Autonomy and Artificial Intelligence: A Threat or Savior? (eBook)

Contributing authors: Kevin Barry, Patrick Benavidez, Chris Berka, Joseph Coyne, Boris A. Galitsky, Peter Gerken, Sri Nikhil Gupta Gourisetti, Rachel Hingst, Ayanna Howard, Mo Jamshidi, W.F. Lawless, James Llinas, Jonathan Lwowski, Ranjeev Mittu, Ira S. Moskowitz, Michael Mylrea, Anna Parnis, John J. Prevost, Adrienne Raglin, Signe A. Redfield, Paul Robinette, Galina Rogova, Stephen Russell, Alicia Ruvinsky, Mae L. Seto, Sarah Sherwood, Ciara Sibley, Donald Sofge, Douglas Summers Stay, Maja Stikic, Catherine Tessier, Alan R. Wagner.

Preface 5
Spring 2015: Foundations of Autonomy and Its (Cyber) Threats—From Individuals to Interdependence 6
Spring 2015: Organizing Committee 6
Spring 2015: Program Committee 6
Spring 2015: Invited Keynote Speakers 7
Spring 2015: Regular Speakers 8
Spring 2016: AI and the Mitigation of Human Error—Anomalies, Team Metrics and Thermodynamics 9
Spring 2016: Organizing Committee 9
Spring 2016: Program Committee (duplicates the spring 2015 symposium) 9
Spring 2016: Invited Keynote Speakers 9
Spring 2016: Regular Speakers 10
Questions for Speakers and Attendees at AAAI-2015 and AAAI-2016 and for Readers of This Book 10
Contents 13
Chapter 1: Introduction 15
1.1 Background of the 2015 Symposium 15
1.2 Background of the 2016 Symposium 16
1.3 Contributed Chapters 17
References 24
Chapter 2: Reexamining Computational Support for Intelligence Analysis: A Functional Design for a Future Capability 26
2.1 Motivation 26
2.2 Goals and Requirements 27
2.3 Future Directions in Intelligence Analysis 28
2.3.1 Reviews of Open Literature and Operational Environments 28
2.3.2 Analytical Rigor in Intelligence Analysis/Argument Mapping 29
2.4 Approaches to Computational Support 30
2.4.1 Paradigms and Methods 30
2.4.2 Argumentation Methods 32
2.4.3 Computational Support to Argumentation: The State of the Art 34
2.4.3.1 Argument Detection 37
Moen et al. (2007) Automatic Detection of Arguments in Legal Texts 37
Mochales-Palau and Moens (2007) 37
Feng and Hirst (2011), Classifying Arguments by Scheme 38
2.4.3.2 Argument Mining 39
Moens (2013), State of the Art in Argument Mining 39
2.4.3.3 Argument Invention 40
Walton and Gordon (2012), the Carneades Model of Argument Invention 40
2.4.3.4 Argument Visualization (a.k.a. Mapping, Diagramming) 41
2.5 Current-Day Computational Support to Argumentation 45
2.5.1 AVERS and CISpaces as Leading Relevant Prototypes 45
2.6 Computational Support for Narrative Development 47
2.6.1 Using Topic Modeling to Assess Story Relevance and Narrative Formation 47
2.7 Developing a Functional Design for an Advanced-­Capability Prototype 51
2.7.1 Looking Ahead: Possible Test and Evaluation Schemes 55
References 56
Chapter 3: Task Allocation Using Parallelized Clustering and Auctioning Algorithms for Heterogeneous Robotic Swarms Operating on a Cloud Network 60
3.1 Introduction 60
3.2 Robotic Swarm Methodology 62
3.2.1 Scenario 62
3.2.2 System Overview 62
3.2.3 Localization of People 63
3.2.4 Building the Map 65
3.2.4.1 Clustering the People 65
3.2.4.2 Parallelization of Clustering Algorithm 67
3.2.4.3 Meta-Clustering the Clusters 69
3.2.5 Auctioning the Meta-Clusters 71
3.2.6 Traveling to the Assigned Clusters 72
3.2.6.1 Traveling Salesman Solver 72
3.2.6.2 Human Interaction with Swarm 74
3.3 Experimental Results 75
3.3.1 Simulation Results 75
3.3.2 Hardware Emulation Results 76
3.3.2.1 Unmanned Ground Vehicle (UGV) 76
3.3.2.2 GPS Emulation 77
3.3.2.3 Traveling Salesman 78
3.3.2.4 CCR K-Means Clustering 79
