Spying on the Wrist Wriggle: What Happens When AI Try to Steal Our Tacit Skills?
AI easily learns data but struggles with “tacit knowledge”: the unspoken habits we rely on at work. To fix this, bosses are tracking workers’ every move. But spying raises massive privacy issues and often makes staff hide their best skills from AI.
Teaching AI to do human jobs is a headache. The main hurdle? We barely understand how we do them ourselves. AI thrives when you give it strict rules or giant spreadsheets. It happily chugs through company databases to figure out how invoices get paid. But work is rarely that neat.
Most jobs rely heavily on unspoken habits, gut feelings, and muscle memory. Experts call this “tacit knowledge”. The philosopher Michael Polanyi summed it up perfectly: we know much more than we can actually say. If you cannot even explain why you do a task a certain way, you stand little chance of explaining it to AI.
Can machines figure it out on their own?
Some tech experts think they can. They argue that our unspoken habits are already buried inside our data. If you feed an AI thousands of old marketing pitches, it soaks up the company’s vibe and judgement without needing a rulebook. It spots patterns we miss. From playing chess to scanning faces, AI has a knack for mastering things we never strictly taught it.
But the physical world is messier, and tiny details matter. You cannot learn who to build a brick wall just by staring at one. A Dutch startup called Monumental wanted to build bricklaying robots. When they asked human builders how they worked, the masons just shrugged and said, “That’s the way I’ve always done it.”
The breakthrough only came when the startup filmed the workers. The footage showed the masons doing a tiny, unconscious wrist jiggle as they pushed a brick into the mortar. That small shake forced the glue into the brick’s pores, making the wall stronger. The masons has no idea they were doing it, but the robots quickly learned the trick.
Should bosses spy on everything we do?
If filming workers solved the bricklaying problem, the obvious next step for companies is to track every tiny thing their staff do. Call centres already do this, recording every conversation and monitoring what agents look at on their screens.
Researchers at the Massachusetts Institute of Technology reckon workers are sitting on a goldmine of undocumented skills. But trying to dig it out is risky. When Meta tried to track staff keystrokes and mouse clicks to train its AI, the workers kicked up a massive fuss.
The MIT study also showed that staff know exactly how much valuable information they keep in their heads, and they are quite happy to keep it hidden from nosy bosses. When researchers told workers their actions would be used to train AI, people suddenly became much less willing to share their data.
Is there a less creepy way to train AI?
Even the most invasive tracking software cannot see inside your head to understand why you made a specific choice. So, instead of spying, some companies use trial and error.
For fuzzy tasks (like deciding if a website looks good or if a research paper hits the mark) they let the AI have a go. Then, human experts grade the computer’s work. Over time, the machine tweaks its approach until it matches human standards.
Companies have always panicked about losing knowledge when their oldest staff retire. But using AI to suck up that wisdom creates a minefield. Who owns the skills inside your head? How much workplace snooping is acceptable? And if machines learn to do all the hard stuff, how will humans ever get the chance to build that unspoken expertise for themselves?
Comments ()