Deceptive AI (Communications in Computer and Information Science)

✍ Scribed by Stefan Sarkadi (editor), Benjamin Wright (editor), Peta Masters (editor), Peter McBurney (editor)

• Publisher: Springer
• Year: 2021
• Tongue: English
• Leaves: 182
• Category: Library

✦ Synopsis

This book constitutes selected papers presented at the First International Workshop on Deceptive AI, DeceptECAI 2020, held in conjunction with the 24th European Conference on Artificial Intelligence, ECAI 2020, in Santiago de Compostela, Spain, in August 2020, and Second International Workshop on Deceptive AI, DeceptAI 2021, held in conjunction with the 30th International Joint Conference on Artificial Intelligence, IJCAI 2021, in Montreal, Canada, in August 2021. Due to the COVID-19 pandemic both conferences were held in a virtual mode.
The 12 papers presented were thoroughly reviewed and selected from the 16 submissions. They present recent developments in the growing area of research in the interface between deception and AI.

✦ Table of Contents

Preface
Organization
Contents
Conceptual Frameworks
Characterising Deception in AI: A Survey
1 Introduction
2 The Origin of the Deceptive Categories in the AI Communities
3 Illustrating the Categories
3.1 Imitating
3.2 Obfuscating
3.3 Tricking
3.4 Calculating
3.5 Reframing
3.6 Exceptions and Exclusions
4 Conclusion
References
The Role of Environments in Affording Deceptive Behaviour: Some Preliminary Insights from Stage Magic
1 Introduction
2 The MindTrails Game
3 How MindTrails Affords Deceptive Techniques from Stage Magic
3.1 Given and Hidden Stories
3.2 Confidence
3.3 Relevance and Irrelevance
3.4 Attentional Misdirection
3.5 Time Displacement
3.6 Convincers
4 Discussion
References
Deceptive Dreams: Nudging for Better Sleep
1 Introduction
2 Nudging, Heuristics and Cognitive Biases
3 Scenarios and Prototypes
3.1 Scenario 1 – Rise and Shine
3.2 Scenario 2 – Step Tracker
3.3 Scenario 3 – the Coffee Break
3.4 Scenario 4 – Night Lights
3.5 Scenario 5 – Display Brightness
3.6 Scenario 6 – Setting the Alarm
4 The Two Types of Deceptions
5 Discussion
6 Conclusion
References
Agent-Based Approches
Wolves in Sheep’s Clothing: Using Shill Agents to Misdirect Multi-robot Teams
1 Introduction
2 Related Work
3 Robot Models
3.1 Behavior Overviews
3.2 Mathematical Models
3.3 Robot Missions
4 Misdirection Simulations
4.1 Results
4.2 Discussion
5 Conclusions and Ongoing Work
Appendix I
Appendix II
References
Learning to Deceive in Multi-agent Hidden Role Games
1 Introduction
2 Background and Related Work
3 Rescue the General
3.1 Teams and Objectives
3.2 Observational Space
3.3 Scenarios
3.4 Reward Structure
4 Bayesian Belief Manipulation Model
4.1 Theory of Mind
4.2 Assumptions
4.3 Notation
4.4 The Model
4.5 Direct Policy Prediction
4.6 Deception Bonus
5 Training and Evaluation
5.1 Agent Architecture
5.2 Training Details
5.3 Evaluation
6 Effect of Deception on Agents' Behaviour
6.1 Role Prediction
6.2 Performance of Deceptive Agents
6.3 Deceptive Tendencies in Unincentivised Agents
7 Conclusions and Future Work
A Scenario Configuration Options
B Encoder Architecture
C Hyperparameter Search
References
Modelling Strategic Deceptive Planning in Adversarial Multi-agent Systems
1 Introduction
2 Related Work
3 Modelling Strategic Agent Deception
3.1 Agent Behaviour Models in Operations Research
3.2 A Strategic Deceptive Planning Model
3.3 Novel Deceptive Behaviour Discovery
3.4 Deception Feasibility
3.5 Determining Exploitable Uncertainties
4 Conclusion
References
Logic-Based Approches
E-Friend: A Logical-Based AI Agent System Chat-Bot for Emotional Well-Being and Mental Health
1 Introduction
2 Related Work
3 Design of Architectural Framework for a Logical Based Intelligent Agent System Chat-Bot
3.1 E-Friend General Approach
3.2 A Master-Slave AI Design
3.3 The E-Friend Intelligent Agent System Chat-Bot
4 Answer Set Programming Solution Proposal
5 Conclusions and Future Work
References
Deception in Epistemic Causal Logic
1 Introduction
2 Epistemic Causal Logic
3 Deception in Epistemic Causal Logic
3.1 Deception by Lying
3.2 Deception by Bluffing
3.3 Deception by Truthful Telling
3.4 Deception by Omission
4 Various Aspects of Deception
4.1 Intentional Deception
4.2 Indirect Deception
4.3 Self-deception
5 Related Work
6 Concluding Remarks
References
Influencing Choices by Changing Beliefs: A Logical Theory of Influence, Persuasion, and Deception
1 Introduction
2 Logical Framework
2.1 Syntax
2.2 Semantics
2.3 Axiomatization
3 Agency and Rationality Types
4 Influence, Persuasion and Deception
5 Conclusion
References
Machine Learning Approches
Towards a Framework for Challenging ML-Based Decisions
1 Introduction
2 Challenges and Justifications
3 Outline of the Framework
3.1 Challenges
3.2 Justifications
3.3 Generation of Evidence
3.4 Possible Outcomes
4 User Study
5 Related Works
6 Conclusion
A User Study: Methods and Results
B Rule Search Algorithm
References
Impostor GAN: Toward Modeling Social Media User Impersonation with Generative Adversarial Networks
1 Introduction
2 Background and Related Work
3 Methodology
4 Dataset Description
5 Experiments
6 Evaluation
7 Conclusion
References
Studying Dishonest Intentions in Brazilian Portuguese Texts
1 Introduction
2 Related Work
3 Data Overview
4 Linguistic Features
4.1 Word-Level Analysis
4.2 Sentence-Level Analysis
5 Final Remarks and Future Work
References
Author Index