3.3.2.5 Meta-Clustering 79
3.3.2.6 Auction Algorithm 80
3.4 Conclusions 80
References 82
Chapter 4: Human Information Interaction, Artificial Intelligence, and Errors 83
4.1 Introduction 83
4.2 Human Information Interaction 85
4.3 HII and Artificial Intelligence 95
4.4 HII, AI, and Errors 99
4.5 Conclusion 107
References 108
Chapter 5: Verification Challenges for Autonomous Systems 114
5.1 Introduction 114
5.2 Autonomy 115
5.2.1 Benefits of Autonomy 116
5.3 Verification 117
5.3.1 Verification Implications of Autonomy 118
5.3.2 Example System 120
5.4 Challenges 121
5.4.1 Models 121
5.4.2 Abstraction 123
5.4.2.1 Fidelity 123
5.4.2.2 Requirements Generation 124
5.4.3 Test 126
5.4.3.1 Scenarios 126
5.4.3.2 Metrics and Performance Evaluation 128
5.4.3.3 Intersection of Scenarios and Metrics 131
5.4.4 Tools 131
5.5 Summary 135
References 137
Chapter 6: Conceptualizing Overtrust in Robots: Why Do People Trust a Robot That Previously Failed? 139
6.1 Introduction 139
6.2 Conceptualizing Overtrust 140
6.3 Robot Guidance Versus Existing Guidance Technology 142
6.3.1 Experimental Setup 142
6.3.2 Results 145
6.4 Human-Robot Trust in Virtual Simulations 147
6.5 Repairing Broken Trust 151
6.5.1 Experimental Setup 151
6.5.2 Results 153
6.6 Overtrust of Robots in Physical Situations 158
6.6.1 Experimental Setup 158
6.6.2 Results 160
6.7 Discussion 161
6.8 Thoughts on Future Work 163
References 164
Chapter 7: Research Considerations and Tools for Evaluating Human-Automation Interaction with Future Unmanned Systems 166
7.1 The Current Environment and Future Vision 166
7.2 Calibrating Trust in Automation 168
7.3 DoD Plans and Guides 169
7.4 Supervisory Control Research and Testing Environments 170
7.4.1 The Adaptive Levels of Automation Test Bed and Research 171
7.4.2 The Research Environment for Supervisory Control of Heterogeneous Unmanned Vehicles Test Bed and Research 173
7.5 Supervisory Control Research Limitations and Challenges 174
7.6 Assessing Human-Automation Performance 175
7.6.1 The Value of Eye Tracking 176
7.7 Supervisory Control Operations User Testbed (SCOUT) Overview 179
7.8 Summary 183
References 185
Chapter 8: Robots Autonomy: Some Technical Issues 188
8.1 Introduction 188
8.2 What Is a Robot? 188
8.3 Autonomy 190
8.3.1 What Is Autonomy? 190
8.3.2 Authority Sharing 192
8.4 Autonomy and Authority Sharing: Some Questions 193
8.4.1 The Robot 193
8.4.1.1 Situation Tracking: Interpretation and Assessment of the Situation 193
8.4.1.2 Decision 194
8.4.2 The Human Operator 195
8.4.3 The Operator-Robot Interaction 195
8.5 Autonomy and Authority Sharing Ethical Challenges 198
8.5.1 Why Imbue a Robot with Ethics? 199
8.5.2 A Careful Approach Is Needed 199
8.5.3 Thought Experiments Usefulness 200
8.6 Conclusion: Some Prospects for Robots Autonomy 201
References 202
Chapter 9: How Children with Autism and Machines Learn to Interact 204
9.1 Introduction 204
9.2 From Hypersensitivity to Limited Interaction with the World 206
9.2.1 Hypersensitivity 206
9.2.2 Active Learning in Computer Science 207
9.2.3 Learning Repetitive Patterns 209
9.2.4 Self-Stimulation 210
9.2.5 Not Paying Attention to What Is Important 212
9.2.6 From Hyper-Sensitivity to Self-Stimulation of an Engineering System 213
9.3 Building and Revising Hypotheses in Active Human Learning 215
9.4 Building Teams Having Learned to Interact 218
9.4.1 How Trust Develops in a Baby 218
9.4.2 Measuring Skills of Reasoning About Mental World 219
9.4.3 A Cooperation Between CwA in the Real World 222
9.5 Rehabilitating Autistic Interactions 224
