Advances in Human-AI Collaboration
Description
Detailed guide on how humans and AI systems work in tandem, focused on the successful deployment and use of applications
Advances in Human-AI Collaboration offers a comprehensive exploration of AI technologies and applications in the field of Industrial and Systems Engineering. The book incorporates knowledge about AI technology, methodologies, and tools for enabling human-AI collaboration at work, covers the effective design and use of systems and operations utilizing human-AI collaboration to benefit productivity, quality, and customer satisfaction, and provides readers with the skills necessary to effectively implement and consider human-AI collaboration across a variety of settings.
This book delivers insights on a wide range of topics including similarities and differences of human and artificial intelligence, effort in creating algorithms versus meeting user needs and enabling improved decision support, sentiment analysis and language models, AI tutors and their design, engagement, and theory building, situation awareness of AI models in relation to human performance, and fact-checking beyond machine learning and predictive accuracy.
Written by a team of highly qualified academics with significant experience in the field, Advances in Human-AI Collaboration includes information on:
- Autonomous vehicles and delivery systems, covering sensors and perception as well as adoption rate and safety projections
- Chat-based customer service, covering theory-based interventions to enhance public services and examples of intentional human-technology interaction
- Ship safety, covering increased automation and machine vision to enable collision avoidance
- Strategies for moving beyond passive writing assistance and writing-related best practices
- Vulnerabilities in technology-centered design including biased and distorted data, with examples of real-world accidents
Advances in Human-AI Collaboration is an essential read for industry practitioners and corporate researchers concerned with using AI in integrated system design and operation. The book also provides essential knowledge for academics researching AI and integrated systems.
More details
Other editions
Additional editions
Persons
VINCENT G. DUFFY is a Professor of Industrial Engineering and Agricultural & Biological Engineering at Purdue University.
WALDEMAR KARWOWSKI is a Pegasus Professor and Chairman in the Department of Industrial Engineering and Management Systems at the University of Central Florida.
GAVRIEL SALVENDY is a University Distinguished Professor at the University of Central Florida.
Content
List of Contributors xvii
Preface xxi
Section I Fundamentals in Human-AI Collaboration 1
1 Human Interaction with Intelligent Automation: A Continuum of AI Levels from Which Designers and Users Can Choose 3
Thomas B. Sheridan and William B. Rouse
1.1 Introduction 3
1.2 Background 3
1.2.1 Traditional Automation 3
1.2.2 Human Interaction with Traditional Automation 4
1.2.3 Artificial Intelligence 6
1.2.3.1 Evolution of AI 6
1.2.3.2 Spectrum of AI 7
1.3 Human Interaction with Intelligent Automation 7
1.3.1 AI for Robotic Automation 8
1.3.2 Augmentation and Adaptation 8
1.3.3 Levels of AI Assist in Automation 9
1.3.4 Policy Issues Regarding AI and Automation 10
1.4 Three Scenarios 10
1.4.1 Driverless Cars 10
1.4.2 Management of Autonomous Airplanes 11
1.4.3 Healthcare Decision Support 11
1.4.4 Comparison of Scenarios 13
1.5 Discussion 13
1.5.1 Continuum of Levels 13
1.5.2 Transition Management 14
1.5.3 Human-Centered Intelligent Automation 15
1.6 Conclusions 15
References 16
2 Human-AI Interaction Fundamentals 17
George Margetis, Stavroula Ntoa, Asterios Leonidis, and Constantine Stephanidis
2.1 Introduction 17
2.2 Human-Centered AI 18
2.2.1 Explainability and Understandability 18
2.2.2 Equity and Fairness 19
2.2.3 Human-AI Collaboration 20
2.2.4 Ethical Considerations 20
2.3 Interaction Styles 22
2.3.1 Traditional Interaction Styles in AI 22
