SHS/COMEST-10EXT/18/3 Paris, 16 July 2018 Original: English COMEST CONCEPT NOTE ON ETHICAL IMPLICATIONS OF THE INTERNET OF THINGS (IoT) Within the framework of its work programme for 2018-2019, COMEST decided to address the topic of the Internet of Things (IoT), reflecting on the ethical considerations of IoT in relation to society, science and sustainability. This work builds on the COMEST Report on Robotics Ethics (2017) and is undertaken with the participation of UNESCO s Communication and Information (CI) Sector. At the 10 th (Ordinary) Session of the COMEST in September 2017, the Commission established a Working Group to develop an initial reflection on this topic. The COMEST Working Group met in Enschede, the Netherlands, in March 2018 to discuss the content of its reflection. The Working Group s discussion was enriched by a public workshop on the Ethics of the Internet of Things (IoT) Technologies organized by the University of Twente in conjunction with its meeting. Within the context of the 10 th Extraordinary Session of COMEST that will take place from 10 to 14 September 2018 at UNESCO Headquarters in Paris, the COMEST Working Group decided to prepare a concept note providing a synopsis of its reflection so far. As it stands, this concept note does not necessarily represent the final opinion of COMEST and it is subject to further discussion within the Commission in 2018 and 2019. This document also does not pretend to be exhaustive and does not necessarily represent the views of the Member States of UNESCO.
2 COMEST CONCEPT NOTE ON ETHICAL IMPLICATIONS OF THE INTERNET OF THINGS (IoT) TABLE OF CONTENT INTRODUCTION CHAPTER I: INTERNET OF THINGS TECHNOLOGIES CHAPTER II: ETHICS OF THE IoT TECHNOLOGY (THE MICRO LEVEL) CHAPTER III: SOCIAL PRACTICES (THE MESO LEVEL) CHAPTER IV: POLITICAL (THE MACRO LEVEL) CHAPTER V: RECOMMENDATIONS
3 INTRODUCTION COMEST CONCEPT NOTE ON ETHICAL IMPLICATIONS OF THE INTERNET OF THINGS (IoT) 1. What are the ethical implications of the Internet of Things? Technologies are increasingly taking on an environmental role: rather than being used by people, they have come to be an environment for human practices, in which humans are immersed. Sensor networks are integrated in our material environment, while algorithms and artificial intelligence make sense of the generated data, and influence our public and private lives. Wearable sensors can be used to track the physical and mental condition of patients after surgery, in order to improve their recovery, and make care more effective and efficient. Animals can be tracked via sensor systems in order to study their behaviour, or to prevent poaching. Video and audio systems can detect disturbances in public spaces, and change the color of the light accordingly, to influence people s moods. Shopping windows can nudge people towards buying products by tracking their gaze and highlighting the product that seems to draw their attention. 2. All of these technologies seem to bring many benefits, but at the same time they raise important ethics questions. There can be biases in the sensors and in the algorithms that make sense of the data, for instance: detection systems might not work equally for every type of skin, and the boundary between normal and deviant behaviour on the street might be hard to draw. The transportation of data brings specific vulnerabilities, in terms of privacy, safety, and security. And the systems might have important implications for the ways we deal with education, care, and policing. What will the new world of smart homes, smart cities, smart hospitals, and smart schools imply for social interactions, power relations, scientific practices, city life, healthcare, and teaching? How are these technologies different from other types of technologies? How can we evaluate them adequately? And how can they be designed in a value-sensitive way? CHAPTER I: INTERNET OF THINGS TECHNOLOGIES 3. What is the Internet of Things (IoT)? Since the technology is evolving, a formal definition may be too strict, so we will focus on the understanding of the use of this term. Formally speaking, the Internet is a protocol for communicating computer networks, already developed more than 50 years ago. However, people use the term Internet to describe the ability to communicate with other people and to share data and information, using computers as their terminals. Traditional Internet is controlled by people, who should have access to a computer (including its mobile version), to actively log-in, and then to decide how to use it. That is, the Internet is a tool for people to connect with each other and to connect people to information and databases, where the users have full control on how such information and resources are used. 4. The internet of things basically does the same, replacing people by things. That is, by IoT technology things can communicate, share information and data. Moreover, when programmed to do so, the things can make decisions and operate without the involvement of a human. IoT technology is built up of three technological layers: a sensing layer, a networking layer (connectivity) and a data layer. Measurements are collected by sensors and then are communicated by the networking layer to the data layer. 5. Taking these three elements in reverse order, the technology of the Data layer is intimately related to big data. The challenge comes from the need to deal with a large amount of data, arriving at a data center from different, heterogeneous sources. The technology involves both computing power and data storage, enabling classification and labeling of the data, as well as processing it. The processing and the decision making are commonly done using artificial intelligence and machine learning algorithms for specific use cases.
