The Case for Human Curation in the Age of Algorithms
Updated
There's a question worth sitting with: when was the last time you discovered a piece of music that genuinely surprised you? Not a variation on something you already liked — a lateral step into adjacent territory — but something that opened a door you didn't know was there?
For a lot of people, that experience has become rarer. Not because music has become less interesting. Because the systems that deliver music to us have become very good at knowing what you'll accept, and optimising for that.
What Algorithms Are Actually Optimised For
Algorithmic recommendation systems are engineering achievements. The behavioural data processed to predict your next listening choice — play counts, skip rates, listening duration, time of day, device type, geographic location — is genuinely impressive. These systems work.
But "works" needs a definition. Recommendation algorithms are optimised for engagement: keeping you listening, preventing skips, bringing you back tomorrow. That means they're rewarded for giving you music that resembles music you've already shown you like.
This is not a conspiracy. It's the design. An algorithm that challenged your taste aggressively would produce more skips — all the metrics going the wrong way. So the system learns that the safe move is to stay close to existing tastes.
The cumulative effect is a kind of contraction. Your musical world becomes more accurately mapped — the recommendations more precise — but also, quietly, smaller. Research on recommendation systems and filter bubbles suggests that heavy algorithmic reliance tends to reduce breadth of consumption even as it increases satisfaction with individual choices.
You're more comfortable. But the territory has shrunk.
What Human Curation Is Actually Doing
A human curator — a radio music director, a record shop buyer, a magazine editor, a playlist compiler with a genuine point of view — is doing something structurally different. They are not optimising for your engagement. They are expressing taste.
That distinction matters more than it might initially seem.
A music director at a jazz radio station plays a record because they think it's good, because it fits the hour, because it says something in the context of what came before it. The judgment is aesthetic, not algorithmic. It accounts for things that data cannot easily capture: the feeling of a particular track in late evening, the way a lesser-known artist's work sounds different when placed next to a more famous peer, the editorial instinct that a certain genre needs more space in the programme this week.
Human curation is inherently opinionated. And opinion is what makes it useful for discovery. When a curator puts something in front of you that you didn't ask for and didn't know you wanted, the implicit message is: I think you should hear this. That's a different relationship than you've indicated a preference for this type of content.
The serendipity of radio — particularly of well-themed radio — comes directly from this. You didn't search for it. You didn't seed it. A person decided it belonged there, and you happened to be listening. That asymmetry is where genuine discovery lives.
The Difference Between Discovery and Confirmation
There's a useful distinction between two kinds of music experience: discovery and confirmation.
Confirmation is what algorithmic recommendations mostly deliver. You signal a taste; the system extends it. The loop is tight and satisfying. Discovery is what happens outside the loop — a friend plays something at dinner, a radio programme goes somewhere unexpected, a station you've never heard of is on in a café. You're just listening, and something is happening.
In an environment where almost all music delivery is algorithmic, confirmation crowds out discovery — it's more reliable and comfortable. Discovery requires tolerance for the unfamiliar, which recommendation systems are designed to minimise.
Human curation protects space for discovery. It makes an editorial bet that you'll find something valuable in music you wouldn't have chosen yourself.
Why Radio Remains the Most Honest Format
Radio was built on human curation before any other model existed. The programme director, the music director, the DJ — these people made choices, expressed taste, took editorial positions. Radio history is full of people who changed music culture not because an algorithm flagged something as popular, but because they believed in it and had access to an airwave.
Internet radio inherits this tradition while extending it globally. The stations you discover exploring radio by country or genre exist because someone decided to run them, chose what to play, and committed to a sound. That commitment — that point of view — is what makes a station feel like something rather than a data set.
The collections on 72FM work on the same principle. A collection like Women in Jazz or Late Night R&B isn't a metadata cluster. It's an editorial decision: these stations belong together, they share a sensibility. That decision was made by a person.
A Note on What Algorithms Do Well
Algorithmic recommendation does things human curation genuinely cannot. No human team can personalise recommendations to hundreds of millions of users simultaneously. Algorithms also democratise discovery in ways manual curation historically didn't — a small artist with no industry connections can surface in a recommendation feed if their music resonates.
The case for human curation is not that algorithms are bad. It's that algorithms alone produce a narrower musical life than a combination of algorithmic and human-curated listening does. The two belong together.
The Practical Upshot
If your listening has started to feel circular — if you can predict what a recommendation system will give you — that's a signal worth paying attention to. It might mean the system is working exactly as designed, and what you actually want is something a system isn't designed to give you.
That's when themed radio tends to feel most valuable. Not as a replacement for the music you love, but as a window to music you haven't found yet — built by people with opinions and the conviction that some things are worth hearing even if you didn't ask for them.
Browse the themed collections or read more about why discovery and algorithmic playlists answer different questions.
Summary
Algorithms are genuinely impressive. They know what you like, and they're good at giving you more of it.
But "more of what you like" and "discovering what you might love" are different things. Human curation — the editorial instinct, the committed point of view, the bet that you should hear something you didn't ask for — does something that optimisation systems aren't built to do.
Radio, at its best, has always been this. A person behind a programme, choosing music not because data predicted it would work, but because they thought it was worth hearing.
That tradition is still alive. And it's still one of the most reliable routes to the music that changes your taste.