2.3.2 Natural Language Interaction 23
2.3.3 Gesture-Based Interaction 23
2.3.4 Multimodal Interaction Models and Future Trends 24
2.4 Interaction Contexts 25
2.4.1 Affective Computing and Interaction with AI 25
2.4.2 Decision-Making and Recommender Systems 25
2.4.3 Generative AI and Large Language Models 26
2.4.4 Human-Robot Interaction 26
2.4.5 Application Domains 27
2.5 HCI Aspects in Human-AI Interaction 28
2.5.1 Usability and User Experience 28
2.5.2 Cognitive and Social Factors 29
2.5.3 Errors and User Trust 29
2.5.4 Challenges and Limitations 30
2.6 Summary and Conclusions 31
References 31
3 Human-AI Interaction Fundamentals 41
Woei-Chyi Chang, Sogand Hasanzadeh, and Vincent G. Duffy
3.1 Introduction 41
3.2 Entities Involved in Human-AI Interaction 41
3.2.1 Humans 41
3.2.1.1 Demographic and Dispositional Factors 42
3.2.1.2 Situational and Learned Factors 42
3.2.2 Ai 43
3.2.3 Environments 43
3.3 Human-AI Interaction Levels 44
3.3.1 Coexistence 44
3.3.2 Cooperation 44
3.3.3 Collaboration 45
3.4 Critical Elements of Human-AI Interaction 45
3.4.1 Trust 45
3.4.1.1 Trust Spectrum 46
3.4.1.2 Trust Calibration 46
3.4.1.3 Trust Measurements 47
3.4.2 Communication 48
3.4.3 Privacy and Data Security 49
3.4.4 Personalization and Adaptability 50
3.5 Conclusions 51
References 51
4 Guidelines for Human-AI Interaction and User Experience 57
Helmut Degen
4.1 Introduction 57
4.2 Social Needs 59
4.2.1 Laws and Regulations from Selected Regions and Countries 60
4.2.2 EU AI Act 60
4.2.3 Chinese Artificial Intelligence Laws 62
4.2.4 US AI Executive Order 14110 62
4.2.5 Common Social Needs 63
4.3 Guidelines to Address Selected Common Social Needs 65
4.3.1 Ethical Use 65
4.3.1.1 Social Need Details 65
4.3.1.2 Guidelines 66
4.3.2 Product Quality Over Product Lifecycle 66
4.3.2.1 Social Need Details 66
4.3.2.2 Guidelines 67
4.3.3 Human Oversight 69
4.3.3.1 Social Need Details 69
4.3.4 Effective, Ethical, Robustness, Accuracy, Safety, and Security 70
4.3.4.1 Social Need Details 70
4.3.4.2 Guidelines 71
4.3.5 Explainability and Trustworthiness 71
4.3.5.1 Social Need Details 71
4.3.5.2 Guidelines 72
4.3.6 Diversity, Fairness, Impartiality, Privacy, Health 77
4.3.6.1 Social Need Details 77
4.3.6.2 Guidelines 80
4.4 Discussion and Future Outlook 81
4.4.1 Discussion 81
4.4.2 Outlook 82
4.4.3 Acknowledgments 82
4.4.4 Funding 83
4.4.5 Conflict of Interest 83
References 83
5 Human Intelligence vs. Artificial Intelligence 91
Peiran Liu and Denny Yu
5.1 Characterizing Similarities and Differences Today 91
5.1.1 Historical Perspectives of Human Intelligence 91
5.1.2 Measurement of Human Intelligence 93
5.1.3 Background on Artificial Intelligence 94
5.1.4 Similarity and Differences-A Brief Discussion 95
5.2 Limitations in Human and Artificial Intelligence Today 96
5.2.1 Perception: Sensory Systems and Sensors 96
5.2.2 Cognition 98
5.3 Trustworthiness 100
References 102
6 Advances in Human-AI Collaboration: Training with AI Support 107
Yubin Xie, Siu Shing Man, Ronggang Zhou, and Alan Hoi Shou Chan
6.1 Introduction 107
6.1.1 What is an AI Agent? 108
6.1.1.1 Concept of AI Agent 108
6.1.1.2 Type of AI Agent 108
6.1.1.3 Characteristics of AI Agent 108
6.1.2 Training with AI Support 109
6.1.2.1 Background 109
6.1.2.2 Definition of AI Training 109
6.1.3 Advantages of Training with AI Support 110
6.1.3.1 Efficiency 110
6.1.3.2 Personalized 111
6.1.3.3 Cost Saving 111
6.1.3.4 Continuous Learning 111
6.1.4 Challenges and Ethical Considerations in Training with AI Support 112
6.1.4.1 Transparency and Interpretability 112
6.1.4.2 Fairness and Bias 112
6.1.4.3 Other Challenges 112
6.2 Application of AI in Training 113
6.2.1 Knowledge Base Construction 113
6.2.1.1 Concept of Knowledge Base 113
6.2.1.2 Foundation of Knowledge Base 114
6.2.2 Training Needs Analysis 115
6.2.3 Feedback on Training Results 116
6.2.4 Algorithm Aversion and User Experience 116
6.3 Case Studies of Training with AI Support 117
6.3.1 Case Study 1 117
6.3.2 Case Study 2 118
