. Griffiths Department of Psychology Stanford University, Stanford, CA 94305 jbt,gruffydd @psych.stanford.edu Abstract We use

More information

Introduction to Game Theory. Outline. Topics. Recall how we model rationality. Notes. Notes. Notes. Notes. Tyler Moore.

Introduction to Game Theory. Outline. Topics. Recall how we model rationality. Notes. Notes. Notes. Notes. Tyler Moore. Introduction to Game Theory Tyler Moore Tandy School of Computer Science, University of Tulsa Slides are modified from version written by Benjamin Johnson, UC Berkeley Lecture 15 16 Outline 1 Preferences

More information

CSE250A Fall 12: Discussion Week 9

CSE250A Fall 12: Discussion Week 9 CSE250A Fall 12: Discussion Week 9 Aditya Menon (akmenon@ucsd.edu) December 4, 2012 1 Schedule for today Recap of Markov Decision Processes. Examples: slot machines and maze traversal. Planning and learning.

More information

Marks. bonus points. } Assignment 1: Should be out this weekend. } Mid-term: Before the last lecture. } Mid-term deferred exam:

Marks. bonus points. } Assignment 1: Should be out this weekend. } Mid-term: Before the last lecture. } Mid-term deferred exam: Marks } Assignment 1: Should be out this weekend } All are marked, I m trying to tally them and perhaps add bonus points } Mid-term: Before the last lecture } Mid-term deferred exam: } This Saturday, 9am-10.30am,

More information

Epistemic Foundations for Set-algebraic Representations of Knowledge

Epistemic Foundations for Set-algebraic Representations of Knowledge Epistemic Foundations for Set-algebraic Representations of Knowledge Satoshi Fukuda September 10, 2018 Abstract This paper formalizes an informal idea that an agent s knowledge is characterized by a collection

More information

Choice Theory. Matthieu de Lapparent

Choice Theory. Matthieu de Lapparent Choice Theory Matthieu de Lapparent matthieu.delapparent@epfl.ch Transport and Mobility Laboratory, School of Architecture, Civil and Environmental Engineering, Ecole Polytechnique Fédérale de Lausanne

More information

AIS 2016 AIS 2014 AIS 2015

AIS 2016 AIS 2014 AIS 2015 Interne score ASMF journal subject category Economics or Business, Finance on ISI Web of Knowledge TI selected Marketing or OR journal Journal 2013 2014 2015 2016 AVERAGE Advances in Water Resources 1,246

More information

Neural Networks & Fuzzy Logic

Neural Networks & Fuzzy Logic Journal of Computer Applications ISSN: 0974 1925, Volume-5, Issue EICA2012-4, February 10, 2012 Neural Networks & Fuzzy Logic Elakkiya Prabha T Pre-Final B.Tech-IT, M.Kumarasamy College of Engineering,

More information

ESSENCE 2014: Argumentation-Based Models of Agent Reasoning and Communication

ESSENCE 2014: Argumentation-Based Models of Agent Reasoning and Communication ESSENCE 2014: Argumentation-Based Models of Agent Reasoning and Communication Sanjay Modgil Department of Informatics, King s College London Outline Logic, Argumentation and Reasoning - Dung s Theory of

More information

SpringerBriefs in Statistics

SpringerBriefs in Statistics SpringerBriefs in Statistics For further volumes: http://www.springer.com/series/8921 Jeff Grover Strategic Economic Decision-Making Using Bayesian Belief Networks to Solve Complex Problems Jeff Grover

More information

Desire-as-belief revisited

Desire-as-belief revisited Desire-as-belief revisited Richard Bradley and Christian List June 30, 2008 1 Introduction On Hume s account of motivation, beliefs and desires are very di erent kinds of propositional attitudes. Beliefs

More information

Today s Outline. Recap: MDPs. Bellman Equations. Q-Value Iteration. Bellman Backup 5/7/2012. CSE 473: Artificial Intelligence Reinforcement Learning

Today s Outline. Recap: MDPs. Bellman Equations. Q-Value Iteration. Bellman Backup 5/7/2012. CSE 473: Artificial Intelligence Reinforcement Learning CSE 473: Artificial Intelligence Reinforcement Learning Dan Weld Today s Outline Reinforcement Learning Q-value iteration Q-learning Exploration / exploitation Linear function approximation Many slides