4 6. The Networking layer is based on existing communication technologies (e.g., cellular, WiFi, Bluetooth) and computer communication protocols (e.g. Internet, Ethernet). Since an IoT sensor does not actively log in into the internet as a human does, issues regarding control over the connectivity of things may arise. However, once a sensor is connected, the connectivity issues are mostly generic. 7. In general, there are two types of sensors in the Sensing layer: designated sensors, and opportunistic sensors. Designated, special purpose IoT devices are usually developed to be small, cheap, reliable and of low power consumption. They must be able to sense specific physical features, and to communicate the information they sense. Such sensors can be located in specific locations, attached to a device or to a living creature; they can be wearable by a human, or even be taken as pills to sense in-body phenomena. In many cases, sensors are integrated into novel technologically based agents such as robots. However, when a person swallows a sensing pill, or buys Alexa, or wears a sensor, he/she knows that they are being monitored. Opportunistic sensors, on the other hand, are based on unintended use of existing technology as sensors. Many such examples exist, especially using a variety of builtin options in smartphones, carrying both environmental (e.g. location) and social information about the user. 8. While each of these layers in IoT is built on existing technologies, their integration and the new horizon of applications thereby opened bring in not only novel opportunities, but also, through their potential societal impact, new challenges. The end uses of IoT can be divided into three categories: a. Industrial applications. Here the purpose is to improve production processes. These applications are in line with the ongoing trend for increasing automation in routine, labor demanding jobs. It is attractive in a number of areas, e.g., the Food Supply Chain (FSC) where quality of food can be increased while decreasing costs. b. Public space. IoT can be used to improve public services. One leading example for this category is the Smart City, in which city services are provided based on real time data, continuously collected by a network of sensors (e.g., street cameras) covering the city. c. Consumer. IoT s private applications aim to improve individual well-being by e.g., enabling home appliances to sense and to be activated according to personal needs. 9. While the three layers of technology in these IoT categories are basically the same, each has different objectives, communities of stakeholders, legal and regulatory contexts, governance arrangements, and public expectations. The main technological challenges in all IoT applications are in controlling safety, privacy, cyber security, and avoiding misuse of the technology. The ethical challenges are however different, with issues such as avoiding bias, the interface between private and public, social and distributed justice. These challenges require more attention in IoT categories dominated by the for profit sector. 10. IoT systems belong to the emerging family of AIS autonomous, intelligent systems. For such systems Ethically Aligned Design (EAD) is crucial. IoT design, which considers the ART components (Accountability, Responsibility and Transparency), is central to enabling the implementation of specific ethical recommendations. 11. After several centuries in which technologies became ever more present in society and our daily lives, the current generation of digital technologies is rather actually moving to the background. They are becoming part of the material environment in which we live, changing this environment from a mute and stable background to an interactive and impactful context in which human practices and experiences are immersed. This new type of human - technology relations requires further analysis, in order to conceptualize their role in society and to anticipate and evaluate their impact.
5 CHAPTER II: ETHICS OF THE IoT TECHNOLOGY (THE MICRO LEVEL) 12. In this chapter we will, first of all, focus on the ethical aspects of technological building blocks of IoT, i.e. on various vulnerabilities arising from the way these building blocks are designed, produced and used. Whereas broader ( macro-level and mezo-level ) ethical implications of IoT (its social and political consequences in particular) will be discussed in chapters III and IV, respectively, this chapter will remain limited to its narrower ( micro-level ), but just as important, ethical moments. Special attention will be devoted to vulnerability and ethical aspects of four technological elements of IoT: (1) sensors, (2) networks, (3) data collection and storage, and (4) capacity for autonomous decision-making. Sensors 13. As optimal functioning of IoT systems depends on them being able to detect changes in their particular environments, they require accurate sensors, such as cameras, microphones, thermometers or motion sensors. Ensuring reliable functioning of such sensors is an ethical, as well as legal ( product liability issue) imperative. However, there are also other important ethical questions that will be discussed in this context. For example, one of them is the unintended, but nonetheless probable bias (e.g. along racial, ethnic or gender lines) when it comes to production and functioning of particular sensors (especially those designed to recognize human faces). Also, ethically relevant is the question about who has access to or control over sensors of particular IoT systems. The later question the question of the misuse or the responsibility for use of sensors is particularly important as long as such sensors can be put to dual use or used opportunistically (employed to collect data different than those they were originally designed to collect). Networks 14. Since IoT systems are basically networks of interconnected and mutually interacting devices ( things ), the nature of such networks gives rise to a number of specific ethical questions. One of the most prominent questions is the one about safety. Safety issues arise primarily due to the fact that human life is becoming more and more intertwined with IoT technology in areas as diverse as industry, transportation, urban life, home, health care, elderly care, scientific research, etc. Safety issues are pertinent to IoT systems because their complexity and, especially, the interconnectedness of their respective things or devices makes them highly vulnerable to accidental failures (technological unreliability) or intentional misuse (hacking and tapping data). Data 15. IoT systems (e.g. those