6.4 Summary 118
Acknowledgments 119
References 119
7 Human-AI Teaming 123
Tianyi Yuan, Minqian Yang, Dian Yu, and Pei-Luen Patrick Rau
7.1 Theoretical Foundations 123
7.1.1 Definitions 123
7.1.2 History and Trends 124
7.1.3 Theoretical Frameworks for Human-AI Teaming 124
7.2 Decision-Making Process in Human-AI Teaming 126
7.2.1 Division of Labor 126
7.2.2 Attribution 127
7.3 Communication in Human-AI Teaming 128
7.3.1 Factors Influencing Communication in Human-AI Teaming 128
7.3.2 Benefits and Challenges 129
7.3.3 Communication Strategies in Human-AI Teaming 129
7.4 Trust 130
7.4.1 Factors Influencing Trust in Human-AI Teaming 130
7.4.2 The Role of Trust in Human-AI Teaming 131
7.4.3 Strategies to Enhance Trust in Human-AI Teaming 131
7.5 Explainability and Explainable AI in Human-AI Teaming 132
7.5.1 Human-AI Teaming and the Role of Explainability 132
7.5.2 Challenges and Trends in Explainable AI for Human-AI Teaming 132
7.6 Ethical and Social Implications 133
7.6.1 Ethical Dimensions of Human-AI Teaming 133
7.6.2 Social Implications of Human-AI Teaming 134
References 135
8 Human Teaming with Automation and Advanced Agents 143
Barrett S. Caldwell, Rua M. Williams, and C. Nuela Enebechi
8.1 Introduction 143
8.2 Clarifying Automation and Autonomy 144
8.3 Teaming Metaphors and Models 146
8.4 Dynamics of Expertise and Function Allocation 148
8.5 Challenges in Ascribing Trust Dynamics 151
8.6 Impacts of Culture and Bias 152
8.7 Conclusion: Design for Robust and Resilient Human-Agent Teams 154
References 155
9 Integrating Generative Design and Ergonomics: A Data-Driven Approach with Digital Manikins 159
H. Onan Demirel, Xingang Li, and Zhenghui Sha
9.1 Introduction 159
9.2 Literature Review 161
9.2.1 Data-Driven Generative Design Process 161
9.2.2 Deep Learning Methods for Cross-Modal Tasks 162
9.2.3 Computational Human Factors via Digital Human Modeling 162
9.3 Methodology 163
9.3.1 Human-Centered AI-Assisted Concept Generation 164
9.3.2 Concept Evaluation Using Digital Human Modeling 164
9.3.3 Iterative Concept Refinement and Concept Modification 164
9.4 Result 165
9.4.1 Human-Centered AI-Assisted Concept Generation 165
9.4.2 Human-Centered AI-Assisted Concept Evaluation 166
9.5 Discussions 167
9.6 Conclusions 169
References 169
Section II Application of Human-AI Collaboration 173
10 AI-Enabled Accessible Travel in Autonomous Vehicles: Promises, Perceptions, and Prototypes 175
Brandon J. Pitts, Qiyue Wang, and Bradley S. Duerstock
10.1 Introduction 175
10.2 Travel-Limiting Disabilities and Challenges with Current Transportation Systems 176
10.2.1 Travelers with Disabilities 176
10.2.2 Aging populations 177
10.3 Development of Autonomous and Shared Vehicles 177
10.4 Perceptions of AVs and SAVs 179
10.4.1 Travelers with Disabilities 179
10.4.1.1 Mobility Impairments 179
10.4.1.2 Visual Impairments 180
10.4.1.3 General (Other Types of Disabilities) 181
10.4.2 Aging Populations 181
10.5 Interactions with AVs and SAVs 182
10.5.1 Travelers with Disabilities 183
10.5.2 Aging Populations 184
10.6 Technologies and AI Solutions to Overcome Barriers to Using AVs 185
10.7 Case Study: The EASI RIDER Innovation 188
10.8 Future Research and Development Needs 193
10.9 Conclusion 193
Acknowledgments 194
References 194
11 Cobot: Collaborative Robots 201
li liu
11.1 The Origins and Development of Collaborative Robots 201
11.1.1 Overview of Early Industrial Robots 201
11.1.2 From Industrial Automation to Cobots 202
11.1.3 The Birth and Evolution of Cobots 203
11.2 Technical Foundations of Cobots 204
11.2.1 What Is a Collaborative Robot? 204
11.2.2 The Role of AI in Cobots 204
11.2.3 Multimodal Perception Systems 205
11.2.4 Human-Robot Interaction and Safety Design 206
11.3 Application Scenarios of Cobots 207
11.3.1 Cobot Applications in Manufacturing 207
11.3.2 Cobots in Logistics and Warehousing 207
11.3.3 Innovative Applications in Healthcare and Service Industries 209
11.3.4 Cobots in Agriculture and Environmental Management 210
