Making an impact: supporting effective decision-making in the information environment

The growth of artificial intelligence (AI) capabilities has ushered in a new era of human-machine collaboration. In particular, the potential of using AI-powered automated technology to support decisionmaking at all levels of warfare is becoming an ever-closer reality. However, for the planning and execution of operations in the information environment, the sensitivity and human- centric focus of these operations means it can be particularly important to maintain human decision-making at the core. The aim of the paper is to demonstrate how AI-powered semi- automated analysis tools can support human decision-making by effectively analysing data and suggesting possible decision options. This paper presents a semi-automated decision support tool for information-related activities (IMPACT) that harnesses the use of AI in an optimal manner while ensuring humans maintain meaningful control of decisions. Using IMPACT as a tangible example, this paper argues that while there is room for more AI-powered tools to help support decisionmaking in planning operations in the information environment, a balance needs to be struck with human cognition. The analysis and examples presented in this paper shows how such a balance can be struck to ensure humans remain at the heart of decision-making. 1.0 INTRODUCTION The explosive growth of possibilities within artificial intelligence (AI) is ushering in a new era of humanmachine collaboration. Across many different sectors of everyday life, AI is making significant progress, including “narrower” AI systems being deployed to make predictions and support decision-making with success in non-military domains, such as in the commercial sector 1, healthcare 2, and education 3. For the military, using artificial intelligence and automated decision-making has been an accelerating topic of discourse across the last decade. Indeed, there has been a rapid growth of opportunities for AI technology to supplement human decision-making at all levels of warfare 4. This drive has been reflected in recent doctrines that have acknowledged this rising implementation, specifically in the operational environment 5. 1.1 Literature Review The promise of AI-powered technology to bolster command-and-control (C2) decision-making processes within the information domain has been particularly highlighted in recent years 6. Indeed, AI-enabled capabilities have been touted as able to significantly enhance the potential of forces to effectively operate in the information environment. For example, machine-learning tools that can process and analyse large volumes of data quickly can help decision-makers observe and orient more efficiently and accurately. Moreover, there are possibilities for machine-learning algorithms to identify patterns in data that might be too subtle or complex for human analysts to detect, potentially enhancing intelligence gathering by providing new insights that improve situational awareness and anticipation of an adversary’s actions 7. For instance, it can provide a more comprehensive and up-to-date picture of the battlefield than humans through a greater ability to process and integrate data from various sources. It can even be used to suggest optimal courses of action, which is particularly useful in conflict scenarios where quick and precise decisions in the information environment are critical 8. Making an Impact: Supporting Effective Decision-Making in the Information Environment 8 - 2 STO-MP-HFM-377 While the promise of emerging AI technologies has been appealing to militaries, several voices have also urged for caution. James Johnson, for example, argues for a cautious and balanced approach to integrating AI in the military domain, emphasising the need for AI to complement, rather than replace, human judgment and decision-making: “military decision-making is different as it is non-linear, complex, and occurs in uncertain environments. In C2 decision-making, commanders’ intentions, the rules of law and engagement, and ethical and moral leadership are critical to effective and safe decision-making in the application of military force” 9. Johnson’s caution broaches a larger discussion of ‘meaningful human control’ within the field of AI ethics, which scrutinises the increasingly autonomous nature of machines in our everyday life, and posits that development of such technology should not come at the expense of human oversight. This is particularly so for fully-automated systems, such as autonomous vehicles or robots, that are fully capable of operating independently without human intervention. Yet, increasingly, this discourse is focusing on the ethics of deploying AI-powered technologies in military settings. Recent examples, such as the Israeli deployment of AI to identify Hamas targets in the war on Gaza 10 or the use of autonomous drones by the US military in conflict zones 11, have pushed important ethical and moral questions to the forefront regarding the tension between developing efficient and useful AI technology and retaining human accountability in morally complex situations 12. Artificial intelligence technologies are often viewed as “devoid of a social dimension” 13. Indeed, how far can we responsibly embed AI in military decision-making for complex conflict situations with potentially significant civilian impact? In the information environment, targeting human audiences requires decision- making grounded in values, ethics, and a nuanced understanding of human nature and psychosocial dynamics. Removing humans from these decisions poses the risk of neglecting these vital aspects, which can result in the dehumanisation of intended audiences. For the planning and execution of operations in the information environment, the sensitivity and humancentric focus of these operations means it can be particularly important to maintain human decision- making at the core. On a broader level, the embedding of AI-powered technologies poses strategic and ethical dilemmas. Current knowledge of automation bias, for example, might point to the potential danger of intelligence staff becoming overly reliant on machine decision-making. AI-powered technologies cannot understand the emotional impact of certain suggested decisions on audiences targeted by the operations, and will likely fail to adequately address the ethical dilemmas that may be posed by certain recommendations. Johnson suggests that “policymakers risk being blind-sided by the potential tactical utility – where speed, scale, precision, and lethality coalesce to improve situational awareness – offered by AI-augmented capabilities, without sufficient regard for the potential strategic implications of artificially imposing nonhuman agents on the fundamentally human endeavour of warfare” 9. Yet on a more specific level, AI technology might struggle with tasks that are not routine and without clear parameters. Human qualities such as experience, flexibility, creativity, and empathy are crucial for effective decision-making - these are qualities that an AI-based system lacks. Difficulties in the utilisation of AI in the information domain have been demonstrated by, for example, research gauging Iranian citizens’ sentiments towards the Iranian elections. Although the automated analysis of social media was seen as promising for “assessing public opinion or outreach efforts and forecasting events such as large-scale protests”, the authors also highlighted problems with AI technologies accurately interpreting text within the context of the exercise, and the necessity of maintaining human supervision 14. Again, James Johnson highlights that “because these quantitative models are isolated from the broader external strategic environment of probabilities rather than axiomatic certainties characterised by Boyd’s “orientation,” human intervention remains critical to avoid distant analytical abstraction and causal deterministic predictions during the nonlinear chaotic war” 9. There are also technical limitations associated with current AI technology that present pitfalls in their applicability in decision-making processes. At present, AI technologies exist at different levels of maturity, with many uses of AI still in infancy. Analyses such as aspect sentiment analysis are relatively robust Making an Impact: Supporting Effective Decision-Making in the Information Environment STO-MP-HFM-377 8 - 3 compared to an AI-powered chat agent that is fully integrated into all steps of a decision-making process. This is because AI tasks relating to reasoning – be it mathematical, logical or causal – are still in early development and concerns have been raised over the risk of increased hallucinations (i.e. false or misleading results generated by AI models) 15. As such, more research and experience is required before such AI functions can be fully integrated into decision-making systems. That is not to say, however, that simpler, more robust information analyses, like sentiment analysis, cannot play an important role in supporting decision-making. For this reason, semi-autonomy – the combination of human and machine actions, with both elements typically coordinated by a computer system – can be a more realistic and ethically preferable option for military contexts. The desire here is that semi-autonomous systems develop to efficiently support human operators, so that this collaboration can outperform either system alone. Vold discusses how human and AI systems can become so closely integrated that they eventually work in symbiosis to