Transparency for Robots and Autonomous Systems: Fundamentals, technologies and applications (Control, Robotics and Sensors)

Cover
Contents
Related Publications
List of figures
List of tables
About theAuthor
Preface
Acknowledgements
Transparency for robots and autonomous systems
1 Introduction
	1.1 Motivation
	1.2 Motivating concerns
		1.2.1 Human cognitive bias
		1.2.2 Robots in Western culture
		1.2.3 Moral confusion
		1.2.4 Societal concerns
		1.2.5 Real robots
		1.2.6 Public policy for AI and autonomous systems
	1.3 Robots and transparency
	1.4 Thesis
	1.5 Structure of this book
		1.5.1 Chapter 2: Transparency in the wider context of trust
		1.5.2 Chapter 3: A transparent robot control architecture
		1.5.3 Chapter 4: The impact of transparency using real-time displays
		1.5.4 Chapter 5: Transparency using audio – the muttering robot
		1.5.5 Chapter 6: The effects of appearance on transparency
		1.5.6 Chapter 7: Synthesis and further work
		1.5.7 Chapter 8: Conclusions
	1.6 Contributions
2 Transparency in the wider context of trust
	2.1 Summary
	2.2 The meanings of trust
		2.2.1 Informed trust
		2.2.2 Uninformed trust
		2.2.3 Naive trust
		2.2.4 So-called trustless systems
		2.2.5 Trust as behaviour
		2.2.6 Distrust and mistrust defined
	2.3 A model for trust
	2.4 Accountability
	2.5 Responsibility
	2.6 The role of transparency in informed trust
		2.6.1 Organisational transparency
		2.6.2 System transparency
	2.7 Transparency as a driver for trust
3 A transparent robot control architecture
	3.1 Summary
	3.2 Introduction
		3.2.1 From ethology to robots
	3.3 The R5 robot
	3.4 The Instinct Planner
		3.4.1 Enhancements and innovations
		3.4.2 Multiplatform
		3.4.3 Memory management
		3.4.4 Instinct testing environment
		3.4.5 Instinct transparency enhancements
		3.4.6 Instinct command set
		3.4.7 Creating reactive plans with iVDL
		3.4.8 Plan debugging and transparency
	3.5 Modelling robot control using the Instinct Planner
		3.5.1 Kinds of minds
		3.5.2 The Instinct Robot World
		3.5.3 Conjectures
		3.5.4 Methods
		3.5.5 Outcomes
		3.5.6 Discussion
	3.6 Conclusions and further work
		3.6.1 Instinct and ROS
4 The impact of transparency using real-time displays
	4.1 Summary
	4.2 Mind models and transparency
	4.3 Technologies used: reactive planning and robot transparency
		4.3.1 The transparent planner
		4.3.2 Robot Drives and behaviours
		4.3.3 Real-time plan debugger
		4.3.4 Mobile Augmented Reality
	4.4 Methods: the robot experiments
		4.4.1 Experiment one – Online Video
			4.4.1.1 Robot videos
			4.4.1.2 Demographic and post-treatment questionnaires
		4.4.2 Experiment two – Directly Observed Robot
	4.5 Results
		4.5.1 Main findings
		4.5.2 Qualitative outcomes
	4.6 Experiment three – MobileAugmented Reality
	4.7 Discussion
	4.8 Conclusion and further work
5 Transparency using audio – the muttering robot
	5.1 Summary
	5.2 Transparency through spoken language
	5.3 The muttering mechanism
		5.3.1 Transparency execution stack
		5.3.2 The Speak Rules Engine
		5.3.3 The Vocaliser
	5.4 Experimental methods
		5.4.1 Post-treatment questionnaire
		5.4.2 Affect Questions
	5.5 Results
		5.5.1 Main findings
		5.5.2 Affect – self-report of Feelings
	5.6 Discussion
	5.7 Conclusions and further work
6 The effects of appearance on transparency
	6.1 Summary
	6.2 Introduction
	6.3 Appearance and anthropomorphism
	6.4 Experimental methods
		6.4.1 Amazon Mechanical Turk
		6.4.2 Buddy the Robot
		6.4.3 Robot videos
		6.4.4 Online questionnaire
		6.4.5 Impression of the robot – the 'Godspeed' questions
		6.4.6 Work batches in Mechanical Turk
	6.5 Results
		6.5.1 Participant demographics
		6.5.2 Main results
		6.5.3 Failure to identify 'Sleeping'
		6.5.4 Participant confidence
		6.5.5 Affect – self-report of feelings
		6.5.6 Godspeed results
	6.6 Discussion
	6.7 Conclusions and further work
7 Synthesis and further work
	7.1 Summary
	7.2 Tools, machines, robots and algorithms
	7.3 Robot and AI ethics
		7.3.1 Facts versus values
		7.3.2 The science of morality
		7.3.3 An empirical basis for morality
	7.4 The case for transparency as a fundamental design consideration for AI systems
	7.5 Transparent minds
	7.6 Hybrid architectures
	7.7 Humans expect intentional explanations
	7.8 Further work
		7.8.1 Anthropomorphism, transparency and traceability
		7.8.2 Explainable AI, policy and regulation
		7.8.3 Technologies
	7.9 Conclusion
8 Conclusions
	8.1 The effectiveness of transparency to improve mind models
	8.2 Transparency as a fundamental design consideration for AI systems
	8.3 Robot technical architecture
	8.4 Conclusion
Appendix A. Instinct Planner command set
Appendix B. R5 robot command set
Appendix C. Instinct plan for the R5 robot
Appendix D. ABOD3 displays
Appendix E. ABOD3-AR mobile augmented reality
Appendix F. Google Forms questionnaire
References
Index
Back Cover