Human-Machine Shared Contexts

Front Matter
Copyright
Contributors
Preface
	Questions for speakers and attendees at our symposium on AI and readers of this book
	Fault modes
	Detection
	Isolation
	Resilience and repair
	Consequences of cyber vulnerabilities
Introduction: Artificial intelligence (AI), autonomous machines, and constructing context: User interventions, ...
	Introduction
		Overview
		Problem 1: The generation of new data
		Problem 2: Explanations by machines of what they have learned or can share (building context)
		Problem 3: The mutual determination of context
		Possible solutions
			Potential solution #1. An AI apprentice update
			Potential solution #2. Digital twins
			Potential solution #3. Metrics
			Potential solution #4. Calibration
			Potential solution #5
		Summary
	Introduction of the chapters from contributors
	References
	Further reading
Analogy and metareasoning: Cognitive strategies for robot learning
	Background, motivations, and goals
	Using social learning and analogical reasoning in cognitive robotics
		Related work
		Approach
			Demonstration phase
			Assistance phase
			Mapping inference phase
		First experiment
			Discussion
		Summary
	Using reinforcement learning and metareasoning in cognitive robotics
		Related work
		Approach
			Architecture and concepts
			Object layer
			Deliberative layer
			Metareasoning: The processes
		Algorithm
		Second experiment
			Gold hunting in Minecraft
			Cup grasping in Gazebo
		Summary
	Conclusions
	Acknowledgments
	References
	Further reading
Adding command knowledge ``At the Human Edge´´
	Introduction
	Characteristics of the three systems
		Information collection systems
		Command and control systems
		The role of machine learning
		Information decision systems
	Strategies
	An example, agile C2 scenario
		The general situation
		The specific problem
		The mission
		The contingencies
		Mission sequence
		Aftermath
	Background of the approach
	Type considerations
	References
Context: Separating the forest and the trees-Wavelet contextual conditioning for AI
	Introduction
	Artificial intelligence, context, data, and decision making
	Wavelets and preprocessing
		Wavelet decomposition
		Illustration: Visualizing decomposed wavelet data
	A preferential transformation for initial resolution-scale
		A heuristic to suggest a decomposition-level
	Evaluating the preferred decomposition-level selection technique
		A management decision problem application
		Applying DWT to the management decision problem application
		Running the simulation
	Results and discussion
	Conclusion
	References
A narrative modeling platform: Representing the comprehension of novelty in open-world systems
	Introduction
	New system-level representations
	Taxonomy
		Display zones
		Discrete objects
		Discrete objects
		Operators
	2D versus 3D
	Examples
		Example 1: Analyzing propaganda-Red Riding Hood as a Dictator Would Tell It
		Parallel processes
		Nested meaning
		Grouping
		New unity capabilities
		Conflicting inferences
		Unity versus Keynote
		Example 2: The influence of personal history in a patient with PTSD
			Entering and recording text
			Representing nested situations
	Challenges
	Higher-level structures
		Analogy
		Governance
		Scale
		Unexpected information
		The visualization of unexpected systems
	Surrounding research and foundations
	Conclusion
	Acknowledgments
	References
Deciding Machines: Moral-Scene Assessment for Intelligent Systems*
	Introduction
	Background
		Spooner\'s grudge
		Decomposing ``three-laws safe´´
		Pathway to moral action
	Moral salience
		Mind perception
		Interpersonal harm
		Human biases in moral processing: Cognitive and perceptual
		Minds and their vulnerabilities
		Affordance/danger qualification of objects
		Moral salience estimation
			Perceiving moral salience
			Material
			Structural/functional and behavioral
			Computer vision for moral salience
		Moral salience integration
	Mode of interaction
		Stance adoption
		The phenomenal stance
		Human biases in moral processing
			Stance confusion: Android minds
			Fallacies in stance adoption
			Intelligent systems push our Darwinian buttons
	Reasoning over insults and injuries
		Harm and damage
		Harm ontology
		Danger-vulnerability matrix
		Insults and injuries
		Learning to detect suffering and avoid harming
	Synthesis: Moral-Scene Assessment
	Application
		Values-driven behavior generation for intelligent systems
		Reasons-responsiveness in physical therapy
		Implementing nonmaleficence
			Entity labeling
			Scenario: Deposit tool in toolbox
			Scenario: Hand tool to human
	Roadmap
	Acknowledgments
	References
	Further reading
The criticality of social and behavioral science in the development and execution of autonomous systems
	Introduction
	Autonomous systems: A brief history
	Limitations of cognition and implications for learning systems
	Considering physical, natural, and social system interdependencies in autonomous system development
	Ethical concerns at the intersection of social and autonomous systems
	Conclusion
	Acknowledgments
	References
Virtual health and artificial intelligence: Using technology to improve healthcare delivery
	Introduction
	The end-to-end healthcare experience
	Digital health solutions
	Architecture of an end-to-end digital health solution
		Information
		Consultation
		Follow-up
	The role of AI in virtual health
	HealthTap: AI methods within a virtual health platform
	Limitations and future directions
	References
An information geometric look at the valuing of information
	Introduction
	Information geometry background
	A brief look at Riemannian geometry in general
	Fisher information and Riemannian geometry
	A simple Fisher space-normal distribution: Two parameters
