The Cognitive Cost: Calculating What AI Subtracts from Our Skills
Exploring the hidden costs of outsourcing our skills to technology in the age of AI.
With each technological leap forward, we enthusiastically embrace the new capabilities it provides. AI is no different, promising superhuman productivity, tireless repetition, and a generative exoskeleton that enables us to create beyond our innate skills. But as we bask in AI's potential, are we overlooking what it might take away?
This nagging concern was raised by reader and friend Sanjay Castelino in an email last week:
"One of the things I've been pondering the past few days is whether things like ChatGPT are going to reduce the average person's ability to reason.
As I use some of the LLMs I think, I'm just putting in raw facts, the rough positioning that I want, and the AI is spitting out a nice email. I'm not even a sophisticated user. If we fast forward this and the AI was listening to the meeting where I got the action item and we could in fact assign this action to the AI, will we get to the point where the average person loses their proficiency at reasoning?
Much as we lost our ability to write well with ink, will the need to reason for the average business person just go away?
"
Sanjay hits on an important issue - even as AI augments our capabilities, does overreliance on it atrophy the very human skills it is meant to enhance? What starts at first as a handy tool to make things easier starts to mature to a point where we don't just use them for convenience - we lose the ability to function without them entirely.
This question sparked an email exchange between Sanjay and me. We discussed how the rise of GPS technology and turn-by-turn navigation apps has decreased our use of our wayfinding abilities.
Inspired by our discussion, this article will explore this phenomenon further through the lens of past technological shifts. We'll examine how the convenience and efficiency offered by new tools and systems often come with hidden costs—the gradual deterioration of skills that were once essential for everyday life.
How a Guiding Blue Dot Took Over the World
GPS originated from the U.S. military's pursuit of satellite-based navigation in the 1960s. Spurred by the Soviet Sputnik launch in 1957, research into transit satellite positioning led to establishing of the NAVSTAR GPS program in 1973 to develop a global navigation system.
Initially, the GPS signal was intentionally degraded for non-military users through "Selective Availability" - introducing inaccuracies up to 100 meters. This made the technology virtually unusable for turn-by-turn navigation in cities. But in 2000, President Bill Clinton discontinued Selective Availability by executive order, giving civilians access to unobstructed GPS signals and higher accuracy.
This opened the floodgates. GPS rapidly integrated industries. The late 1990s saw the first automotive navigation devices hit consumer markets from companies like Garmin, Magellan, and TomTom - albeit expensive standalone units costing hundreds.
Mass-scale adoption came when Google Maps was released. Two years later, they dropped a bombshell - free voice-guided turn-by-turn GPS navigation on Android, starting with the Motorola Droid in October 2009.
According to The Wall Street Journal, Garmin's shares plummeted 16% the next day, while TomTom's dropped 20%. Consumers no longer needed pricey dedicated devices when their smartphones could navigate with a simple, intuitive blue dot.
Getting from A to B became as easy as following the guiding blue dot to the red pushpin on the map. Physical maps and memorizing routes were rendered obsolete by a magic blue orb that could effortlessly chart courses through the world's roadways.
The Cognitive Cost: Use It or Lose It
In the era before GPS, visiting new places required active preparation and engagement. We had to consult physical maps, think through routes, and commit key turns and landmarks to memory. Getting from A to B was a cognitive exercise that kept our spatial reasoning and mental mapping skills sharp.
But in today's digital world, we've outsourced that mental labor to our smartphone's magic blue dot. Simply follow the turn-by-turn directions, and you'll arrive at your destination without any real spatial awareness of how you got there.
This cognitive outsourcing comes at a price. The part of the brain associated with spatial memory and navigation - the hippocampus - is also deeply linked to the ability to form memories. Some anthropologists have even suggested that navigation needs were the evolutionary starting point for memory itself.
Mnemonic techniques like the "memory palace" rely on visualizing intricate mental maps of locations to store and recall information. And the arduous process of becoming a licensed London taxi driver, with its notoriously difficult "The Knowledge" test on the city's 25,000 streets, has been shown to physically alter drivers' brains.
In a 2000 study, researchers found that successful London cabbies had significantly larger hippocampi than non-drivers of the same age, as their brains adapted to the incredible spatial skills required for the job. Just as a muscle atrophies from disuse, the cognitive skills required for exploring and understanding the physical world around us deteriorate when we rely too heavily on technology's virtual maps and directions. A Harvard study in 2020 revealed that regular GPS users exhibited poorer spatial memory, landmark recognition, and cognitive mapping abilities compared to non-users. The more they outsourced navigation, the steeper their skills declined.
The convenience afforded by GPS comes at the cost of keeping our inner compass calibrated.