11.4 The Future of Collaborative Robots 211
11.4.1 The Integration of Humanoid Robots and Cobots 211
11.4.2 Ethical, Legal, and Safety Challenges 212
11.4.3 The Integration of Cobots with Industry 4.0 213
11.5 Summary 214
References 214
12 AI Chat-Based Customer Services and Systems 217
Qin Gao and Jinhan Zhang
12.1 Introduction 217
12.2 Development of AI Chatbots in Customer Services 218
12.3 Customer Service Affordances of AI Chatbots 219
12.4 Factors Affecting Consumer Experience with AI Chatbots 225
12.4.1 Functional and Utilitarian Features 225
12.4.2 Interaction Experience Features 225
12.4.3 System Quality Features 227
12.4.4 Individual Differences 228
12.5 Impacts of AI Chatbots on Customer Service Experience 229
12.6 AI Chatbot in Public Services 230
12.7 Challenges of Using AI Chatbot 232
12.7.1 Technical and Functional Challenges 232
12.7.2 Data Security, Privacy, and Ethical Challenges 232
12.7.3 Organizational and Implementation Challenges 232
12.8 Conclusions 233
References 233
13 AI in Healthcare and Medicine 239
Tianrong Chen, Zhenzhen Xie, and Calvin Or
13.1 Introduction 239
13.2 What Is AI? 240
13.2.1 Definition and Scope of AI 240
13.2.2 Advancements of AI 240
13.2.3 Role of AI in Healthcare and Medicine 241
13.3 AI in Healthcare 241
13.3.1 AI in Chronic Disease Management 241
13.3.2 AI in Home Rehabilitation 242
13.3.3 AI in Fall Prevention and Detection 243
13.3.4 AI in Mental Health Support 244
13.4 AI in Medicine 245
13.4.1 Diagnosis and Clinical Decision Support 245
13.4.2 Medical Imaging 246
13.4.3 Surgery 247
13.4.4 Predictive Analytics 248
13.4.5 Clinical Workflow Management 248
13.5 Considerations and Challenges 249
13.5.1 Data Privacy and Security 249
13.5.2 Data Fragmentation and Limited Availability 250
13.5.3 Transparency and Interpretability of AI 250
13.5.4 Human-AI Interaction 250
13.5.5 Regulatory Frameworks and Guidelines 251
13.6 Future Trends and Opportunities 252
13.7 Conclusions 252
References 253
14 AI in Human Resource Management 263
Xinyu Fu, Fiona Fui-Hoon Nah, Songbo Liu, Kairui Zhang, Zixin Huang, Ruilin Zheng, and Weiqi Xie
14.1 AI-Driven Human Resource Processes 264
14.1.1 Human Resource Planning 264
14.1.2 Recruitment and Selection 265
14.1.3 Learning and Development 266
14.1.4 Performance Management 268
14.1.5 Compensation and Benefits 269
14.1.6 Employee Relationship Management 270
14.1.7 Summary 271
14.2 How AI Reshapes Team Dynamics 272
14.2.1 AI Reshapes the Teaming Process 272
14.2.2 Impact of AI on Team Performance 273
14.2.3 AI and Leadership in the Team 273
14.2.4 Summary 274
14.3 Ethical Considerations and Challenges 274
14.3.1 Employees' Attitudes Toward AI 274
14.3.2 Employees' Adaptation to AI 275
14.3.3 Human-Machine Relationship in HRM 275
References 276
15 Kansei Engineering: Current Challenges and Future Trends with Advances in Artificial Intelligence 287
Qing-Xing Qu, Fu Guo, and Vincent G. Duffy
15.1 Introduction 287
15.2 Advances in Intelligence Product Development 289
15.3 Kansei Design in Different Interaction Modalities 292
15.3.1 Kansei Design in Unimodal Interaction 292
15.3.2 Kansei Design in Bimodal Interaction 293
15.3.3 Kansei Design in Multimodal Interaction 293
15.4 Challenges Encountered and Opportunities for Future Research 295
15.5 Conclusions 296
References 297
16 AI in Collaborative Writing 303
Nina Jiang and Vincent G. Duffy
16.1 Introduction 303
16.2 Literature Review 304
16.2.1 Data Collection 304
16.2.2 Trend Analysis 304
16.2.3 Co-citation Analysis 304
16.2.4 Timeline Analysis 306
16.2.5 Word Cloud 308
16.2.6 Geography Network 309
16.3 Fundamentals in Machine Learning and Text Generation 309
16.3.1 Language Model 309
16.3.2 Natural Language Generation 310
16.4 User Perception of the Machine Role 311
16.5 Strategies for Moving Beyond Passive Writing Assistance 313
16.6 Collaborative Writing Best Practice Examples 314
16.7 Discussion 315
16.7.1 Evolution of AI Tools for Collaborative Writing 315
16.7.2 Impact of AI on Collaborative Writing 316
