The present-day aviation system as a whole is of incredible complexity. It is the long-term result of a large number of individuals working together to improve several critical aspects: network safety, efficiency, resilience, etc. As any complex system, the components are numerous and variate and the relationships between them can be quite unclear. Engineers construct safer and more efficient aircraft. A different community of professionals independently ensures that the various aircraft do not collide en-route, all while distinct hierarchical companies compete for the slot attainment that will suit their business strategy. Also at the same time, pilots need to navigate weather complications during landing and airport operational people struggle to prepare the aircraft as fast as possible without neglecting safety. This non-exhaustive list of processes exists for every single flight, with an ever increasing number of activities operating on a daily basis.
The point is clear: the human capacity to provide logical explanation to emergent behaviour such as delay propagation is now limited. One might understand that one fight delay that is caused by a late passenger arrival can delay another flight’s departure. It is impossible to maintain a logic while causalities are spreading throughout the whole network. This is only one example of the limited capacities of the human “manual” analysis when the system’s complexity is high. The word “manual” has been highlighted on purpose. It is not our logic (or more specifically, the logic of the expert analysing a problem) that is at the core of the problem, but the fact that it has to be done manually is what takes indefinite time.
This is why algorithms and software were invented. It was understood a long time ago that a machine can do what we can’t do. Just look at the time needed to calculate the multiplication of two large numbers and compare it to the time spent by a computer solving the same equation. This machine integration is everywhere, as ATC controllers use them, companies use them for price tagging or optimising crew allocation, etc. However, computers only do what they are programmed to do (within its own limitation of speed and memory etc.). Until recently, machine programming was used to control and monitor. With the increasing number of data (that was before manageable for a singular expert person), computers are now used to execute expert analysis in a more complex environment.
In a sense, machine learning is just a prolongation of the expert mind, allowing them to project their logic into a more complex situation. Naturally, the expert, if lacking themselves the proper formation, must sometimes work hand in hand with new data scientists. This new kind of association is the future of aviation, as growing computer abilities is guided with expert area knowledge, fostering the knowledge discovery process and benefiting all the components of the system.