designed to optimize traffic, health monitoring wearables and smart homes) are about to collect huge amount of data (especially thanks to their likely dual or opportunistic use, as already mentioned). Given the high commercial value of various types of data (both Big and Small ), the manner in which IoT systems will be designed to collect, store and process data is of extreme ethical importance. For example, an IoT system may or may not be designed to coarse grind (blur) images of people collected by its cameras, giving thus rise to not only to safety issues, but also to privacy issues (privacy issues related to IoT will be discussed more extensively in chapter 3). Storage of data collected by IoTs also has to be done in an ethically acceptable way (e.g. allowing anonymization of data about identifiable individuals or groups of people, deleting data after given period of time, not collecting unnecessary data and similar). Agency 16. A fourth ethical question to be addressed in this chapter is captured by the word agency. Whereas the impact of IoT systems on human agency and behaviour will be addressed in chapters III and IV, this chapter will address the agency of such systems
6 themselves. The problem is already known from the debates about robotics and AI: should such systems be designed so as to make decisions and act autonomously? Given their capacity to bring about not only benefits, but also harm to human beings, should there not be human in the loop bearing the legal and moral responsibility for possible harm or damages? The problem, again, is find balance between the ethical demand to keep the human in the loop and the necessity of excluding human from the loop in order to increase technological efficiency of such systems. This requires further ethical reflection about the possibility that certain IoT systems, especially those based on sophisticated AI technologies, might develop unforeseen or emergent properties and start making decisions or acting in ways nontransparent and untraceable to their original programming and design. CHAPTER III: SOCIAL PRACTICES (THE MESO LEVEL) 17. This chapter will start with a narrative, based on real life (no science fiction), which addresses as many diverse aspects and activities in daily life as possible: a day in the life of a person living in a world of IoT. The narrative will include issues of understanding and sensemaking (how does IoT shape our understanding of ourselves, the other, the world?); humantechnology relations; gender; sexuality; freedom; health; the generation gap; and cultural bias. 18. After this, the chapter will discuss the role of IoT technologies in various social domains: private life, the public sphere, social relations, cultural frameworks, sustainable processes, healthcare, scientific practice, education, and the functioning of the state. It will discuss issues of privacy (and how the concept itself is changing meaning as a result of technological developments), data ownership (and donorship), the representation and interpretation of human beings and their patterns of behaviour, cultural impacts, differences and bias, normalization, empowerment, dual use, misuse, and issues of well-being and the quality of life. 19. Third, the chapter will focus on sense-making: how does the IoT change our understanding of ourselves (self-understanding, existential questions, the quantified self), others (social relations, friendship, intimacy, public life); and the world (IoT in scientific practice, everyday experience). 20. Fourth, on the basis of this inventory of social impacts of IoT, the report will identify the main ethical issues in the social sphere, with a specific focus on issues of access, privacy, the quality of social practices, and the relations between the public and the private sphere. CHAPTER IV: POLITICAL (THE MACRO LEVEL) 21. This chapter focuses on the macro-level issues regarding the Internet of Things. How do IoT systems have implications on a societal and political level? How does it affect the functioning of the state, of companies, and political processes? The chapter investigates the relations between IoT technologies and politics by distinguishing between internal and external relations. 22. At the internal level, IoT technologies have implications for political structures (power, democracy, etc.) by offering new ways to read the city, changing the character of public spaces, opening new possibilities for both state surveillance and citizen participation. Central issues here are citizen participation and democratic representation; state surveillance and state responsibility; the future of democratic values, citizenship and technological literacy and education. 23. At the external level, the main question is how societies can learn to deal with the political implications of IoT technologies. Here, the central issues are security and system vulnerability; law, regulation, and human rights; the future of public infrastructures; the public/private distinction and ownership of data; new governance systems and the redistribution of state and citizen roles.
7 24. After this analysis, the report will analyse the ethical issues raised at the internal and external level, potentially including power relations; bias; democracy; human rights; changing responsibilities; power; data property; democracy; weaponization; surveillance; the changing relationships between society and the state; governance of IoT; access; legal regulation; cross-border issues; biopolitics; the ethics of big data. 25. Special emphasis will be put on the question of how the values of UNESCO - human rights, global justice equity, gender equality, SDGs can guide the way in which we deal with the structural impact of IoT? Also, this chapter will put the political significance of IoT technologies in the context of the fourth industrial revolution, and address the relations between IoT technologies and the economy (implications for commerce, and for economical values). CHAPTER V: RECOMMENDATIONS 26. This chapter will formulate recommendations. It will at least include recommendations on the following issues: Opportunistic IoT: existing infrastructure (like networks, cell phones et cetera) can be used to build opportunistic IoT systems what are the ethical implications? Ethics by design: is it possible to design ethical values in IoT systems? Full transparency: is it realistic to aim for full transparency in the collection and processing of data? Values: which values should we give a central place? The values of the French Revolution (freedom, equality, solidarity)? How to take value dynamism into account (value change via technology)?