Tech systems amplify variety and that's a problem

Thursday, December 1, 2022

I recently read "Designing Freedom" by Stafford Beer. It has me thinking a lot about the systems we have in place and something clicked for why they feel so wrong despite being so prevalent. I'm not sure what any solutions look like yet, but outlining a problem is the first step, so let's go.

Systems background

First, some background. What's a system? And what's variety?

A system is a group of components and their interactions. Systems are often used as models for the real world, allowing us to pick out the most important elements and interactions.

Everything you interact with is part of a system and can be modeled as such. A company is a system. You can model it with employees as the components, or you can model it with departments as components. The economy is also a system! You can model it with consumers and companies and the government as various components which interact.

The variety of a system is number of possible states of a system, or of one of its components. Consider monitoring your home's temperature. If you have one temperature probe at your thermostat, that's lower variety than if you have a temperature probe in each room to measure them independently. Similarly, if you have a probe in each room but you average them, the aggregate measure has lower variety than the raw data.

With variety in a system, interactions can amplify (increase) or attenuate (decrease) that variety. We saw one example of attenuation: Taking an aggregate of some measurements attenuates the variety. Likewise, if you decide to add more probes in each room which are not aggregated, that would amplify variety.

Lots of tech amplifies variety

Technology can do either job: it can attenuate or amplify variety for us. The thing is that so much of our technology today amplifies variety. Let's look at it through a few examples.

Global news amplifies variety. It used to be that we could see just what's happening in our local town through a newspaper delivered (maybe) once daily. Now we get a firehose of news from all over the world at a moment's notice. Local news once daily was pretty low variety, and instant global news is almost unfathomably high variety.

Social media amplifies variety. It's pretty clear that social media amplifies variety in much the same way as global news. Most social media apps are structured in a way where you can follow basically any person with a public feed. You can follow any famous person you'd like, and you're encouraged to. Instead of having a local view of your local friends, you get to see this gigantic stream of information from all over the world.

Okay, media amplifies variety. What about our other technology? Let's look at tools we use for work.

Chat apps like Slack amplify variety. Slack encourages you to join a lot of channels, so instead of a small drip of information you get a firehose of it. And with most cultures encouraging open channels by default, there is a lot of information to take in.

Tools like GitHub Copilot or GPT-3-based writing assistants amplify variety. We're seeing an explosion of tools which use GPT-3 to help you write code or write prose. This amplifies variety in two ways. It increases the variety while you're using the tool, because it usually readily shows you suggestions, so you've increased the state of the system while you're using it. It also increases the variety of the system that you're using the tool to contribute into. If you use a GPT-3-based tool to produce documents more quickly at work, that actively increases the variety of the corpus of documents. You write more docs more quickly, so that adds to the firehose again.

Amplifying variety is one side of the coin. It's not inherently good or bad, just a property of the system. But variety in the wrong places in a system can be bad, and can lead to undesirable outcomes. These outcomes can range from mundane (getting overwhelmed by sensor readings) to catastrophic (total collapse of a company).

Tech-amplified variety is causing problems

In this case, I think that the our tech-amplified variety is causing tangible problems today. This amplified variety definitely has a lot of good points. It's great that we have cultures where there's more openness in companies and you can see whatever information you want to see. It's great that we're lowering the barriers to writing code and prose and making those easier. It's great that we can connect with people from all over the world.

It comes with a cost, and it comes with an opportunity. Every problem begs a solution.

The cost of this amplified variety is that we're pushed beyond human limits. The human brain is finite. There is only so much information we can process, only so much we can take in.

Our brains aren't designed to take in these massive firehoses of information.

Here are some of the problems that I experience and see within these high variety systems, along with how I mitigate some of the problems as they affect my daily life:

  • Global news induces anxiety and depression. We see all the problems of the world on broadcast, and we see the good news from only our immediate circles. This imbalance contributes to mental health crises.

    Mitigation: stop consuming global news on a regular basis. We're taught that it's important to consume global news to "stay informed", but practically speaking, there's nothing actionable I can do with this information anyway. I have to disconnect to preserve my mental health. I'm still on the lookout for a way to get news summaries, in context, on a slower cadence.

  • Social media distracts from other tasks. When I was on social media (my vice was Twitter), it was what I turn to for a brief distraction. With high variety, there's always something new and interesting on it.

    Mitigation: get off social media. For me, the mitigation is to go cold turkey off of it. I'd like to find a balance here, and I hope that with the rise of Mastodon, we may see humane social media which isn't a dopamine factory by design.

  • Too many documents, emails, and chat threads to read at work. There's simply too much information produced in even a small company for me to process all of it.

