Ethics and Digital Transition
Description
More details
Other editions
Additional editions
Persons
Marie-Hélène Abel is Professor and Head of the Computer Science department at the University of Technology of Compiègne, France. She was President of the IEEE-SMC French chapter from 2014 to 2021.
Nada Matta is Professor of Knowledge Engineering and Management at the Université de Technologie de Troyes, France. She was Co-Chair of the IEEE-SMC French chapter from 2022 to 2024.
Hedi Karray is Programme Manager for Artificial Intelligence at the European Innovation Council, and a former professor at the University of Technology Tarbes Occitanie Pyrénées, France. He was Co-Chair of the IEEE-SMC French chapter from 2022 to 2024.
Inès Saad is Professor of Information Systems at Amiens Business School and a researcher at the UPJV's MIS laboratory, France. She is also the Chair of the ICIKS conference and the IEEE-SMC French chapter.
Content
Foreword xi
Jean-Gabriel GANASCIA
Introduction xv
Nada MATTA, Marie-Hélène ABEL, Hedi KARRAY and Inès SAAD
Chapter 1 Digital Ethics: Empowering Agents and Taking Care of Systems 1
Bruno BACHIMONT
1.1 Introduction 1
1.2 Technology and ethical neutrality 3
1.3 What are the ethics of technology? 5
1.3.1 Contemporary ethics 5
1.3.2 The ethics of moral agents 7
1.4 Calculation, algorithm 10
1.5 The ethical challenges of digital technology 14
1.5.1 Complexity and technology: temporal alienation 14
1.5.2 Data, algorithms: alienation of property 16
1.5.3 Moral agents and their acolytes 18
1.5.4 Ethical situations 19
1.5.5 Empowering agents 20
1.5.6 Taking care of patients 22
1.5.7 Taking care of witnesses 23
1.6 Conclusion 24
1.7 References 25
Chapter 2 Bias, Discrimination and Decision-Making: Fate or Responsibility? 29
Florence SÈDES
2.1 Why this question? 29
2.2 Bias, history and typology 31
2.3 Types of bias 35
2.4 Bias and fairness - a measure of ethics? 37
2.4.1 A mixed picture 37
2.4.2 How can we define algorithmic fairness? 40
2.4.3 And in practice... how can these biases be corrected? 42
2.4.4 Fairness and interpretability: finding the right balance? 44
2.5 "We are open, the door is just very heavy" 44
2.6 Biases: fatality or responsibility? Fatality and responsibility&... 46
2.6.1 Synthetic data for a qualified bias correction approach 47
2.7 Debiasing machines 48
2.8 References 49
Chapter 3 Digital Technology and Artificial Intelligence: How Can We Facilitate the Ethical Control of their Use? 51
Alain MILLE
3.1 Introduction 51
3.2 Expected properties of an ethics-oriented DTD-AI 54
3.2.1 Property I: construction and memorization of knowledge of situated action for the technician (Simondon) as tertiary retentions (Stiegler) 59
3.2.2 Property II: supporting the process of discussing and regulating activity 61
3.3 Illustrations of DTDs integrating activity reflexivity for individual and collective documentation purposes 65
3.3.1 Support from M-Traces for a corporate community of practice 66
3.3.2 Support for human learning activity 68
3.3.3 Example of an activity involving a ChatGPT-type DTD-AI 69
3.4 Conclusion 71
3.5 References 76
Chapter 4 Ethical Autonomous Agents: Literature Review and Illustration for Markov Decision Processes 81
Grégory BONNET, Nadjet BOURDACHE, Abdel-Illah MOUADDIB and Mihail STOJANOVSKI
4.1 Introduction 81
4.2 Issues specific to the integration of ethics 83
4.2.1 A high-level approach to ethics 83
4.2.2 Programming ethical autonomous agents 85
4.3 Ethical autonomous agents: state of the art 87
4.3.1 Qualitative implicit architectures 87
4.3.2 Qualitative cognitive architectures 88
4.3.3 Quantitative cognitive architectures 93
4.3.4 Learning an ethical model 95
4.3.5 Intermediate conclusion 97
4.4 Ethical Markov decision processes 98
4.4.1 Markov processes and ethical contexts 99
4.4.2 Modeling ethics 102
4.4.3 Solving E-MDPs 105
4.5 Establishing ethical principles in E-MDPs 107
4.5.1 Theory of divine command 107
4.5.2 Prima facie duties 109
4.5.3 Virtue ethics 110
4.6 Conclusion 114
4.7 References 115
Chapter 5 Ethics and Ecology in Production Systems 121
Emmanuel CAILLAUD and Lou GRIMAL
5.1 Introduction 121
5.2 Ecological context 122
5.3 Integrating ethical issues into an industry in ecological transition 126
5.3.1 Attempts to meet standards 128
5.3.2 From environmental ethics to ecological ethics 132
5.3.3 Ecological ethics and technology 133
5.4 Proposed areas of work 134
5.4.1 Competence of production system players 134
5.4.2 Purpose of the production system 135
5.4.3 Returning to planetary limits with available technical systems 136
5.5 Conclusion 137
5.6 Acknowledgments 138
5.7 References 138
