Towards Trustworthy Artificial Intelligent Systems

Description

This present book provides valuable insights on the technical, societal and legal challenges posed by the use of artificial intelligent systems in a plethora of different applications, from embodied robotic systems to ML algorithms. Engaging with concerns about equity, privacy, surveillance and respect for human dignity, “Towards Trustworthy Artificial Intelligent Systems” highlights the fundamental factors on which stakeholders’ trust relies, identifying benchmarking, standardisation and certification as milestones grounding and consolidating that future trust. The multidisciplinary approach followed will make this book a valuable resource for all those involved in the production and deployment of AIs, as well as for academia and legal practitioners.  Preface Acknowledgments Contents Artificial Intelligence: A Concept Under-Construction, A Reality Under-Development 1 A Concept Under-Construction 2 Natural Cognition: What Does [Intelligence] Stand for? 2.1 Basic Assumptions 2.2 The Three Common Key Existential Factors 2.3 The Specificity of Human Cognition 2.4 Representing Life Dynamics: The Need for a Non-mechanistic Approach 3 The Task of Replicating Intelligence 3.1 First Steps and the Way Forward 3.2 The Natural, the Artificial and Their Hybrid Forms 3.3 When Algorithms Couple with Data: The Digital Umwelt 4 Conclusion References In Machines We Trust? 1 Introduction 2 The Biological Roots of Trust 3 Trusting: The Glue of Interpersonal Relationships 4 The Concept of [Trust]: Defining Semantic Boundaries and Scope of Reference 5 Trusting Products 5.1 General Framework 5.2 The Case of Artificial Intelligent Systems 5.3 The Role of Benchmarking, Standardisation and Certification 5.4 Certification 6 Conclusions References Informational Privacy and Trust in Autonomous Intelligent Systems 1 Introduction 2 GDPR and Its Aims 3 GDPR & AIS—Case Examples 3.1 Ex-Ante Obligations: TOMs and DPIAs 3.2 Automated Decision Making 3.3 Transatlantic Transfers of Personal Data 4 Conclusions Reference Ethical Risk Assessment for Social Robots: Case Studies in Smart Robot Toys 1 Introduction 2 Ethical Risk Assessment 3 Case Study with a Fictional Smart Toy: RoboTed 3.1 Physical Risks 3.2 Psychological Risks 3.3 Privacy and Security Risks 3.4 Environmental Risks 3.5 Discussion of ERA for RoboTed 4 Case Study with Existing Smart Toy: Purrble 4.1 Physical Risks 4.2 Psychological Risks 4.3 Privacy and Security Risks 4.4 Environmental Risks 4.5 Discussion of ERA for Purrble 5 Discussion and Conclusions References Practical and Open Source Best Practices for Ethical Machine Learning 1 Introduction 2 Ethical Machine Learning 3 Best Practices in Machine Learning 3.1 Definition and Scope 3.2 Technical and Organisational BP 4 Conclusion References Recruitment AI Has a Disability Problem: Anticipating and Mitigating Unfair Automated Hiring Decisions 1 Introduction 2 Disability Marginalisation and Employment 3 Recruitment AI 4 Exclusion by Design and Discriminatory Use 4.1 Biased Systems 4.2 Improper Implementation and Use 5 AI on the Market: The Risks of Discrimination 5.1 ATS and CRM Systems 5.2 CV/Resume Screeners 5.3 Conversational Agents 5.4 Pre-employment Assessments 5.5 AI Interviewing 6 Intervention Recommendations 6.1 Executive Level Stakeholders 6.2 Human Resources Stakeholders 6.3 Procurement Stakeholders 6.4 Information Technology Stakeholders References Developing and Evaluating Complex Interventions: The Case of Robotic Systems in Cognitive Rehabilitation Therapy 1 Introduction 2 Complex Interventions: Definition and Scope 3 Robotics in Cognitive Rehabilitation Therapy (CRT) 4 Conclusions References COVID-19 Contact Tracing Applications in Portugal: Effectiveness and Privacy Issues 1 Introduction 2 COVID-19 Smartphone-Based Contact Tracing in Portugal 3 Contact Tracing Technologies 4 Analysis and Conclusions References Robots That Look After Grandma? A Gerontechnology Point of View 1 Robots That Look After Grandma 2 Gerontechnology, the Intersection Between Technology and Aging 3 Robots Used in Caring for Older Adults in Everyday Lives At Home 4 Building Ambient Intelligence to Support Older Adults and Caregivers References Making Maximally Ethical Decisions via Cognitive Likelihood and Formal Planning 1 Introduction 2 Cognitive Calculi 2.1 Deontic Cognitive Event Calculus 2.2 Inductive Deontic Cognitive Event Calculus 2.3 Cognitive Likelihood 3 Highly-Expressive Automated Planning 4 Selecting Plans Using Cognitive Likelihood 5 Case Study: The Miracle on the Hudson 5.1 Recounting US Airways Flight 1549 5.2 The Setup 5.3 The Arguments 5.4 The Framework, Applied 6 Related Work 7 Conclusion References The (Uncomputable!) Meaning of Ethically Charged Natural Language, for Robots, and Us, from Hypergraphical Inferential Semantics 1 Introduction 2 A Household-Robot Case Study 3 Prior Framework for Engineering of Robots Rodney and Ralph 4 Montagovian/Model-Theoretic Approach to Meaning, Rejected 4.1 MTS in Summary 4.2 Two Fatal Defects in Model-Theoretic Semantics 5 Proof-Theoretic Semantics (PTS) 5.1 The Basic Idea 5.2 But What About PTS for Natural Language? 6 Hypergraphical Inferential Semantics (mathscrH mathscrI mathscrS) 6.1 Intuitive Kernel of mathscrH mathscrI mathscrS via Buffalo-Buffalo-ellipsis… 6.2 Hypergraphical Natural Deduction 6.3 An Example: The Meaning of `Emma Helped' 7 Applying mathscrH mathscrI mathscrS to the Robot Case Study 8 Questions/Objections and Replies 8.1 Objection 1: Redundancies & Irrelevancies 8.2 Objection 2: Termination 8.3 Objection 3: How are Divine-command Theories and Codes Formalized? 8.4 Objection 4: OS-Level Ethical Controls Mysterious 9 Natural Language Understanding Is Provably Uncomputable 9.1 What of Prior and Related Work? 9.2 Descriptive Complexity 9.3 Computational Cognitive NLU 9.4 Computational Learning Theory 10 Future Work; Further Objections References Reinventing Kantian Autonomy for Artificial Agents: Implications for Self-driving Cars 1 Introduction 1.1 Robot Ethics as Applied Ethics 1.2 Agency 1.3 Moral Agency 1.4 Can Animals Be Moral Agents? 1.5 Autonomous Robots 2 Self-driving Vehicles 3 Conclusions References Realizing the Potential of AI in Africa: It All Turns on Trust 1 Introduction 2 The Benefits of AI Depend on Adoption and Trust 3 AI in Africa 4 Accelerating the Exploitation and Adoption of AI in Africa 5 The Downsides of AI 6 AI Education in Africa Is the Key to Progress 7 Conclusions References Index