By Luminae | Fact & Friction | Season 1, Episode 2
The moment curiosity stops feeling like a choice
You open a platform to watch one short clip. The next recommendation looks useful, funny or just unusual enough to deserve another minute. Then another appears. Twenty minutes later, you know far more than you intended about a topic you did not set out to explore – and you may no longer remember why you opened the app.
That experience is often described as falling down an online rabbit hole. It feels like curiosity doing its ordinary work: one question leading to another. But curiosity is only part of the explanation. The path is also being shaped by a recommendation system deciding what to place in front of you next.
What recommendation systems learn from
Recommendation systems use signals from behaviour. Depending on the service, these can include what you click, how long you watch, where you pause, what you replay and which topics repeatedly hold your attention. The system does not need to understand your reasons in the way another person would. It needs patterns that help rank the next likely item.
This is why a feed can feel uncannily responsive. Each action supplies another clue. A recommendation that holds attention can influence the recommendations that follow, creating a feedback loop between what the system offers and what the user does.
It is important not to turn this into a story about a single all-powerful algorithm. Platforms use different systems and objectives, and a recommendation can be helpful. The practical issue is that engagement incentives can make continued attention a powerful measure of success even when the user would have preferred to stop.
Filter bubbles are only part of the story
The familiar filter-bubble account says that personalisation repeatedly shows people material that fits their existing preferences. Echo chambers add a social dimension: communities can reinforce shared views and dismiss outside ones. Both ideas can be useful, but neither completely describes the movement of a rabbit hole.
A rabbit hole is dynamic. It is not necessarily a room filled with the same opinion. It can be a sequence in which each item feels like the logical next step while the overall journey becomes narrower, more novel or more emotionally intense. The system is not required to have a political destination in mind. It can simply keep selecting material that appears likely to sustain engagement.
Why the journey can become more intense
Novelty, surprise and emotion can hold attention. In the episode, Harry describes watching videos in which people apparently stealing parcels are met by exploding packages. The first clip produces a quick emotional response. Later clips offer stronger retribution, confrontation and eventually scenes that appear staged or implausible. The sequence still feels connected, but the character of the material has changed.
This is a personal example, not proof that every feed follows the same path. Its value is explanatory: the next recommendation can feel natural at the level of one click even when the whole chain leads somewhere the user never intended to go.
The same pattern can matter beyond entertainment. Repeated recommendations may make a topic feel unusually urgent or a claim unusually credible. The episode cautiously connects this to misinformation and conspiracy narratives. Recommendation pathways are not the sole cause of those beliefs, but they can be part of the environment in which repetition, emotion and apparent confirmation accumulate.
Signals that a feed may be steering the journey
A rabbit hole is easier to interrupt once you can see its shape. Useful signals include forgetting the original reason you opened the platform; recommendations becoming steadily more dramatic or specific; a topic suddenly feeling urgent; and each item seeming like an obvious next step even though you did not choose the overall direction.
Emotion is another signal. The point is not to suppress a reaction or assume that emotional content is false. It is to notice when anger, vindication, anxiety or excitement has become the engine of the next click.
How to add friction without abandoning curiosity
The answer is not to reject curiosity or treat every recommendation as manipulation. Curiosity is essential to learning. The aim is to make the journey more deliberate.
Pause before the next recommendation. A few seconds can turn an automatic click into a choice. Look for the pattern across several recommendations rather than judging each post in isolation. Ask whether the subject is becoming narrower, stranger or more emotionally charged. Set a simple intention before opening a platform: what are you here to find, and when will that task be complete? If the feed has become absorbing, show one item to another person and ask what they see. An outside perspective can puncture the private logic of a personalised sequence.
These are practical editorial suggestions, not a promise that a single habit will defeat every recommendation system. Their value is that they make attention visible again. Awareness introduces friction, and friction creates room for choice.
- Pause before the next recommendation.
- Notice the pattern across several items.
- Set an intention before opening the platform.
- Introduce an outside perspective.
For classrooms, families and individual users
Teachers can compare two different recommendation pathways and ask students which signals may have shaped each one. Parents can begin with curiosity rather than accusation: “What did you open the app to do, and where did it take you?” Teenagers and students can run a short experiment by stating an intention before opening a feed and noting whether the session stayed on course.
The goal is not surveillance or blame. It is shared language for a common experience. A person who can name the mechanism is better placed to decide whether to follow it.
A question before the next recommendation
When you follow a chain of recommendations, are you deliberately exploring a subject, or has the feed quietly taken over the route? The distinction may not be obvious in the moment. That is precisely why a pause matters.
Listen to Fact & Friction and explore more of Luminae’s work at www.luminae.org.
Evidence and editorial note
This article is grounded in the episode transcript and supplied claim register. Claims about narrowing or intensification are deliberately conditional. The practical countermeasures are presented as proportionate editorial guidance; the supplied evidence for their direct effect on recommendation loops is emerging.
From Fact & Friction
This Insight accompanies “Rabbit Holing: How Curiosity Gets Hijacked”.
Listen to the companion episode
Explore the connections
- Why You Can't Stop Scrolling: How Social Media Algorithms Shape Your Feed
- Why Online Content Makes You React: How A/B Testing Shapes Headlines and Feeds
Further source reading
YouTube: how recommendations work
Recommendation signals help explain a mechanism; individual experiences and outcomes vary.
Explore the research
Academic research notes supporting this Insight are available on request. Use our contact form and tell us which topic interests you.