Free beta access. 5 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

Uncertainty over AI military application puts NATO at the center of the broader debate over democracy and emerging technology – cooperation now sets international norms.

Gilli ’20 (Andrea is Senior Researcher at the NATO Defense College where he works on issues related to technological change and military innovation. In the past, Andrea has been visiting and post-doctoral fellow at Johns Hopkins University and Columbia University as well as Stanford University, “NATO-Mation”: Strategies for Leading in the Age of Artifical Intelligence,” https://www.ndc.nato.int/about/organization.php?icode=126, 6/25/22)-lf

NATO has an interest, and a moral obligation, to promote the adoption of its values in the realm of AI. Democratic values inform the Alliance’s goals, thus giving meaning to its material capabilities – including its military power. At the same time, the integration of AI into the fields of security and defence poses unique moral, ethical, legal, and safety-related questions.85 It is thus imperative that the Alliance actively considers and operationalizes AI ethics, regardless of the degree and scope of AI integration within NATO and its Allies. The common principles and values pronounced in the Atlantic Treaty represent the foundations on which NATO was built. Such principles and values – democracy, freedom, rule of law, individual rights, free markets – are the bond underlying the transatlantic community, which in fact predates the Atlantic Alliance.86 Norms and values can strongly shape the international system.87 Given the intense international competition in the technological, military, economic, and normative domains, embedding democratic values into AI is as much a strategic imperative for the Alliance as it is a functional one.88 In addition to signaling to domestic populations that NATO and its Allies follow through on their commitments to uphold values as the basis of the political and military Alliance, incorporating AI ethics into the “NATO-mation” agenda also serves as a bulwark against the incursion of unwelcome illiberal values in the course of future technological development.89 It is worth noting that NATO’s role in the AI ethics sphere differs from that of many other organizations, such as national governments and the European Union, because NATO is not a regulatory body. NATO complies with existing laws and regulations, including the Laws of War, which nations and the international community created. This means that regulatory and normative questions such as the development and deployment of autonomous weapons systems will not be determined at NATO level. Nevertheless, there is still room for NATO to play a clear role. Indeed, given doubts and worries about the adoption of AI for military purposes, NATO can help generate more public support and engagement by clearly defining ethical boundaries and moral guidelines. The uses of AI in military operations – ranging from logistics to maintenance, from recruitment to retainment, from intelligence, surveillance and reconnaissance to medical tests and medical evacuation, and more90 – go beyond the discussions on lethal autonomous weapons systems that have dominated European debate about AI in military affairs. Accordingly, the range of ethical questions relevant to NATO extend beyond focusing on the tip of the spear. Seemingly mundane uses of AI, such as in human resources or decision support, can still pose distinct ethical questions the Alliance should be prepared to handle. Addressing security risks,91 minimizing bias in systems,92 developing trust,93 and respecting privacy are fundamental tasks for the Alliance to ensure the future effectiveness of AI, whether in battle or in other functions. The age of intelligent machines requires the Alliance to reiterate its commitment to values as new moral and ethical questions emerge, because algorithms do not have a conscience, personal preferences or moral agency. These statistical machines have no understanding of good and bad, or fair and unjust.94 All an algorithm can do is achieve its human-defined reward function, not provide any context or information on whether the right question is being asked.95 Instead of giving moral agency to algorithms, humans and organizations can view AI as a “moral entity”. This means that we humans are dutybound to adhere to our moral code of conduct when interacting with the systems, rather than shirking human responsibility to computers.96 At the organizational level, this means that the design, development and deployment of AI should be “ethically aligned” with the Alliance’s values and goals.97 This is important because AI is data-intensive, unpredictable and brittle.98 While a system may work well in the context in which it was trained, it may break in an unfamiliar setting. The reliance on BD in the current second wave of ML also creates fallibilities,99 given that the algorithms can scale up harm if the data over- or under-represents certain groups.100 For neural networks in particular, it may be impossible to explain or interpret results. Some refer to this as the “black-box” problem, meaning that the outcomes of these complex AI systems are opaque to humans that either want to reproduce the good, or prevent the bad from recurring. While traditional software can be debugged to solve a performance issue, the lack of linear causality between a programmer’s inputs and the AI system’s outputs means that it is difficult to track bias and reliability. Creating organizational processes to minimize these concerns across the AI lifecycle is critical to responsible use of the technology. Considering AI ethics also means developing trust in systems. Organizations developing AI applications should think deeply about user expectations related to transparency and disclosure. Ultimately, these norms will change the distinction between human, AI-assisted and AI interactions.101 “Calibrating” trust is especially important for AI-assisted decision making,102 a concept whose value will only increase given the emphasis on human-machine teaming in Allied militaries.103 In many cases, the unethical or unacceptable outcomes of AI systems pertain not to moral dilemmas, but to the reliability and robustness of the systems at