9.5.1 Teaching Hide-and-Seek Game 224
9.5.2 Learning to Navigate Environment 226
9.5.3 A Literary Work Search System 227
9.6 Discussion and Conclusions 232
References 234
Chapter 10: Semantic Vector Spaces for Broadening Consideration of Consequences 236
10.1 Designing for Safety 236
10.2 Understanding Intent 238
10.3 Expressing Intent 239
10.4 Problem 1: An Encoding for Concepts 240
10.5 Semantic Vector Spaces 242
10.6 Problem 2: Distributional Semantic Vector Spaces 244
10.7 What Needs to be Done 248
10.7.1 Learning More Complex Relations 248
10.7.2 Distributional Semantics 249
10.7.3 Semantics from Images, Video, and Other Data Streams 249
10.7.4 Combining Two Vector Spaces to Better Capture the Knowledge Learned from Each 249
10.7.5 Encoding the Meaning of Natural Language Phrases and Sentences as Vectors 250
10.7.6 Modifying a Semantic Vector Space as New Information Is Learned Without Destroying Already Existing Structure 250
10.7.7 Performing Reasoning Within Vector Spaces 250
10.7.8 Ways of Discovering and Representing Knowledge About Physical Consequences 251
10.8 Conclusion 251
References 251
Chapter 11: On the Road to Autonomy: Evaluating and Optimizing Hybrid Team Dynamics 253
11.1 Introduction 253
11.2 Teaming Platform 255
11.3 Teaming Studies 260
11.4 Neurophysiologic Synchronies 260
11.5 EEG Predictors of Team Performance 261
11.6 Narrative Storytelling 262
11.7 Tutoring Dyads 263
11.8 Quality of Surgical Operations 265
11.9 Discussion 265
11.10 Future Research Directions 267
References 268
Chapter 12: Cybersecurity and Optimization in Smart “Autonomous” Buildings 271
12.1 Introduction 272
12.2 Smart Building Opportunity 273
12.3 Smart Building Challenges 274
12.4 AI Enabled Building Automation Is Blurring the Lines Between Information Technology and Operations Technology 279
12.5 AI Enabled Autonomous Building Automation to Enhance Security 279
12.5.1 AI Enabled Threat Identification and Mitigation 280
12.5.1.1 Theoretical Concept: AI Based Identification System 280
12.5.1.2 AI Based Security Learning System: Theoretical Concept 280
12.5.2 AI Enabled Cybersecurity Protection 282
12.5.2.1 The Role of AI in Cybersecurity Protection: Theoretical Concept 282
12.5.3 AI Enabled Cyber-Physical Intrusion Detection System 285
12.5.3.1 An Integrated AI Based IDPS: Theoretical Concept 286
12.5.4 AI Enabled Cyber Incident Response 288
12.5.4.1 An Autonomous AI Cybersecurity Response System: Theoretical Concept 290
12.5.5 AI Based Building Recovery System 290
12.5.6 AI Based Building Recovery System: Theoretical Concept 291
12.6 Use Cases 291
12.6.1 AI to Mitigate Insider Threat: Cognitive Ubiquitous Sensing and Insider Threat 291
12.6.2 AI Enabled Smart Buildings Cybersecurity and Business Optimization 292
12.6.3 Uber for Cyber and Energy 293
12.6.4 Blockchain for Power Grid Resilience: Exchanging Distributed Energy at Speed, Scale, Autonomy and Security 295
12.6.5 Social Engineering Autonomy for Cyber Intrusion Monitoring and Real-Time Anomaly Detecting (SCI-RAD) 298
12.7 Conclusion and Future Research 298
References 300
Chapter 13: Evaluations: Autonomy and Artificial Intelligence: A Threat or Savior? 303
13.1 Introduction 303
13.1.1 Mathematical Model of Autonomy: Entropy of Teamwork 305
13.1.2 Entropy Production 309
13.1.3 Emotion 310
13.1.4 Evaluations 311
13.2 Introduction. Safety and Human Error 313
13.2.1 Human Error 314
13.2.2 The Role of AI in Reducing Human Error 315
13.2.3 Roles with AI 316
13.2.4 Forecasts with AI and Interdependence 317
13.2.5 Evaluations 318
References 320