More information

Don t Plan for the Unexpected: Planning Based on Plausibility Models

Don t Plan for the Unexpected: Planning Based on Plausibility Models Don t Plan for the Unexpected: Planning Based on Plausibility Models Thomas Bolander, DTU Informatics, Technical University of Denmark Joint work with Mikkel Birkegaard Andersen and Martin Holm Jensen

More information

Bayesian Learning Extension

Bayesian Learning Extension Bayesian Learning Extension This document will go over one of the most useful forms of statistical inference known as Baye s Rule several of the concepts that extend from it. Named after Thomas Bayes this

More information

Lecture Notes 1: Decisions and Data. In these notes, I describe some basic ideas in decision theory. theory is constructed from

Lecture Notes 1: Decisions and Data. In these notes, I describe some basic ideas in decision theory. theory is constructed from Topics in Data Analysis Steven N. Durlauf University of Wisconsin Lecture Notes : Decisions and Data In these notes, I describe some basic ideas in decision theory. theory is constructed from The Data:

More information

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability Due: Thursday 10/15 in 283 Soda Drop Box by 11:59pm (no slip days) Policy: Can be solved in groups (acknowledge collaborators)

More information

Lecture 16 Deep Neural Generative Models

Lecture 16 Deep Neural Generative Models Lecture 16 Deep Neural Generative Models CMSC 35246: Deep Learning Shubhendu Trivedi & Risi Kondor University of Chicago May 22, 2017 Approach so far: We have considered simple models and then constructed

More information

Overview, cont. Overview, cont. Logistics. Optional Reference #1. Optional Reference #2. Workload and Grading

Overview, cont. Overview, cont. Logistics. Optional Reference #1. Optional Reference #2. Workload and Grading Course staff CS389L: Automated Logical Reasoning Lecture 1: ntroduction and Review of Basics şıl Dillig nstructor: şil Dillig E-mail: isil@cs.utexas.edu Office hours: Thursday after class until 6:30 pm

More information

The Social Value of Credible Public Information

The Social Value of Credible Public Information The Social Value of Credible Public Information Ercan Karadas NYU September, 2017 Introduction Model Analysis MOTIVATION This paper is motivated by the paper Social Value of Public Information, Morris

More information

Coherence with Proper Scoring Rules

Coherence with Proper Scoring Rules Coherence with Proper Scoring Rules Mark Schervish, Teddy Seidenfeld, and Joseph (Jay) Kadane Mark Schervish Joseph ( Jay ) Kadane Coherence with Proper Scoring Rules ILC, Sun Yat-Sen University June 2010

More information

CS 188: Artificial Intelligence

CS 188: Artificial Intelligence CS 188: Artificial Intelligence Adversarial Search II Instructor: Anca Dragan University of California, Berkeley [These slides adapted from Dan Klein and Pieter Abbeel] Minimax Example 3 12 8 2 4 6 14

More information

BELIEFS & EVOLUTIONARY GAME THEORY

BELIEFS & EVOLUTIONARY GAME THEORY 1 / 32 BELIEFS & EVOLUTIONARY GAME THEORY Heinrich H. Nax hnax@ethz.ch & Bary S. R. Pradelski bpradelski@ethz.ch May 15, 217: Lecture 1 2 / 32 Plan Normal form games Equilibrium invariance Equilibrium

More information

Nonlinear Dynamics between Micromotives and Macrobehavior

Nonlinear Dynamics between Micromotives and Macrobehavior Nonlinear Dynamics between Micromotives and Macrobehavior Saori Iwanaga & kira Namatame Dept. of Computer Science, National Defense cademy, Yokosuka, 239-8686, JPN, E-mail: {g38042, nama}@nda.ac.jp Tel:

More information

Reinforcement Learning. Introduction

Reinforcement Learning. Introduction Reinforcement Learning Introduction Reinforcement Learning Agent interacts and learns from a stochastic environment Science of sequential decision making Many faces of reinforcement learning Optimal control

