I grew up thinking that on August 29th 1997, an automated system called “Skynet” would destroy all life on Earth. Whilst the apocalyptic predictions from the movie Terminator may not have happened after all, artificial intelligence certainly has. From the self-driving cars of Tesla, to the algorithms tailoring specific advertisements to your web browser, today’s automated systems are more clever, agile, intuitive (and intrusive) than ever before.
Born out of a need to satisfy the skills gap left by an ageing and shrinking population, early this year a large Japanese insurance firm announced that it was replacing the functions previously performed by 34 employees with IBM’s revolutionary Watson Explorer AI. It’s an incredible achievement and a giant leap in automated systems that will forever change the industry in perhaps the same way that the autopilot system revolutionised the airline industry. But whilst we continue to be amazed by automation and the opportunities these technologies provide us with, should we be so quick to adopt these systems without considering the broader impacts?
In order to predict how increasingly automated systems will affect our working lives in the future, we need to first learn the lessons of the past. In the 1950’s, Professor James Bright studied the human impact of automation in industry. Despite being heralded as a way of upskilling employees and boosting productivity by removing routine manual tasks, Bright found that automation was actually deskilling and demotivating employees. He concluded that the positions no longer required skilled operators as the “skill” could be built into the machine. As a result, skilled employees became bored and their skill level inevitably withered over time. Evidence is mounting that AI and automated systems are now having the same deskilling effects on the workers of today, including highly trained professionals, managers and specialists.
In 2007, British aviation researcher Dr. Matthew Ebbatson theorised that airline pilots were experiencing “skill fade” as a result of an over-reliance on autopilot. Ebbatson conducted experiments in a Boeing flight simulator where he asked pilots to perform a variety of manoeuvres under different conditions typically performed by the autopilot. The results of the experiments led Ebbatson to conclude that “Flying skills decay quite rapidly without relatively frequent practice”. As the standard industry indicator of an airline pilot’s competence is measured in “hours of flight” rather than “active flight time”, and given that up to 85% of flight time is controlled by the aircraft’s autopilot system, it’s hard to believe that airline pilots are getting “frequent practice” they require.
Minor skill fade in airline pilots can be disastrous. Autopilot-related pilot errors have been implicated in several recent air disasters, including Air France Flight 447 in 2009. The pilot incorrectly responded to system prompts after the autopilot was disengaged due to iced up wind speed sensors, ultimately sending the Airbus A330 into a stall before plunging into the Atlantic. The conclusions from the investigation of this incident prompted the Federal Aviation Administration to issue a warning that pilots “have become accustomed to watching things happen and reacting, rather than being proactive” and urged pilots to spend more time flying “by hand”.
Whilst the importance of Ebbatson’s findings cannot be overstated, they should not be viewed in isolation to the airline industry. Physicians can miss critical details when they rely heavily on software prompts to guide them through a patient exam (rather than following the patients narrative thread). Architects and designers can lose the aesthetic sensitivity of their craft when they rely heavily on computer aided design over more traditional, tactile methods. Engineers using automated systems can mistakenly allocate resources to the symptom of an asset fault, whereas a trained plant engineer would have first identified the problem through a root cause analysis.
Our skills get sharper when we regularly use them to overcome a diverse range of complex challenges. However, the primary goal of automated systems is to eliminate the need for us to be faced with this complexity, leaving us to complete passive, mundane tasks such as monitoring and data entry. Whilst it all sounds bleak, the choice between improving our skillset and increasing our exposure to automation doesn’t have to be mutually exclusive. In his book “The Glass Cage: Automation and Us” Nicolas Carr introduces the concept of “human-centered automation” Where the talents of people take precedence. Systems are designed to keep the human operator in what engineers call “the decision loop”—the continuing process of action, feedback and judgment-making. That keeps workers attentive and engaged and promotes the kind of challenging practice that strengthens skills.
Machine learning algorithms and augmented reality have given birth to a significant development in the human-centred approaches known as adaptive automation (AA). AA monitors people’s physical and mental states and makes decisions to shift tasks and responsibilities between them and the computer. For example, if the system senses that an operator is having difficulty with a procedure, the system intervenes and reduces the workload to allow the operator to focus. When the system senses that the operator is paying less attention or becomes bored, it ramps up the person’s workload to capture their attention and build their skills.
It’s natural to be amazed by the development in automation and the power of machines, but we must also take care not to underestimate our own talents. If we continue to allow our own skills to fade by relying on machines to do our work, we are going to become far less capable and competent, resulting in a world better suited to machines than us.
Written by Sam Byrnes, Coriolis Ltd
Nicolas Carr “The Glass Cage: Automation and Us.”
Wall street journal – Automation makes us dumb