	The statistical manifold NfΣ
		Hf Revisited
		Nfμ,ΣD(σ11)
		Nfμ,ΣD(σ11,,σnn)
	Value of information and complexity
	Allotment of resources
		C[NfΣD(σ1,σ2)]
	Conclusion
	Acknowledgments
	References
AI, autonomous machines and human awareness: Towards shared machine-human contexts in medicine
	Introduction
	Current state of medical education and its challenges
		Biomedical information explosion
		Digital health technology expansion
		Technology insertion barriers
		Medical learning challenges
	Potential AI application for medical education
		Medical education use-cases
			Example use case-USMLE Step-2 clinical knowledge question
	Shared human-Machine contexts in medical education
		``Crunchy´´ technology questions
		Of men and machines-Context matters
		The AI hopes that float the health care boats
		Medical learning contexts
		Decision support
	References
Problems of autonomous agents following informal, open-textured rules*
	Informal, open-textured rules
	Obstacles of IORs
		Formal languages
		Asymmetry
		Indeterminacy
	Interpretive arguments
		Statutory interpretation: A working model
		Formality of law
		Types of interpretive arguments
	Conclusion: Ameliorating the problems of IORs
	References
Engineering for emergence in information fusion systems: A review of some challenges*
	Introduction
	Technical foundations
		Data fusion and IF systems
		Systems engineering
		Machine learning
	Widespread impacts of emergence
		Emergence and IF systems
		Machine learning and emergence
		Systems engineering and emergence
	Emergence challenges for future IF systems
		Decision models for goal-directed behavior
		Information extraction and valuation
		Decision assessment
		Operator/human issues
		Integration of emergence from social sciences
		Interdependence
		Emergence
		Machine learning in adversarial environments
	Conclusions and future work
	References
Integrating expert human decision-making in artificial intelligence applications
	Introduction
	Background
	Decision-making background
	Problem domain
	Approach
	Technical discussion of AHP
		SME matrix (comparison ratios)
	Some matrix definitions
		SMEs give actual values
	Exponential additive weighting
	Procedure
	An example with R code
	Conclusion
	Acknowledgments
	References
A communication paradigm for human-robot interaction during robot failure scenarios
	Introduction
	Related work
	Interaction design
	Experiment methodology
		Communication protocol
		Design
		Hypotheses
		Independent variables
		Dependent variables
		Participants
		Analysis methods
	Results
		Robot usage
		Game-playing zone
		Balloons game
		Robot interactions
		Post-run questionnaires
		Post-experiment questionnaires
		Experiment scores
	Discussion
		H1 is supported by the results
		H2 is partially supported by the results
		H3 is supported by the results
		Effects of run ordering
	Future work
	Conclusions
	Acknowledgments
	References
On neural-network training algorithms
	Introduction
	The one-dimensional case
		Convergence and rate of convergence
		Constrained minimization
	The n-dimensional case
		Convergence and rate of convergence
		Solution of equations versus minimization
		Steepest descent and conjugate gradients
		Constrained minimization
	Implications for neural-network training
	Summary
	References
Identifying distributed incompetence in an organization
	Introduction
	Defining DI
	Observing organizations with DI
		DI and management
		DI and whistleblowing
		Financial crisis of 2007 and DI
		DI and competitive rating
		DI and other forms of irrationality
		Aggressive DI
		DI and unexplainable machine learning
	Detecting DI in text
		An example of discourse-level DI analysis
		A DI detection dataset
		Discourse-level features of DI
		Implementation of the discourse-level classifier
		Detection results
	Conclusions: Handling and repairing DI
	References
	Further reading
Begin with the human: Designing for safety and trustworthiness in cyber-physical systems
	Introduction
	The Three Mile Island accident
	The analytical framework
		Autonomy
		Agency
		Assurance
		Metrics
		Interfaces
	Discussion and conclusions
	Acknowledgments
	References
Digital humanities and the digital economy
	Motivation
	What is digital humanities?
		Artificial readers
		Cyberspace of books
		An ontological augmentation
	The twin space
		``Cloudy´´ digital humanities
		The twin space and consilience
	The digital economy
	Reinventing individuality
		Individuality and the bottom-up history
		Identity and blockchain
	Matching
		Personal web of everything
		Great transformation and smartness
		Feedback loops
	Concluding remarks
	Acknowledgments
	References
Human-machine sense making in context-based computational decision
	Introduction
	Basic features of decision based mechanisms
		Computational contexts
		The problem solving loop sequence
		The causal break and Hopf algebras
		Double S-curves and disruption
	Human-machine agents and characteristics
		Human-machine interactions
		Sense making and computational contexts
		Decision and failures of deductive systems
		A plausible integrating model: The continuous inference
		Perspectives
	Conclusion
	References
	Further reading
Constructing mutual context in human-robot collaborative problem solving with multimodal input
	Introduction
	UIMA
	Information processing architecture
	Object detection
	Spatial relation processor
	Speech processing
	Natural language processing
		Gesture
		Naming a block
		Modifiers
		Commands
		Colors
	Gesture recognition
	Confidence aggregation
		Spatial relationships
	Communication unit
	Memory
	Constructing shared context
		Shared context in annotators
		Context accumulated in the CAS object
	Conclusions
	Acknowledgments
	References
Index