Chapter 6 Operational Ethics in Industrial Systems of the Future: Methodological Elements 141
Damien TRENTESAUX, Lamia BERRAH and Karine SAMUEL
6.1 Introduction 141
6.2 Ethics: definition, typologies and paradigms 144
6.3 Performance management for 4.0 industrial systems and its ethical risks 147
6.3.1 Overview 147
6.3.2 State of the art on ethics in industrial systems of the future 149
6.4 Toward the operational integration of ethics in the 4.0 industrial systems 153
6.4.1 Scope 154
6.4.2 Methodological elements 154
6.5 Industrial testimony 159
6.6 Conclusion 164
6.7 Acknowledgements 165
6.8 References 165
Chapter 7 AI for Industry: Transforming the Daily Lives of Maintenance Operators 171
Anne DOURGNON, Eunika MERCIER LAURENT and Alain ANTOINE
7.1 Genesis of this innovation 171
7.2 Background of this innovation 173
7.2.1 A low-key innovation 173
7.3 The naysayers 174
7.4 Dealing with hazards 176
7.5 A decisive demonstration 177
7.6 The keys to success 178
7.7 Overcoming obstacles 179
7.8 Outlook 180
7.9 Acknowledgments 181
7.10 References 181
Conclusion 183
Hedi KARRAY
C.1 From ethics to responsibility 183
C.2 Principles and challenges behind responsible AI 184
C.3 Responsible AI: a threat or an opportunity? 186
C.4 Responsible AI challenges 187
C.5 An ecosystem for responsible AI 189
C.5.1 Actions for responsible AI 190
C.5.2 Policy measures: initiatives and regulations 191
C.5.3 Technical progress 191
C.5.4 Public awareness and education 192
C.6 References 192
List of Authors 195
Index 197
Introduction
Ethics and Digital Transition Challenges and Investigations
Ethics was originally defined in antiquity as "moral principles" (Singer 1986; Aristotle 2019; Frey and Wellman 2008) dictating the virtues of behavior (Hursthouse 1999). Today, it is increasingly identified with principles of deontology and social rules linked to the consequences of actions (Bonhoeffer 2012; Siau et al. 2020). The dimension of evaluating activity and systems has therefore become important in order to respond to these principles.
Furthermore, society's digital transition is tending to exploit systems emanating from artificial intelligence (AI) and data processing. AI techniques tend to simulate behavior and, above all, reproduce thoughts and actions. The consequences of these actions must then be assessed in the light of social rules and ethics. In fact, AI techniques are mainly based on sampling and data analysis, on the one hand, and cognitive rules and procedures, on the other hand. The results of these approaches modify our everyday behavior by introducing new elements generated by the connections of massive knowledge processing and algorithms. Deep and machine learning, as well as ChatGPT1, are among the main examples of this invasion of our activity.
The main question debated in this book is: "what are the different aspects to be considered when assessing ethical principles in approaches to digital transition and intelligent systems in particular?". To answer this question, we first explain the main technologies, techniques and uses of intelligent systems. We then explore the works addressing ethics in digital transition to highlight the ethical dimensions to be taken into account in the development of these systems. These are extracted from a survey of digital researchers. Finally, we introduce a summary of the seven chapters of this book that address these issues.
These investigations are being carried out as part of the activities of the French chapter of IEEE SMC2, where a number of initiatives focus on studying the relationship between digital technology and human activity.
I.1. The digital transition and its challenges
Information technologies increasingly offer processing approaches that enable us to understand the internal and external socio-economic ecosystem. On the one hand, these techniques make it possible to capture and exploit data and information produced by an activity and/or existing in the environment, and, on the other, to provide decision-support tools. Socio-economic players are therefore called upon to grasp these technologies and integrate them into their organizations (Hesse 2018).
The digital transition is defined as the integration of information processing technologies, while conveying a profound change in habits, to enable an understanding of the ecosystem, thereby leading to better organizational performance (Hesse 2018; Zacklad 2020). We can cite, as an example, the massive exploitation of teleworking support tools (Zoom®, Microsoft Teams®, Webex®), notably during the Covid-19 health crisis. Similarly, our understanding of the environment is currently raising social awareness, leading to more sustainable action.