16.7.3 Challenges and Limitations 316
16.7.4 Future Directions 317
16.8 Conclusion 318
References 318
17 Addressing AI Vulnerabilities Through Human-Centered Approaches and Risk Frameworks 325
Sarvesh Sawant, Aasish Bhanu, Beau G. Schelble, and Kapil Chalil Madathil
17.1 Introduction 325
17.1.1 AI Vulnerabilities 325
17.1.2 Frameworks for Identifying and Mitigating AI Vulnerabilities 327
17.2 Types of AI Vulnerabilities 328
17.2.1 Technical Vulnerabilities 329
17.2.2 Human-Centered AI Vulnerabilities 330
17.3 Mitigating AI Vulnerabilities Through Human-Centered Design 331
17.3.1 Human-Centered Design Principles 331
17.3.2 Transparency and Explainability in AI 332
17.3.3 Human-in-the-Loop Design 333
17.3.4 User Education and Training 333
17.3.5 Ethical Considerations 334
17.4 Summary 335
References 335
Index 341
1
Human Interaction with Intelligent Automation
A Continuum of AI Levels from Which Designers and Users Can Choose
Thomas B. Sheridan1 and William B. Rouse2
1Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA, USA
2McCourt School of Public Policy Georgetown University, Washington, DC, USA
1.1 Introduction
Intelligent automation will make use of artificial intelligence (AI) technology to provide far greater capability than traditional automation. It will pervade many aspects of our lives-in our factories, our businesses, our homes, and our hospitals. It will allow for new flexibility of interaction with human users, and in many cases, it will make decisions on its own, based on its knowledge of what is safe and what its users want.
To provide contrast, this chapter briefly reviews traditional automation and highlights aspects of human interaction from the human factors literature, including salient taxonomies and human-computer simulation experiments that reveal the limitations of both the automation and the humans. Factors related to task requirements, safety, and time and cost requirements are discussed. The chapter then discusses how advances in AI and robotics are changing automation and providing augmentation and adaptation in information management, intent inferencing, error-tolerant interfaces, adaptive aiding, and intelligent tutoring. We propose a novel scale of levels of AI sophistication that, when combined with the traditional levels of automation scale, provides new ways for designers to consider alternatives. Several types of related policy issues are then considered.
Finally, as examples of human interaction with intelligent automation, we discuss three scenarios: driverless cars, management of autonomous aircraft, and healthcare decision support. Considerations of how the automation supports presentation, planning, and delivery are specified in each example for four temporal stages of use. Management of the evolution toward these developments is discussed as well.
1.2 Background
1.2.1 Traditional Automation
Definition of automation. Automation is the creation and application of technologies to produce and deliver goods and services to improve the efficiency, reliability, and/or speed of many tasks that were previously performed by humans. (from Technopedia Dictionary). Typically, automation involves a feedback loop consisting of sensing, memory, computation, and control to perform a given task according to a specified criterion or objective function.
1.2.2 Human Interaction with Traditional Automation
In 1951, Psychologist Paul Fitts proposed what came to be known as the Fitts List:
Men are better at: (1) detecting small amounts of visual, auditory, or chemical energy; (2) perceiving patterns of light or sound; (3) improving and using flexible procedures; (4) storing information for long periods of time and recalling appropriate parts; (5) reasoning inductively; and (6) exercising judgment. Correspondingly, machines are better at (1) responding quickly to control signals; (2) applying force, smoothly and precisely; (3) storing information briefly; (4) erasing it completely; and (5) reasoning deductively.
