Teach A Man To Fish
When can cognitive atrophy be a good thing?
We’ve all heard the proverb: “Give a man a fish, feed him for a day. Teach a man to fish, feed him for life.” As we continue into an AI-powered future, this ancient wisdom has never been more relevant. Are we creating AI systems that feed us fish, or ones that teach us to become better fishers?
Consider GPS navigation. While undeniably convenient, our increasing reliance on turn-by-turn directions is eroding our natural wayfinding abilities. This “cognitive displacement” — where technology gradually replaces rather than enhances human capabilities — isn’t inherently bad. After all, the industrial revolution gave us vacuum cleaners and dishwashers, freeing us from mundane physical labor. But as AI moves beyond physical tasks into cognitive domains, we need to think carefully about what we’re willing to outsource.
A tweet I recently encountered captured this tension perfectly: paraphrased, “I don’t want AI to do my art or writing — I want it to do my laundry and dishes.” This sentiment reflects a crucial distinction: there are onerous tasks we want to delegate entirely and others where we want AI to stay away — or possibly help us do better. The challenge lies in where to draw that line.
I propose a simple litmus test for human-centered AI design: Is this a human activity that enriches our lives, or is it one that is tedious, onerous, or artificial? If it is the latter, then I welcome automation. If it is the former, hands off. Writing poetry, making music, or even navigating our environment are deeply human activities that contribute to our cognitive development and emotional well-being. These are areas where AI should augment rather than automate — teaching us to fish rather than serving us fish.
And what’s more, where this line is drawn can be very individual. Some of us enjoy writing, whereas others see it as a necessary evil. Cooking, laundry, or housework can be therapeutic for some, and an abomination for others.
This also mean that we should not reject AI assistance. For one thing, there are areas where I am perfectly fine having my cognitive abilities atrophy entirely: doing my taxes, memorizing phone numbers, or learning obscure keyboard shortcuts. But for the skills that are indeed worthwhile, we should design AI systems that scaffold human learning and development. Imagine navigation apps that gradually teach you to recognize landmarks and understand spatial relationships, or writing assistants that help you become a better writer rather than just generating text for you. This is the true promise of human-centered AI: tools that make us better versions of ourselves, not just more efficient consumers of technology.
The industrial revolution freed our bodies from repetitive physical labor. The AI revolution — if it deserves to be called this — promises to free our minds from repetitive cognitive labor. But we must be careful not to let it also take away the very mental activities that make us human. As we design the next generation of AI tools, let’s focus on teaching people to fish — creating systems that enhance human capabilities rather than replace them, that augment rather than automate, that empower rather than supplant.
