The importance of friction

With the glut of online articles about how generative AI tools like ChatGPT are making us dumber (which, as I said in a previous column, isn’t precisely true), an interesting conversation started around the role of friction in our lives as technology makes things easier and easier for us.

Friction, in the physics sense, is a force that happens between two surfaces as they slide or roll against one another. Friction is a force you need to overcome if you are sliding a piece of furniture across a floor. Friction generates heat which is why you can (with a great deal of effort) start a fire by rubbing two sticks together. Friction helps your fingers grip your mug and prevents you from dropping your precious morning coffee all over the rug.

Metaphorically, friction can mean anything in our lives that makes something more difficult. In this sense, technology has greatly reduced the amount of friction in our day-to-day lives. The electric light made illuminating our living spaces easier and safer. The internal combustion engine made journeys that previously took days achievable in hours. Email allows us to communicate with the other side of the world in seconds rather than the weeks or months it would take to deliver a physical letter. Before these innovations, those tasks were much more difficult taking more time, energy, and effort.

However, the removal of all friction from our lives isn’t necessarily a good thing. Working through a process with friction, such as learning about a new topic, can be where we gain new insights. Reading widely, running into dead ends and conflicting information, and puzzling through a problem that’s difficult to understand forces us to think critically and generate our own ideas. This cognitive friction can help us develop our thinking and deepen our knowledge.

AI powered chatbots acting as answer machines, however, smooth out the cognitive friction that comes with learning something new. When prompted to summarize a topic, a chatbot will spit out whatever echoes of that topic are in its training data, mimicking the synthesis that comes with research and learning within seconds. While words may have been produced quickly, for the person using the tool no new ideas or insights have been made. This is backed up by the MIT study earlier this year by Kosmyna, et al that measured brain activity while using LLMs – the least amount of brain activity was seen in the participants who used only an LLM for the assigned task, and recall was the lowest when asked to summarize their learning.

Friction when learning something new can also help build resilience. There’s a somewhat pop-psychology term known as the “trough of despair” which has been applied to a number of fields including learning and change management. The trough of despair occurs when you’re learning something new and are overcoming the feeling that you’re never going to “get it.” The trough of despair is filled with confusion and frustration but is ultimately temporary. With some effort and perseverance, it is usually followed by an increase in understanding and competence. Getting through the trough of despair, knowing that despite feeling like you can’t figure things out but ultimately will, makes enduring that phase the next time you learn something new easier.

While friction in our lives can be important, it doesn’t mean we want needless friction. This isn’t a “when I was in school I walked uphill both ways through the snow, and so can you” situation. We want the tools we use to work well, safely, and without unnecessary barriers. Efficiency can still be a goal, while recognizing that learning things and learning them well requires an investment of time that may feel frustratingly inefficient. Going back to MIT and Kosmyna, et al, this doesn’t preclude the use of generative AI in the learning process (the group that went from brain-only to furthering their work with an LLM outperformed all groups in their study). But do I think it makes the case that ChatGPT should probably not be the first stop on the path to learning.

A little bit of friction in our lives can be a good thing. It can teach us to move through difficulty and discomfort in the pursuit of a goal. It can help us generate new ideas and better hold on to the knowledge that we’ve gained. So, when we’re looking at the technology we use that reduce the friction in our lives, we should ask at what cost.

Learn more

Friction (Wikipedia) Last accessed 2025/09/16

Friction Is The Point: What AI Will Never Understand About Being Human 2025. Jason Snyder (Forbes) Last accessed 2025/09/16

When the mind stumbles, it grows 2025. John Nosta (Psychology Today) Last accessed 2025/09/16

How to use AI without losing our minds 2025. Ngaire Woods (Manila Times) Last accessed 2025/09/16

Impact of AI on PhD research and learning 2025. Morten Rand-Hendriksen (TikTok) Last accessed 2025/09/16

Disappearance unclear question 2025. Victoria Livingstone and Jeppe Klitgaard Stricker (UNESCO) Last accessed 2025/09/16

Correction log

Nothing here yet.