The Fitts list elicited much discussion, and as technology of sensing, memory, computation, and control improved, it became clear that machines could perform many of these functions better than people.
In a study on underwater robotics and their relation to human operators Sheridan et al., (1978) proposed a 10-level scale shown in Table 1.1 for graded levels of the computer subsuming human function in any given task. This scale was intended as a suggestion, since clearly one could devise different definitions, combine levels, devise graphical representations using flow charts or logic trees, etc.
Certainly, the number of levels can vary, depending on how they are defined. Sheridan later published a simplified 8-level scale (2002). Others have suggested multidimensional scales. There is no basis for a unique levels-of-automation scale.
The original 1978 level of automation publication motivated considerable discussion in the human-systems literature and several alternative order scales. It also resulted in one author asserting that it prematurely constituted an "algorithm" for design of human-machine systems. Kaber (2018) identified existing issues with the levels-of-automation approach and suggested future research directions. Endsley and Kaber (1999) proposed a levels-of-automation taxonomy that included monitoring displays, generating processing options, selecting an "optimal" option and implementing that option.
Table 1.1 Sheridan-Verplank Levels of Automation.
- 1) The human operator does the task and turns it over to the computer to implement.
- 2) The computer helps by determining the options.
- 3) The computer helps determine and suggests options. The human operator can choose to follow the recommendation.
- 4) The computer selects the action, and the human operator decides if it should or should not be done.
- 5) The computer selects the action and implements it if the human operator approves the action.
- 6) The computer selects the action and informs the human operator in case the operator wants to cancel the action.
- 7) The computer does the action and tells the human operator what it did.
- 8) The computer does the action and tells the human only if the human operator asks.
- 9) The computer does the action when told and tells the human operator only if the computer decides the operator should be told.
- 10) The computer does the action if it decides it should be done. The computer tells the human operator only if it decides the operator should be told.
Thirty subjects performed simulation trials involving various levels of automation of a dynamic task. Several automation failures occurred, and out-of-the-loop performance decrements were assessed. Results suggested that human operators benefited most from automation of the implementation portion of the task, but only under normal operating conditions; in contrast, removal of the operator from task implementation is detrimental to performance recovery if the automated system fails. Joint human/system option generation significantly degraded performance in comparison to human or automated option generation alone.
Parasuraman et al. (2000) later introduced the idea of associating levels of automation to four functions, as follows: (1) Sensory processing, referring to the acquisition and registration of multiple sources of information. This stage includes the positioning and orienting of sensory receptors, sensory processing, initial preprocessing of data prior to full perception, and selective attention. (2) Perception and/or working memory, including conscious perception and manipulation of processed and retrieved information in working memory. This includes cognitive operations such as rehearsal, integration, and inference, but these operations occur prior to the point of decision. (3) Decision-making, based on such cognitive processing. (4) Response selection, which involves the implementation of a response or action consistent with the decision choice. These functions are based on a four-stage model of human information processing and can be translated into equivalent system functions: (1) Information acquisition; (2) Information analysis; (3) Decision and action selection; (4) Action implementation.
Wickens et al. (2010) used 14 experiments to examine how performance, workload, and situation awareness are affected by the level of automation when the automation fails. They found: (1) an inverse relationship between routine performance and workload as automation is introduced and the level increases. (2) a weak positive relationship between routine performance and failure performance. (3) A strong mediating role of situation awareness in improving both routine and failure performance.
In conjunction with the development of self-driving vehicles, the Society of Automotive Engineers developed a 6-level automation scale for driving (SAE, 2021):
Note that the human monitors the driving environment in the first three levels of Table 1.2, while the automated system monitors the driving environment in the last three. A major problem with implementation at levels 3 and 4 is that human drivers, if not "in the loop" and constantly attending to the road, will not be sufficiently alert if their attention is suddenly needed (Inagaki & Sheridan, 2012).
Table 1.2 The SAE Levels of Automation Scale.
- 1) No automation. Manual control. The human performs all driving tasks (steering, acceleration, braking).
- 2) Driver Assistance. The vehicle features a single automated system (e.g., it monitors speed through cruise control).
- 3) Partial Automation. The vehicle can perform steering and acceleration. The human still monitors all tasks and can take control at any time.
- 4) Conditional automation. Environmental detection capabilities. The vehicle can perform most driving tasks, but human override is still required.
- 5) High Automation. The vehicle performs all driving tasks under specific circumstances....
