On the Need for Friction

Imagine talking to a medieval farmer, about the concept of excercise. Giving yourself “useless” physical strain, to improve your health? Let’s call this idea artifical physical excercise. That wouldn’t make any sense to someone who would get more than enough strain through just living their life.

And while it will vary greatly, from person to person, how much artificial exercise is needed, everyone agrees that we need to look after our physical health in today’s society.

When discussing this, people will mention which of these they enjoy the most, which they find effective, how to fit it into their lives, etc. Some of it are games, competitions, sports, and more – and it can be your job, a favourite pastime, a hobby, or just something you tolerate.

It can also be adjustments you make to your life, like riding a bicycle to work, or changing your desk setup. Let’s call this incidental physical exercise, in opposition to deliberate physical exercise.

But, in this context, this is my main point about this: It’s very accepted to talk about doing things for your physical health, even though it might not be the most comfortable, fun, or easy.

And similar to how technological improvements increased the need to look after the health of our bodies, it has now made it important to look after the health of our minds as well.

Learning and AI

If you’re at the gym, there are many examples of how technology can enhance the effectiveness of our artificial physical exercise. However, using a forklift to lift weights might be more effective and comfortable, compared to doing it yourself – but it also makes the action completely useless! The point isn’t that the weights get lifted, but that you do it. This is in contrast with a warehouse, where the point is to get the stuff lifted.

One framework here, could be the difference between a tool and a machine. The former will enhance the user, while the latter replaces it. And when using AI, specifically for learning, some discipline is required to make sure that it remains a tool, and doesn’t slip into becoming a machine. This is one reason why I’m highly skeptical of giving young students unfettered access to LLMs.¹

A custom LLM implementation, with guard-rails that helps with this discipline, could be great for students, though. And LLMs also have great potential for teachers, as a machine can be more useful when the main point isn’t exercise.

It’s quite similar to the way we think about calculators in mathematics didactics. They’re obviously useful, and are a tool students need to learn to use. But they have to be used consciously. And when training your thinking, and establishing core understanding, they can be problematic.

Articles and videos with titles like “Junior Devs Can’t Code Anymore” are popping up, and give an interesting example of a border-case: In this case we are in a professional environment, where the work getting done is an important point. But over-relying on a tool, that might not help you develop as a professional, especially early in your career, can be quite dangerous.

As someone who only dabbles in programming, LLMs has made it possible for me to do and learn much more than I could’ve previously. But I’m trying to be conscious about it, by trying to do it myself the next time, really understand the core concepts and code I get in return, etc.

Cognitive health

I want it to be just as common to talk about doing things that are good for your cognitive health² as for your physical health.

One example of this, was a recent video, by the beautifully nerdy YouTube channel Technology Connections, called Algorithms are Breaking How We Think. He’s talking about the value of friction, and I think it explains why we should be careful not to always do things the easiest way. This can be compared to taking the stairs even though the lift is easier – and would be an example of:

Incidental cognitive exercise

One of the problematic aspects of feeds like TikTok’s (and everyone wants to make those nowadays) is that they remove any kind of friction. And I don’t think it’s a good thing that we’re getting used to just having the apps serve us something, without making any choices. “I want to listen to music. Just give me some, Spotify – I don’t care what.” This is neither great for the arts nor our minds, I think.

I’m not saying that there are no good things about algorithms, and that we shouldn’t ever use them. For instance, discovery algorithms can be great to find new people. But it’s a bit like how it’s no problem for Jane Doe to always take the lift at work, if she’s very active in other parts of her life. But I, being a lazy slob, shouldn’t always default to the lift.

Relying less on algorithms isn’t only good for your mind – it also takes back some power from the platforms. If I follow a page on Facebook, it only nudges the algorithm to give me their content – Meta still decides. However, if I follow a blog via RSS, I will get the things they write – because I made a choice.

Junk food for thought

Incidental cognitive exercise is about making choices for “the future you”. But it does require some more effort to curate a Mastodon or RSS feed, as opposed to just opening an app that always gives you unlimited amounts of stimuli.

I think it makes sense to look at the hyper-algorithmic feeds, especially for video (TikTok, Reels, YouTube Shorts), as junk food for the mind. And that’s why I’m so annoyed when people over-emphasise “revealed preference”:

Placing this option in the context of Facebook and Instagram actually suggests that this feature won’t matter very much; both services make it hard to find, and revert back to the default algorithmic feed, and for good reason: users may say they want a chronological feed, but their revealed preference is the opposite.

— Ben Thompson / Stratechery

If I had the choice between a salad and a pile of McDonald’s, my stupid brain would want the burgers and fries every day.³ But I wouldn’t use that as proof that McDonald’s “is better”.

I don’t subscribe to every argument of Jonathan Haidt’s The Anxious Generation (some of which can be heard in this Hard Fork episode.) But if we gave our kids free access to McDonald’s, 24/7, how do we think that would affect their physical health? Would they be equipped to regulate that? And how often would they be too full to bother with the healthy dinner their parents have made?

I know that I can’t have candy in the cupboard, and that I’ll eat better if I make sure to have healthy snacks available. Doing this kind of effort can be compared to curating some (finite) feeds, like Mastodon or RSS, so that mindless scrolling isn’t the only decent option. Other things could be to charge your phone away from your bed (and keeping some good magazines and books close instead), and to only have good games on your devices.

Friction is underrated

There are many great parts about modern technology – including algorithms. It’s much easier to just say: “Let’s throw it all out!” But I think we need to strive towards keeping the good parts, like how social media can help to connect us, and LLMs can be great tutors, without being naïve regarding the negative.

And similarly to how it’s very common to discuss artifical physical excercise (both deliberate and incidental), to improve our physical health, we need to be more conscious about our cognitive health. I want more conversations about cognitive excercise – deliberate and incidental.

It’s very natural for us to want to avoid friction, of any kind. But the simple fact is that it’s good for us! And when technology has removed much of both the physical and mental strain, we have to add some of it back.