AI, ML and Monkeys: Is Aviation Ready For The Paradigm Shift?

Artificial Intelligence. Machine Learning. Neural Networks. Gradient Boosted Trees. Monkeys playing Pong with their minds. It would seem intelligent machines are all the rage nowadays. It is inevitable then that more and more of these emerging technologies are finding their way into the aviation operations landscape. Optimisers, exception based alerting systems and autonomous decision making are already key themes in the aviation systems marketplace. The key question here is whether our existing operational organisations and structures are ready to implement and fully reap the benefits of this new technology.

The Human Touch

It is important to remember that a good system should aid humans in making better decisions. Both human and machine must be in sync to achieve the optimal outcome. The current mode of operations is largely reactive. Disruption events occur and teams scramble to find the solution they think is best. 

The key question here is how teams will need to react when predictive technology is implemented and decisions can potentially be made before an event occurs. Certain teams may choose to be aggressive and act on intelligence pre-emptively, whilst others may revert to a more conservative approach and wait for the event to occur before they act - just to be certain. It is important here from a management standpoint to clearly define and convey what the tolerance for risk is when dealing with predictions. Care must be given to educate teams on the nature of predictions and what the expected behaviour is. The evolution of training and SOPs is key.

Intuitive Design

Change is scary. As we transition into this new era, it is important to make this transition as comfortable as possible for teams. Information overload is a threat given the increased rate of data availability. It is important to have a clear understanding of what is important to an operations team, and what is just noise. False alerting generally tends to create mistrust between user and system.

Data points also need to be kept simple. There is an 87% chance of a delay occurring on flight XYZ of 8 minutes, with a confidence interval of 10 minutes. Confusing? Absolutely. However, this is the typical level of insight that can be churned out by any predictive model. It is important here to define which information should be presented to the user, and which should be kept at a system level. Simple, actionable insight is the key.

Patient & Unwavering

There is tremendous upside to these new systems. However as with anything that is novel, there is bound to be a certain level of uncertainty and error. Predictions will almost never be 100% accurate. However, this should not be a reason to shun these new technologies and revert to the incumbent. It is important that these new systems are battle tested and given the opportunity to grow and improve. After all, knowing something even 1% of the time is better than not knowing 100% of the time.

TLDR

·     Intelligent technologies provide massive upside to the aviation operations landscape

·     Organisations must evolve their teams and the human element to make full benefit of these systems and tech

·     Organisations must also find a sensible way to deal with the risk that comes with predictive decision making