Advances in information processing technologies are bringing about radical changes in activities and behavior, particularly in communication and decision-making (Figure I.1) (Zacklad 2020; Vial 2021).
These technologies are largely based on AI approaches that have proven their worth in supporting ecosystem understanding and decision-making.
Figure I.1 The digital transition of organizations (Vial 2021)
I.1.1. Artificial intelligence techniques
The basic principles of AI approaches are to represent human reasoning and behavior in computational techniques (Fetzer et al. 1990; Dick 2019). For instance, rule-based systems tend to illustrate mainly deduction, case-based reasoning stands for inferences by analogy, while machine learning algorithms tend to simulate induction.
We also mention multi-agent systems that represent the cooperation of bees and ant communities. This similarity in the behavior of living organisms is leading to real collaborations between humans and AI algorithms, and not just to help with decision-making and assistance.
Some AI approaches use cognitive dimensions and propose logical reasoning based on experience feedback knowledge (ontologies, rule-based and case-based systems). Other techniques use statistical data processing, based on data lakes to aggregate features and generate rules and reasoning models (neural networks, deep learning) (Hunt 2014).
These two approaches interconnect through supervised learning, currently referred to as hybrid AI (Figure I.2).
Figure I.2 Main artificial intelligence techniques.
Important questions arise concerning the influence and relevance of these techniques for understanding the ecosystem:
- Are the expertise and data used in these techniques enough complete to suggest models for reasoning and effective decision-making? Are the data global enough to represent real-life situations?
- Can these models be so complete as to claim to represent different aspects of human reasoning?
- Can these approaches recognize and avoid erroneous data and incomplete experiments?
I.1.2. The main applications of AI
From 1955 (the birth of the notion of AI and the Turing test) to the present day, the application of AI has grown exponentially, especially with the increasing computational capabilities of machines. In terms of applications, AI was initially used in healthcare and industry as knowledge- and case-based systems (Dick 2019). Other techniques, such as fuzzy logic and neural networks, are used in image and speech processing (Hunt 2014) and robotics. Similarly, multi-agent systems are used in networking and Cloud Computing.
We can cite a number of applications for these systems in certain fields (Figure I.3):
- natural language processing: translation, information retrieval and text generation for medical, industrial, marketing and legal applications;
- image processing: supervision, cultural and archaeological recognition, facial recognition, medical diagnosis, climate change, augmented reality, digital twins, etc.;
- digital data processing: problem prediction, maintenance, behavior prediction, supervision, customer-market relations, recommendation, etc.;
- the Semantic web: data web, information retrieval, text generation, chatbot, social networks and mutual aid, e-learning, etc.
These tools lead to a transformation in user behavior and a major organizational change while considering the influence of data characterization and the prediction on the ecosystem behavior. These approaches not only represent human behavior, but also emphasize their mutual influences with the environment. We can even mention the possibility of self-evolution of AI systems, just like the reasoning they represent. The apprehension of these techniques increasingly raises ethical questions, which are essential to the integration of these approaches in the socio-economic environment.
Figure I.3 Examples of AI application fields.
I.2. Ethical principles
Ethics, by definition, is related to morality, virtues of behavior and social rules (Hursthouse 1999). The issue of applying theories of morality and virtues as principles of ethics (Siau et al. 2020) is addressed in several sciences, including law, medicine, business and engineering (Frey and Wellman 2008). As a result, principles were then defined, wherein the consequences of these sciences in society are mainly studied. We can note, for example, that in the medical sciences, certain principles have been prescribed such as common goals, fiduciary duties, legal and professional standards and responsibility, as well as methods for transforming these principles into practice. Similarly, the notion of environmental ethics and the sustainable behavior of human in their ecosystem are studied in the engineering sciences (Palmer et al. 2014). In this science, ethical concepts are introduced as responsibility, autonomy, virtue, right and moral status (Powers and Ganascia 2020).
I.3. Ethics and artificial intelligence
The application of theoretical ethical definitions in the digital transition primarily points to the analysis of the nature and social impact of this transition and the justification of this impact (Newell and Marabelli 2015; Majchrzak et al. 2016; Mittelstadt et al. 2016). Companies are invited to manage a trade-off between their performance and ethical principles (Vial 2021). These trade-offs should be at the operational and strategic levels (Zacklad 2020; Vial 2021). The application of ethical principles in the digital transition is strongly linked to the integration of data processing applications and AI technologies.
Currently, several scientific organizations such as the ACM and IEEE have defined certain ethical principles for the digital transition and AI. The OECD (2019) and the European Commission's Artificial Intelligence Expert Group (AI HLEG 2019) have...
