Is it possible to use AI as a support system for emergency decisions when operating aircraft?
After crashing into birds that caused the total loss of both engines, the pilot Chesley Sullenberger, or better known as a Sully, managed to successfully land an Airbus A320 aircraft in the middle of the Hudson River in New York, saving 150 lifes. After several investigative analyzes the flight simulators showed that it would be possible to land safely at a nearby airport and that this decision put people at risk instead of saving them from death.
It seems a hasty attitude, or even negligent, but he followed all the emergency procedures to start the engine again, but until he concluded that all his actions have no effect, he had to make a completely unexpected and creative decision when visualizing the Hudson river like a big airstrip and only possibility of survival. Something that was not added during the aeronautical investigation simulations was the human factor, more specifically the reaction time. The simulations carried out by the FAA’s Research Department were successful, as they already knew the cause of the crash and directed the aircraft immediately for landing, however when considering all procedures as performed by the pilot, adding 15 seconds of reflection, no simulation could land safely at the nearest airport, and sully was considered a hero.
Although aviation safety standards have reduced the number of tragedies over the years, the falls will never be eradicated. Since the Second World War, the majority of flights with fatalities have been in the United States, a country that accounts for a total of 821 air accidents, according to the most recent data on the “Aviation Safety Network” website. The figure below shows the number of fatal accidents since 1946. If the attitude of the sully pilot was any other during his landing on the Hudson River, the crew and all its passengers would be another number of these fatal accidents.
And what does this have to do with Artificial Intelligence acting as a decision support system in emergency situations? Everything. If there was an intelligence system that could not only simulate infinite landing possibilities during this incident, using only machine learning, but also be able to interpret all actions performed by the pilot both in the system and in the face and body, as well as process the output responses of the commands of the aircraft, it might be possible to recommend to the pilot an alternative that was different from returning to the nearest airport and perhaps considering an emergency landing in a location despite being unexpected would protect all passengers.
The neurologist Damasio, in his book,“Descartes’ Error” presents the role of reason for decision making. In some patients with frontal lobe disorders they indicate that the ability to feel has an impaired ability to make decisions, that is, emotions are vital for us to function as rational human decision makers. Thus, a system that simulates several possibilities of successful landings using machine learning is not enough to support the decision making of an emergency situation such as the landing of the sully, in which case the machine needs to think and feel what is happening , to support decisions not only rationally, but also emotionally, which will involve an interpretation in addition to mathematical calculations, because despite ancient philosophy dissociating these two concepts today we have several studies in the area of psychology and neuroscience showing that reason and emotion are fundamental to model something that is close to human intelligence. In cases of operational emergency as occurred with the sully pilot, emotion and intuition was present for decision making, and therefore interpreting all the data and feeling the burden of responsibility and emotions involved must also be considered to model artificial intelligence, the need to save all those people was the main trigger for sully to decide to land in the water, that is, any other action could have caused an even greater accident.
Analyzing now the development of an artificial intelligence model that have all these aspects already discussed previously acting as a decision support system, it is very important to understand the psychology and neuroscience involved behind an artificial intelligence model. Daniel Kahneman in his book “Thinking, Fast and Slow” where we find our immediate reactions to certain events and one slowly, where our most structured thinking, known as system 1 and system 2. Interpretation for modeling an artificial intelligence system that has an effective computation behind its development, emotions and intuitions have a very important value for the development of these types of systems, creating layers of intelligence that is instantaneous like system 1, returning quick responses for decision making in an emergency such as assisting the pilot in carrying out all the main operating procedures necessary to return the aircraft to operation and perhaps suggesting a creative solution, as well as system 2 in which it would be responsible for processing infinite landing possibilities by crossing aircraft operation data with the system, it is essential for an AI system to be able to suggest possible creative or even conventional based actions, depending on the emergency and the problem.
Finally, considering all the aspects discussed above, we can consider that with technological advances in the field of artificial intelligence and effective computing, we can consider AI as a possible assistant for decision making in emergency situations during aircraft operation, precisely because of the power contribute with factors close to what we have today related to decision making. We still have several obstacles to be overcome and new technologies appear every day in the market for modeling artificial intelligence systems, so we can consider these solutions as a support system, without completely replacing the human factor for decision making.
References
[1] A. R, Damasio. Descartes’ Error: Emotion, Reason, and the Human Brain. New York, NY: Gosset/Putnam Press, 1994.
[2]Kahneman, D.Thinking,fast and slow. New York: Farrar, Straus and Giroux, 2011.
[3]PICARD,R. Affective Computing. Cambridge, Massachusetts: MIT Press, 1997.
[4]Sullenberger, C. Highest Duty: My Search for What Really Matters. Harper Collins e — books, 2009.
[5]Sully, the human factor. Wordpress.com. Available in: <https ://iaragiraldipsicologa.wordpress.com/2017/12/06/sully-e-o-fator-humano/>. Access in: 11-15-2020.
[6] Figure 1 Available in:<https://www.homecinemachoice.com/content/sully-miracle-hudson-blu-ray-review>. Access in: 11–15–2020.
[7] Figure 2 Available in: <https://aviation-safety.net/graphics/infographics/Fatal-Accidents-Per-Year-1946-2019.jpg>. Access in: 11–15–2020.
