Assuring Human and Artificial Intelligence are Appropriately Informed in Aviation Systems

Developments driving adoption of new entrants to aviation, namely artificial intelligence (AI) and more ubiquitous information, will redefine roles as they create teams of humans and machines to manage facets of aviation operations. The safety outcome depends on how well the human-machine team is designed and controlled to evolve. This occurs via distributed decision making, where decision makers need shared understanding of how their decisions affect others, and the possible implication of others’ actions on their remaining options. The information available to each decision maker must provide the context and coordination they need to create this shared understanding. The safety controls and mitigations must account for what a human "on the loop" or "over the loop" know about what the AI is thinking and what they add to command a better safety outcome. The machine must "know" what the human is thinking and the outcome they are trying to achieve. And then the machines must understand each other, just as the people must also understand one another’s options and actions. The key to adopting these new entrants is shared understanding of information in terms of inputs, context, assumptions, decision points, options, and desired outcomes. This paper examines recent efforts to capture information needs in model-based system engineering tools. It highlights extensions proposed by the authors to address the human-machine team needs. The extensions include a construct for complex, distributed decision-making environment like aviation where coordination is essential to prevent decisions from undermining the strategy of the mitigation layers.