More information

Chapter Three. Hypothesis Testing

Chapter Three. Hypothesis Testing 3.1 Introduction The final phase of analyzing data is to make a decision concerning a set of choices or options. Should I invest in stocks or bonds? Should a new product be marketed? Are my products being

More information

Intrinsic and Extrinsic Motivation

Intrinsic and Extrinsic Motivation Intrinsic and Extrinsic Motivation Roland Bénabou Jean Tirole. Review of Economic Studies 2003 Bénabou and Tirole Intrinsic and Extrinsic Motivation 1 / 30 Motivation Should a child be rewarded for passing

More information

Markov Decision Processes

Markov Decision Processes Markov Decision Processes Noel Welsh 11 November 2010 Noel Welsh () Markov Decision Processes 11 November 2010 1 / 30 Annoucements Applicant visitor day seeks robot demonstrators for exciting half hour

More information

ECE521 week 3: 23/26 January 2017

ECE521 week 3: 23/26 January 2017 ECE521 week 3: 23/26 January 2017 Outline Probabilistic interpretation of linear regression - Maximum likelihood estimation (MLE) - Maximum a posteriori (MAP) estimation Bias-variance trade-off Linear

More information

School of EECS Washington State University. Artificial Intelligence

School of EECS Washington State University. Artificial Intelligence School of EECS Washington State University 1 } Webpage: www.eecs.wsu.edu/~holder/courses/ai } Email (holder@wsu.edu) } Blackboard Learn (learn.wsu.edu) } mywsu (my.wsu.edu) 2 Readings: Chapter 1 3 } John

More information

Lecture : Probabilistic Machine Learning

Lecture : Probabilistic Machine Learning Lecture : Probabilistic Machine Learning Riashat Islam Reasoning and Learning Lab McGill University September 11, 2018 ML : Many Methods with Many Links Modelling Views of Machine Learning Machine Learning

More information

Introduction to game theory LECTURE 1

Introduction to game theory LECTURE 1 Introduction to game theory LECTURE 1 Jörgen Weibull January 27, 2010 1 What is game theory? A mathematically formalized theory of strategic interaction between countries at war and peace, in federations

More information

Bounded Rationality Lecture 4

Bounded Rationality Lecture 4 Bounded Rationality Lecture 4 Mark Dean Princeton University - Behavioral Economics The Story So Far... Introduced the concept of bounded rationality Described some behaviors that might want to explain

More information

RSMG Working Paper Series. TITLE: The value of information and the value of awareness. Author: John Quiggin. Working Paper: R13_2

RSMG Working Paper Series. TITLE: The value of information and the value of awareness. Author: John Quiggin. Working Paper: R13_2 2013 TITLE: The value of information and the value of awareness 2011 RSMG Working Paper Series Risk and Uncertainty Program Author: John Quiggin Working Paper: R13_2 Schools of Economics and Political

More information

COMP219: Artificial Intelligence. Lecture 19: Logic for KR

COMP219: Artificial Intelligence. Lecture 19: Logic for KR COMP219: Artificial Intelligence Lecture 19: Logic for KR 1 Overview Last time Expert Systems and Ontologies Today Logic as a knowledge representation scheme Propositional Logic Syntax Semantics Proof

More information

Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric?

Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric? Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric? Zoubin Ghahramani Department of Engineering University of Cambridge, UK zoubin@eng.cam.ac.uk http://learning.eng.cam.ac.uk/zoubin/

More information

PLANNING (PLAN) Planning (PLAN) 1

PLANNING (PLAN) Planning (PLAN) 1 Planning (PLAN) 1 PLANNING (PLAN) PLAN 500. Economics for Public Affairs Description: An introduction to basic economic concepts and their application to public affairs and urban planning. Note: Cross-listed

More information

Parameter Estimation. Industrial AI Lab.

Parameter Estimation. Industrial AI Lab. Parameter Estimation Industrial AI Lab. Generative Model X Y w y = ω T x + ε ε~n(0, σ 2 ) σ 2 2 Maximum Likelihood Estimation (MLE) Estimate parameters θ ω, σ 2 given a generative model Given observed

More information