Features

Man and machine

Chris Sheedy
Photo: digitalvision / Thinkstock

Air traffic controllers are helping researchers build intelligent systems that seamlessly blend with the human mind. 

In an air traffic control tower in France, a man watches electronic blips move across a screen, each dot representing a plane in flight. His job: to safely guide the planes across the crowded airspace to their destination.

It’s work where a single lapse in concentration, a minor miscalculation or moment of misjudgement, could result in the deaths of hundreds of people.

But what if the controllers could super-charge their thinking by merging their brain with a machine that looks after, in real time, every calculation, forecast and analytical problem they face?

Their brain’s “bandwidth” would then be freed up to look at the things that humans excel at – spatial awareness, holistic judgement, strategising and understanding human nature. The air traffic controllers at the Eurocontrol Experimental Centre in Bretigny, France, have been participating in UNSW-led research that could lead to that outcome.

The controllers were put through their paces in a simulated environment with multiple sensors attached to and around their scalps to monitor brain activity during typical air traffic control challenges. The research was led by Hussein Abbass, professor of information technology at UNSW Canberra.

“Throughout my career I have been motivated to build an intelligent system that seamlessly blends with the human being,” Abbass explains.

“Usually when people think of such a coming together, the machine is one component and the human is another. My specialisation is artificial intelligence so I want to see the two morphing into one. It is about what I call Cognitive-Cyber Symbiosis, or ‘CoCyS’ (pronounced ‘cookies’).”

Far-fetched? In fact CoCyS is far closer to reality than you might think, Abbass says. In certain industries such as education and training he and his team are ready to roll out systems that will work in unison with professionals to make them more effective and efficient.

Researchers at UNSW Canberra are now analysing the masses of data gleaned about aircraft movements, distance, altitude, speed and weather effects and then combining them with data from the human controllers, all with the objective of discovering how machines might complement the human brain to create better outcomes and a safer environment. A paper detailing the study is published in the Journal of Air Traffic Control.

Abbass admits there is still a huge amount of “noise” in the data collection process. For example, if you are speaking with somebody and they are distracted and look away, even for a split second, how do you figure out why they are doing that or what caused the distraction?

Modern-day computing and its ability to collect and analyse big data makes it possible to answer such questions, as long as enough data are collected.

“We build models to identify the causes of change and steer the environment towards the desired outcomes,” Abbass says. “If you do it for long enough you can develop confidence in what the data are telling you.

“From a single sensor on a person’s skull – and we can go up to 128 if needed – we receive and analyse 2,048 readings per second.”

What can a machine do more quickly, reliably and accurately than a human? What can the human do better if they have a machine doing some of the thinking for them? This is what Abbass is looking into, and its application is far greater than in the field of air traffic control.

“In almost every industry we need to pay attention to the human and give people the technology and the support to make the right decision every time,” says Abbass, who is keen to partner with researchers in cognitive psychology to expand the application of his work.

“This technology can help people do that in real time. And when things go wrong, as they sometimes do in aviation, it can be used to explain why.”

Professor Abbass’ book Computational Red Teaming: Risk Analytics of Big- Data-to-Decisions Intelligent Systems is published later this year by Springer.