Human curation over algorithmic recommendations

Happy new year 2025!

Last year’s big discussion point was definitely artificial intelligence, large language models and recommendation algorithms so let’s start the new year by contrasting all of that with human touch.

I have grown weary about the constant barrage of companies and products pushing computer generated content and recommendations to my face all the time, everywhere. It has created a counter reaction in me to focus even more on the human interactions.

Recommendation algorithms exist for company's benefit, not ours

Corporate social media (Facebook, X, Threads, Instagram, LinkedIn, TikTok, Bluesky and others) is heavily based on recommendation algorithms. It's gotten so bad that these days it's harder to get a list of posts from people you decided you want to follow than it is to get the recommendations.

Instagram stopped showing older, already seen posts a few years ago, claiming to do it for "mental health" to help users avoid endless scrolling. It didn’t take them a long time to fill the new empty space with their recommended feed and ads. So much for the care about mental health.

These algorithms exist to maximise engagement in the platform. And nothing engages more than controversial, surface-level and rage fuelling content. People who support the ideas activate because the thoughts resonate with them and are simple enough to boost and distribute. Those who disagree with the ideas fight back, replying, arguing and trying to tell their side of the story. The more people talk about it, the more it brings people to the discussion and to the platform.

That's generally not the type of discussion I want to see in my life.

I've been very happy in Mastodon ever since I switched over there from Twitter back in late 2022. There I can actually follow people whose posts I care about. And I read my blogs through RSS feeds because that way I can cut the algorithm middle-man from the process.

At the end of 2024, Mastodon started experimenting with recommendations and that got me startled. They started providing profile recommendations as carousels in the feed. Some people enjoy them as they help them find new people, I dislike them because it's bringing the machine recommendations to a space where I prefer to be human curated.

Recommendation algorithms narrow our exposure down

Even if we cut out all the trolling and controversial stuff, I still think recommendation algorithms are a negative thing often enough to be in general something I want to avoid.

There are two reasons for this:

First, these algorithms create even deeper echo chambers. You'll see more of the content that you already see and interact with it. As time goes on, you'll end up seeing mainly things about things that already enforce your interests.

Second, they tend to narrow their recommendations down to the "top performers". Someone who becomes popular in your niche is more likely to be recommended to you than someone who's not well known. This does happen often with human curation too but in machine algorithms there is rarely any force working against this.

Let's say you're into model trains.

The more you interact with model train content in these platforms, the more you'll see model train content and the more you'll be shown the top model train influencers because they gain the most engagement from others whose profiles are similar to yours.

I prefer human curation

I don't want an algorithm, created by corporations with no transparency, to tell me what to watch, read, listen or play.

I'm more and more looking for people who share their suggestions. Whether it's well curated blog or pod rolls, "best movies of the year" lists (by viewers, not by movie media) or delicious recipes, I appreciate the human touch.

Cassidy Williams' post "I miss human curation" resonates a lot with me. Like her, I also want to find my way around the web more organically: from blog to another, directed by the ever-expanding network of people.

I've previously written how I dislike the way services have ruined search by injecting recommendations or nudges to check out something else than what I'm searching and by muddying the perception of where search ends and recommendations start.

I do like the end of year wrap up things (Wrapped, Rewind, Replay and many other names) many services provide. I like them for seeing what I spent my time with but even more, I like that they enable me to have follow up discussions with people as they share theirs. I'm not interested in how many hours someone spent on something but when people share those, it opens up the opportunity for discussion about what media they enjoyed that year and why.

In No 77. of Hello Ruby newsletter, Linda Liukas writes about recommendation algorithms in the context of childrens' content:

While I might buy shoes based on an Instagram ad for myself, I almost exclusively skip everything an algorithm suggests for children, however cute the product might look. Most often, these products are of poor quality, the influencers are removed from the real world, and the result is disappointing.

One thing I appreciate the most about human curation is that it's not limited to things I'm already interested in. I like to explore the world beyond my deepest interests - and within those interests, the world beyond my limited social circles.

A great example of a fantastic curation is from Linda’s Hello Ruby No. 91 where she listed the books she had read in 2024 and categorised them into groups and explained who might enjoy reading those books.

I love getting book recommendations like I did a few years ago when people in our community recommended their favorite sci-fi books. I'm happy to go the extra mile to find authors who are not mainstream best sellers or local celebrities. My natural exposure to any authors or artists from outside US/UK/North Europe is very small and that's a place where human curation shines.

As a blogger always looking for new bloggers, IndieWeb Carnival has been excellent. While it's not one person curating, it's kind of a self-curating thing. I discovered it in January, participated most months myself and found amazing human beings from all around the world through other participants. Since the topics are all over the place, I discovered especially many non-developer blogs that have become permanent residents in my RSS feed.

Coming back to the Mastodon feature, I often hear people talk about how difficult it is to find people to follow in Mastodon. For me, that's quite a non-issue since I mostly follow people I know from outside social media (like blogs, events or IRL) or people who I've discovered in the platform through discussions.

Human curated Starter Packs are one solution, although even they seem to often become kind of top lists of the most popular people on a niche. Anže has been building a starter pack platform for Fediverse and Bluesky has similar thing built in.

Careful curation over information dump

Curation is much more than just long lists of links. The scifi book list from earlier is more of a backlog for myself than curation for others because there's not enough context or personal recommendation involved for it to be very useful to others.

My blogroll is hand picked and manually updated collection of blogs I find great to read with short descriptions of why. I believe it to be way better than dumping my 300+ blog RSS feed into an OPML file and sharing that with people.

I like to curate recommendations for events that I visit if they publish the talks online. Some examples are React Finland 2022, DevRelCon Prague 2022 and 11ty International Symposium on Making Web Sites Real Good. While I find it difficult to not include all the talks, I think good curation requires filtering.

I want to become better at curating myself

If we want to make this happen, we need to be active in curating. I've been trying to improve as a curator over the past year. Barry Hess wrote about this in We are the curators of the web which in part references Manuel Moreale's Curation, search, and the future of the web, both of them great pieces on the topic. Hess says:

In my opinion we must look within for the solution. We are the human beings that are meant to curate the web. We should continue to spread the word about how we use the web, which is very personal and as active participants. We should make clear our desire for technologies that are accessible for non-technical people to be able to join us in controlling their own web presence. We build our websites and share our thoughts and link our links. Over time the web becomes surfable to more and more people.

I want to keep iterating to improve my blog roll this year to be even more usable and valuable. I've been thinking about finding my group of people to form a web ring with. Once a year I restart my Weeklies but then somewhere down the line I fall off from the habit.