hand. As such, building trust is fundamentally tied to building safe and secure systems.104 The problem of bias in AI illustrates the overlap between AI safety and AI ethics: bias is morally problematic, because it can unfairly harm or systematically discriminate against specific groups105 – and it can also be seen as a failure mode that reduces the reliability of a given algorithm in a given context.106 For NATO, this relates to the safety of enterprise tools and weapons systems alike. In deployments, these safety measures would be critical to ensure that AI-enabled weapons systems are used in a manner consistent with the principles of international humanitarian law. As explained in greater detail below, this may also feed into the eventual standardization process. While addressing these and other ethical issues is not impossible, it is also not immediate, especially among different actors who, understandably, may hold variegated views or priorities. Several courses of action are possible: Appoint an Ethics Board. The Atlantic Alliance could benefit from establishing an Ethics Board of experts, along the lines of the European Commission’s High-Level Expert Group on AI or the AI Ethics Committee in the French Ministry of Armed Forces. The aim should be to start an internal discussion about the applications of AI in the military domain, and the board should be empowered to monitor the progressive implementation of ethically aligned processes, principles, and standards over the medium-to-long term so as to avoid “ethics washing”.107 The board could be made of specific personalities, or also represent ethics boards already set up by individual Allies. It is important, however, that a board of this kind remains in place for the medium term: the upcoming integration of AI at entry-level will lead to new applications, and thus to wider adoption within the Alliance; for this reason, an ethics board would be increasingly called on to support trust-building measures, as well as integration of ethics into the front end of development. Similarly, it would play an important support role in providing the necessary governance for technical, moral, legal and ethical questions as the adoption of AI intensifies.108 Lead and shape with ethical principles. The proposed Ethics Board could start its work from scratch, defining a new set of guiding ethical principles for NATO and its Allies, or it could take a pragmatic approach and borrow from the important work already done by several institutions, including the OECD and the European Union.109 As recalled above, work has already been done by international organizations such as the European Union and the OECD, national governments, such as the US Department of Defense and the French Ministry of the Armed Forces, private companies like Google and IBM, as well as non-governmental institutions like the Institute for Electrical and Electronics Engineering.110 Having clear and simple ethical principles is important, both internally, as actors at different levels need to be able to make choices, and externally, as NATO and its Allies may want not only to signal their ethical and moral commitments to their own citizenries, including developers and civil society, but also to shape the international environment. The priority is to ensure that NATO Allies adopt clear ethical guidelines which reflect their values – such as democratic representation, civilian control of the armed forces, individual responsibility, the centrality of human life, and compliance with the Laws of War – and that such values inform international norms, practices, agreements and possible future treaties. Leading in this realm means moulding the future security environment, in particular by embedding democratic values into this pervasive technology.111 What ethical principles should NATO adopt when it comes to AI? This will be up to the Ethics Board, if established, or to other authorities. As recalled, there is, however, a general agreement on some principles:112 • Human centricity: AI must be based on fundamental human rights and be aimed at improving the human condition; • Safety and security: AI must be adopted and deployed in a risk management framework that accounts for and minimizes threats; • Explainability and transparency: algorithms are generally opaque. For this reason it is important that their underlying logic can be understood, and that algorithms are written in a clear way; • Responsibility and accountability: the deployment of AI must ensure that responsibility for actions can be identified, especially in the event of error, meaning that individuals or organizations can be accountable even when responsibility is distributed across many stakeholders; • Reliability: the development of AI must ensure that it is reliable, and thus does not operate unpredictably or, especially, pose unanticipated threats; • Fairness and inclusion: algorithms must be designed to improve the human condition and thus reduce bias and discrimination, not exacerbate them; and • Privacy and data governance: the privacy and sensitivity of data – with regard to individuals, equipment, processes or organizations – must be preserved and protected. Implementation and execution. The transition from principles to action is critical in any activity. Ethics is no exception, and while issuing comprehensive principles is difficult in any organisation, ensuring their implementation is an even greater challenge so as to strike the fine line between ethics washing and ethics “bashing”.113 Indeed, governance is an important factor in ensuring that technology is used responsibly and in line with the Alliance’s desired outcomes.114 In this respect, NATO may have a relevant role in supporting its Allies. A dedicated organization, such as an Artificial Intelligence Integration & Implementation-Enabling Centre (next chapter), could prove particularly useful, not least because it could help Allies share concerns, considerations, solutions and best practices as well as support them with ad hoc training activities.115 Similarly, aligned principles may eventually be translated into standards – and related measures such as benchmarks and annotations – an aspect which will be discussed towards the end of this Research Paper.

Sign in to recut, update, or save this card.