When Automation Bites Back
The cost of losing navigational skills goes beyond individuals getting lost on a hike. Navigation is a critical component for jobs like airline pilots. As cockpit automation proliferates, pilots risk losing the ability to manage the aircraft when technology fails or malfunctions.
Captain Warren VanDerburgh, a veteran American Airlines pilot, warned about this overreliance back in 1997. In a training presentation, he coined the term "Children of the Magenta Line" to describe pilots blindly following the magenta course lines on navigation displays without analyzing raw data or having backup skills if technology failed.
Vanderburgh's warnings tragically manifested in the crash of Air France Flight 447 on May 31, 2009. The Airbus A330 was cruising at 35,000 feet from Rio to Paris when iced-over pitot tubes caused conflicting airspeed data. The automated systems disengaged the autopilot, unexpectedly handing control to the human pilots.
Suddenly, the pilots were wrestling an aircraft they had minimal experience manually flying. The captain had logged 346 hours over the past six months, but only around four hours were actually hands-on without the autopilot engaged (during takeoffs and landings). Unable to understand what was happening as stall warnings blared, the pilots could not recover before crashing into the Atlantic, killing all 228 aboard.
But skills erosion isn't the only issue. The complexity of automation we've built to make the aircraft fly can exceed our ability to understand and correct when they fail.
Take Boeing's 737 MAX aircraft. To accommodate larger engines without requiring pilots to retrain, Boeing designed a new MCAS (Maneuvering Characteristics Augmentation System) system to automatically push the plane's nose down in certain conditions.
But what if it malfunctions? As it did with Lion Air Flight 610 on October 29, 2018. A faulty sensor caused MCAS to repeatedly push the 737 MAX's nose down, even in level flight. The pilots struggled in vain for 11 minutes to regain control, unable to counteract MCAS's automation before crashing into the Java Sea, killing all 189 on board.
The thought of human pilots grappling against their own aircraft's automation as it relentlessly pitched them toward the sea is a haunting image of technology's potential for unintended consequences.
Automation's Invisible Tendrils
The challenges of automation dependency extend far beyond the cockpit. As complexity increases across domains, our ability to fully understand and manage these large systems is being outpaced. To keep up we have to automate roles at every node of the intricate networks that power modern society.
Consider air travel - flying the plane itself is just one node in a vast, globally coordinated network. Each day, over 100,000 flights arrive and depart from the world's 40,000+ airports. Successfully executing this choreography requires precisely navigating runways, airspace, other aircraft, and coordinating with air traffic control in real-time. The complexity is staggering, driving a relentless march toward automating every aspect of air travel operations.
But it's not just the individual nodes being automated - it's the interactions between them as well. Take the stock market as another example. Human traders simply cannot process information and execute trades fast enough to capitalize on micro-pricing movements. As a result, computer algorithms now dominate the equity markets, with a 2020 SEC report finding 78% of all trades being performed by automated trading centers.
As these complex, inscrutable systems interact, their ability to behave in unexpected ways compounds. On May 6, 2010, the world witnessed this firsthand when a "flash crash" temporarily wiped out $6 trillion in market value in just 36 minutes. While the precise catalyst remains unclear, one theory is that various trading algorithms interacted in an unanticipated feedback loop, sparking the freefall.
Our response? Implement more automation to control the automation. Officials announced new automated "circuit breaker" rules to halt trading for 5 minutes on any S&P 500 stock experiencing a 10% price swing in a 5-minute window. A band-aid solution attempting to cool the potential for future flash crashes emerging from a system we don't fully understand.
Co-Evolving with Our Creations
Throughout this article, we've examined the potential for technology to erode skills, from spatial navigation to piloting aircraft. But recognize that our relationship with technology is not a one-way street of skill decay. Just as automation can diminish certain abilities, it also provides us with new opportunities and ways of interacting with the world.
Take GPS, for example. While overreliance on turn-by-turn navigation may reduce our need to exercise traditional mapping and wayfinding skills, it has also given rise to new outdoor activities like Geocaching. In this global treasure hunt, participants use GPS coordinates to hide and seek containers called "caches" in locations all around the world. By combining technology with exploration, Geocaching enables people to discover new places and engage with the environment in novel ways.
This co-evolution with technology is nothing new for our species. The advent of agriculture thousands of years ago undoubtedly eroded some of our hunting and gathering skills. But it also freed up time to develop new pursuits like art, writing, and complex social structures. Each major technological shift brings trade-offs, enhancing certain abilities while diminishing others.
However, the reality is that we are becoming increasingly intertwined with and reliant upon the automated systems we create - and the pace of this entanglement is rapidly accelerating. As AI advances, the scope of skills and cognitive tasks we outsource to machines will only continue to expand. Should we ensure that crucial human capabilities don't atrophy in the process? If so, how?
There are no easy answers, but actively grappling with these challenges is likely the way to the answer. What do you think?