This week’s Polymathmind.ai Substack article from Adam Riley and myself points to Malone’s finding that AI-human teams often underperform the best of either alone. The question this raises is a useful one: when does AI augmentation actually work, and when does it degrade thinking?
We reference research from Lebovitz, Lifshitz-Assaf and Levina, that distinguishes between engaged and unengaged augmentation.
In engaged augmentation, the professional interrogates the AI’s output. They relate the AI’s claim to their own knowledge, judgement and context. They may overrule it, reflectively agree with it, or synthesise something new. The AI sharpens the thinking; the human stays in charge of it.
In unengaged augmentation, the AI’s output is accepted without that interrogation. The human is in the loop on paper. In practice, the AI has done the thinking.
What separates the two is not the technology. It is the human capability being applied.
The capabilities that produce engaged augmentation are the ones we have been describing as Power Skills for three years: contextual sensemaking, abductive reasoning, the discipline to challenge a fluent answer that looks finished. They are the difference between AI as scaffold and AI as substitute.
Most organisations are investing significantly in the technology. Far fewer are investing in the capabilities that determine whether the technology pays off.
Where in your organisation is augmentation engaged – and where has it become something else?
Read full article here…
