The War Between Technological Productivity and Human Skill

As a kid, Matt Beane, SM ’14, PhD ’17, followed his dad, a general contractor, around job sites, learning by watching as tradespeople passed master-level skills and tools down to talented apprentices.

Beane never lost his fascination with how knowledge and learning are transmitted in the workplace. In fact, he’s made a career out of examining how humans work with machines. His new book, The Skill Code: How to Save Human Ability in an Age of Intelligent Machines, offers a step-by-step guide to understanding how skills are developed (or diminished) in the workplace and how to improve learning now.

It’s an urgent call, too.

In The Skill Code, the assistant professor of technology management at UC Santa Barbara argues that those using artificial intelligence (AI) will become incrementally de-skilled by its productivity shortcuts unless they are consciously upskilling at the same time.

Beane, a Digital Fellow at the MIT Initiative in the Digital Economy, wrote a related opinion piece this June that issued a warning: “We’ve started a war between technological productivity and human skill, and skill is losing.”

While an army of computers isn’t rising up, sci-fi style, to take over the earth (or even every human’s job) in any literal sense, Beane is watching with “grave concern” as we use AI and robotics in ways that eliminate the opportunity for novices to learn from experts and practice skills.

The book builds on Beane’s TED Talk, and Harvard Business Review article, plus years of on-the-ground data collection in industries spanning manufacturing lines and logistics warehouses to robotic surgery theaters and even bomb squads.

In a virtual talk about The Skill Code for MIT Sloan Alumni last month, Beane referenced the unsettling scene from the book in which Kristen, a resident surgeon-in-training, scrubs into the OR and wheels her patient in for prostate surgery. Instead of standing at the table operating alongside her attending surgeon, she hooks up a robot and passively observes the surgery from a console in the corner.

The patient was given top-notch care, and the attending surgeon’s job was undoubtedly faster with the robot’s involvement rather than a resident’s—but was Kristen learning to operate?

“Intelligent technologies in particular, but all forms of automation, allow a single expert to do more valuable work and solve a problem faster or at a higher quality,” Beane explained to the talk’s moderator, Jackie Selby, EMBA ‘21.

While previously, a novice provided essential extra hands for stitching up a patient, combing through mountains of data, or sharing critical situational perspective so a focused master bomb tech could detonate a bomb safely, those previously human roles have evolved into roles for new technology. “If you ask an expert to identify the right time to bring a truly optional novice into their work, somebody who by definition will make mistakes and be slower, the practical answer is likely never.”

The three Cs for skill development

Through his detailed, boots-on-the-ground study of work involving AI and robotics, Beane found “the specific conditions for healthy skill development that are built into the expert-novice bond.” The Skill Code is a deep dive into those conditions, the threats they face, and what we can do to keep skills strong into the 21st century.

Spoiler alert: Beane concludes that it’s mostly not going to be in the classroom. Instead, he presents the critical Cs components for skill development: Challenge, Complexity, and Connection:

• Challenge is “Working close to the edge of your capability on a certain task. It’s close to where you are not performing at your best; you are focused and uncomfortable,” Beane said.
• Complexity is “Digesting the broader universe of skills and roles you’re embedded in so you understand the system, can tackle surprise, collaborate more effectively, and grow into new areas.” For instance, a line worker might look up from the repetitive task at which they are an expert and observe how co-workers 20-feet away package and ship an electric vehicle part to dealerships. When the line breaks down in the future, she’ll likely have suggestions for how to fix it efficiently and might take her next job as a line lead.
• Connection is “about bonds of trust and respect.” Beane shows that this is essential to skill development because, “seeking trust and respect gives us motivation to do better, but also earns access to new opportunities for more skills.”

Using AI to learn new skills—not just to become faster

Beane sees default generative AI, though incredibly useful and not to be banned, as “adding rocket fuel” to the threat that workers can no longer access these three Cs.

To strengthen skill development on the job, Beane challenges managers to prioritize direct, collaborative contact between experts and novices—and he doesn’t necessarily mean a long-term formal relationship. Even if you have six hours with someone in a new junior role, you can think through how to interact with them to complete the task but also leave them with more independent capabilities by the end of your interaction.

Beane suggests we build skill development muscle by refusing to let technology dumb us down. He demonstrates how if you use ChatGPT or Gemini the way it is designed, “it will subtly deprive you of challenge, complexity, and connection the more you use it.” There’s a fix for this problem that doesn’t involve banning advancement. “As the user, you need to go into these programs and add custom instructions to nudge yourself into a healthy skill development and the results you’re after.”

Recognizing that in corporations and educational institutions alike, we are using AI in “predictably unimaginative ways,” mostly focused on automating or accelerating tasks we’ve already mastered, Beane challenges his students to use AI instead to accomplish tasks they would never have thought possible.

For example, he pushed undergrad students to use Python to analyze a giant dataset, visualize it, produce plots, and then post the code on the developer platform GitHub. Although they had little or no previous coding experience and initially balked, they all succeeded. “They felt incredibly capable,” Beane shared.

Their experience echoes his own journey at MIT Sloan. “MIT was extraordinarily difficult. I had to learn new genres of thinking and writing, get familiar with research, and produce work beyond A+ according to the world,” he admitted. Yet, it was also incredibly satisfying in the end—and as he discovered later while researching his book, discomfort in educational settings or on the job is “a strong indicator that you’re doing something right.”

At the end of the talk, Beane announced that he is the new CEO of SkillBench, which he co-founded with MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) alumnus Juho Kim, PhD ‘15. The startup will provide companies with fine-grained, research-driven insight about new technologies’ impact or potential impact.

“We are building a platform for firms to be able to cost-effectively measure the extent to which they are trading off between skill development and productivity as they plan to deploy generative AI,” he hinted.