    Mitigation: read a limited subset, and read anything that someone specifically calls my attention to. With limited time, I pay attention to a few "blessed" channels that are highly relevant to my daily work. For the rest, I sometimes skim but usually let it go by unless someone calls my attention to something, which I then go engage with.

One common element of all of my mitigations: They attenuate the variety of the system that's my life.

The problems of high variety come from an imbalance of varieties. The variety my brain needs is much lower than the variety offered by global news, social media, etc. so I have to attenuate those varieties to bring them back to something I can deal with. This is true for systems in general. If a component is experiencing higher variety than it can handle, it's going to experience negative effects. Attenuating variety for that component is the answer. Similarly, if a component can tolerate high variety, you have a bit of waste if you feed it only low variety; it could do so much more.

Big problems come in when variety is left unattenuated and is higher variety than the component it's fed into. I think this is why we see such discord and division in the US, contributed to by social media. But I don't know for sure.

What led us here

Spoiler: it's incentives, it's always incentives. But also, attenuating variety is hard.

We got into this mess because the incentives of our (capitalist) economy lead to prioritizing amplifying variety. And when we try to attenuate it, that's a harder problem, so we can't do it as effectively—especially without resources, because those resource are poured into amplifying variety.

The main incentive in a capitalist economy is making a profit. Right now, the main way that's done through tech is through monetizing your attention.

Anything that's funded through advertising has a clear model: Get you to spend more time in their app, and they make more money.

Other monetization strategies without advertising often end up grabbing attention anyway, though. GitHub has all the dopamine hits on it to get you to use it more, which makes their platform more valuable. They can then use that platform to get enterprise sales and other paid features. They also use that platform's wealth of data to create new products, like GitHub Copilot.

Even products where you pay directly want to keep your attention. Why do brands push their messaging so much into your inbox? To keep your attention so that when you decide to spend money, you spend it with them. And products you're subscribed to, like Netflix, want you to actively use them as much as possible so that you keep paying for them and feel like you got your money's worth.

So, these products are amplifying variety. If it's so nice having attenuated variety, though, why don't we do that? Some consumers would surely pay for that.

The problem is: It's brutally hard. Let's look at the news as one example. If we collect all the news worldwide, that amplifies variety. Now we want to attenuate it to make it consumable without problems. A few ideas for how to do that are:

  • Have a team of writers/editors condense the news down into something intelligible.

    This is likely very expensive (on top of collecting the news, you must pay again to condense it). It can also create the perception of the problem of bias: What gets included and what does not? And any summary will have some human perspective applied for what's important. Doing this with AI isn't a great idea, in my opinion, but that would also create problems with bias.

  • Filter to subsets based on relevance/interest.

    News posts could be tagged with their topics. This is probably already done, as newspapers are organized into sections, and you could also use a topic model to help generate these automatically. This contributes to the problem of filter bubbles, though. It defeats one of the general benefits of being a news-consumer, which is to get broad exposure to more topics. Additionally, it only attenuates in one way by exposing you to fewer topics, but it keeps the variety very high within those topics. So it's an incomplete solution.

In general, I believe it's much easier to create a system that amplifies variety than to create a system that attenuates the newly-created variety. Any form of attenuation is breeding ground for novel new problems, and it's just expensive and hard. Given the difficulty and the incentives at play, is it any wonder that we keep making systems steal more of our attention and amplify variety?

(No, dear reader. It is not a surprise.)

Where do we go next?

Okay, so what do we do?

I don't know. Like I said, it's a hard problem. It's certainly not going to be solved in one schmuck's blog post.

One thing I do know, though, is that we can fight back and we can change the system. Part of how I mitigate the attention-stealing techniques of apps that amplify variety is by practicing mindfulness and being aware of where my attention goes. With that awareness, you can choose to prioritize products (like Sourcehut) which respect your attention over products (like GitHub) which try to take it. You can also prioritize using products like GitHub how you want instead of how the designer wants you to, although that's very hard. They're putting a lot of money into changing your behavior.

The one thing I do know is that by talking about the concepts here and the problems, we can increase awareness. Maybe we can shift how systems are designed. Maybe we can shift how they're used. But we can certainly talk about it, build awareness, and try to collectively come up with solutions.

I think part of the long-term solution is alternative incentives. We see this happening with structures like benefit corporations, which prioritize other incentives in addition to profits. We're seeing a broad global shift toward more focus on sustainability, because consumers are demanding it. We can demand more respect for our attention, and shift the system.

Let's design, build, and buy humane systems which work for us rather than exploiting us.

If this post was enjoyable or useful for you, please share it! If you have comments, questions, or feedback, you can email my personal email. To get new posts and support my work, subscribe to the newsletter. There is also an RSS feed.

Want to become a better programmer? Join the Recurse Center!
Want to hire great programmers? Hire via Recurse Center!