{"course":{"slug":"ai-actually-explained","title":"AI, Actually Explained","url":"https://addaly.com/learn/ai-actually-explained","level":"beginner","track":"start"},"lesson":{"slug":"the-feed-that-learns-you","title":"The feed is a prediction machine","url":"https://addaly.com/learn/ai-actually-explained/the-feed-that-learns-you","position":11,"of":92,"minutes":9,"answer":"A feed predicts what you will do next, not what is good or true, so watch time steers it far more than likes and the only real control you have is what you finish watching.","keyIdea":"A feed predicts what you will do next, not what is good or true, so watch time steers it far more than likes and the only real control you have is what you finish watching."},"body":{"format":"markdown","text":"## What is actually happening when you scroll\n\nEvery time a feed loads, three things happen in under a second.\n\nFirst, **candidate generation**: from millions of possible items, a fast and rough process pulls a few hundred that might suit you. Second, **scoring**: a much more expensive model predicts, for each candidate, how likely you are to do each of several things: watch to the end, rewatch, like, comment, share, or close the app. Third, **ranking**: those predicted probabilities are combined by a weighted formula, and the order you see is the result.\n\nThe weights in that final formula are a business decision, not a learned one. Somebody chose how much a predicted share is worth relative to a predicted long watch. When a platform announces it is \"showing more content from friends\", that is usually those weights being changed, not a new model.\n\nThis structure is published. YouTube described the two-stage design in a 2016 paper; Instagram has published lists of its ranking signals; Twitter released a version of its ranking code in 2023. The details differ, the shape does not.\n\n## What it is optimising for\n\nThis is the sentence to keep. **A feed does not predict what is good for you, or true, or important. It predicts what you will do next.**\n\nEverything else follows. Content that reliably provokes a response is promoted because provocation predicts response. Content that is quietly useful and produces no reaction is demoted, not out of malice, but because the model has no column for \"was quietly useful\".\n\nThe strongest signal on most video platforms is not the like button. It is watch time, and specifically whether you finished and whether you watched again. That is why a short video you rewatch by accident moves your feed more than a video you deliberately liked. If you want to steer a feed, the effective lever is what you actually watch to the end, not what you tap.\n\n## Why short video learns you faster\n\nA fifteen-second clip gives the system a complete signal every fifteen seconds. A forty-minute video gives one signal in forty minutes. Feed the same learning machinery a hundred times more feedback per hour and it converges on your preferences far quicker.\n\nThat is most of why short-form feeds feel uncannily accurate within a day, while a film recommendation service still suggests things you would never watch after a year. It is not a smarter model. It is a faster loop.\n\n## Where the honest uncertainty is\n\nThe filter-bubble story (that these systems trap people in a narrowing tunnel of agreement) is far less settled than headlines suggest.\n\nSeveral large studies, including a set of 2023 experiments run with Meta's cooperation during a US election, found that switching people to a chronological feed changed what they saw a great deal and changed their political attitudes very little over the period measured. Other work finds real amplification effects. Researchers disagree, the experiments are hard, and the platforms control the data.\n\nSo the fair statement is: these systems demonstrably shape attention, and their effect on belief is contested. Anyone who states either side as established fact is going beyond the evidence.\n\n## What you can actually do\n\nUnlike most systems in this block, this one has real controls, and they are free.\n\n- **Watch and search history** is what the model uses. On Google and YouTube it lives at `myactivity.google.com`, where you can view it, delete items, pause collection, or set automatic deletion after three months.\n- **\"Not interested\" and \"Don't recommend this channel\"** feed the model a negative signal directly. They work, though slowly, because one signal competes with thousands.\n- **Logged-out or incognito viewing** gives the system much less to work with, at the cost of worse recommendations, which is exactly the trade being made.\n- On most platforms there is a **following-only or chronological view** buried in the settings. It is usually not the default, and usually reverts.\n\nNone of this is about resisting a manipulation. It is about knowing that a system is predicting your behaviour from your behaviour, and that the only input you control is the behaviour."},"check":{"question":"A user carefully likes only educational videos but keeps watching arguments to the end without liking them. Their feed fills with arguments. Why?","options":[{"text":"The ranking model weights completion and rewatching heavily, so finishing a video is a stronger signal than a like that follows a short view.","correct":true},{"text":"The platform deliberately promotes conflict because outrage is more profitable than education.","correct":false,"tempting":"Assumes an explicit editorial intent where a behavioural prediction plus a weighting choice already explains the result."},{"text":"Likes are used only for the public like count and are not a ranking input at all.","correct":false,"tempting":"Overstates the case; likes are a genuine signal, merely a weaker one than sustained watching."},{"text":"The model has not yet had enough data, and the feed will correct itself once enough likes accumulate.","correct":false,"tempting":"Treats the mismatch as a data shortage rather than as the system correctly learning from the behaviour it was given."}]},"prev":{"slug":"already-around-you","title":"Where it already is, without the label","url":"https://addaly.com/learn/ai-actually-explained/already-around-you"},"next":{"slug":"spam-fraud-and-the-oldest-ai","title":"The models that have been working since 2002","url":"https://addaly.com/learn/ai-actually-explained/spam-fraud-and-the-oldest-ai"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'The feed is a prediction machine', in AI, Actually Explained. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-actually-explained/the-feed-that-learns-you (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"The feed is a prediction machine","courseTitle":"AI, Actually Explained","module":{"index":2,"slug":"where-it-already-is","title":"Where it already is, without a label"},"answer":null,"keyIdea":"A feed predicts what you will do next, not what is good or true, so watch time steers it far more than likes and the only real control you have is what you finish watching.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"what-is-actually-happening-when-you-scroll","runs":[{"text":"What is actually happening when you scroll"}]},{"t":"p","runs":[{"text":"Every time a feed loads, three things happen in under a second."}]},{"t":"p","runs":[{"text":"First, "},{"text":"candidate generation","b":true},{"text":": from millions of possible items, a fast and rough process pulls a few hundred that might suit you. Second, "},{"text":"scoring","b":true},{"text":": a much more expensive model predicts, for each candidate, how likely you are to do each of several things: watch to the end, rewatch, like, comment, share, or close the app. Third, "},{"text":"ranking","b":true},{"text":": those predicted probabilities are combined by a weighted formula, and the order you see is the result."}]},{"t":"p","runs":[{"text":"The weights in that final formula are a business decision, not a learned one. Somebody chose how much a predicted share is worth relative to a predicted long watch. When a platform announces it is “showing more content from friends”, that is usually those weights being changed, not a new model."}]},{"t":"p","runs":[{"text":"This structure is published. YouTube described the two-stage design in a 2016 paper; Instagram has published lists of its ranking signals; Twitter released a version of its ranking code in 2023. The details differ, the shape does not."}]},{"t":"h2","id":"what-it-is-optimising-for","runs":[{"text":"What it is optimising for"}]},{"t":"p","runs":[{"text":"This is the sentence to keep. "},{"text":"A feed does not predict what is good for you, or true, or important. It predicts what you will do next.","b":true}]},{"t":"p","runs":[{"text":"Everything else follows. Content that reliably provokes a response is promoted because provocation predicts response. Content that is quietly useful and produces no reaction is demoted, not out of malice, but because the model has no column for “was quietly useful”."}]},{"t":"p","runs":[{"text":"The strongest signal on most video platforms is not the like button. It is watch time, and specifically whether you finished and whether you watched again. That is why a short video you rewatch by accident moves your feed more than a video you deliberately liked. If you want to steer a feed, the effective lever is what you actually watch to the end, not what you tap."}]},{"t":"h2","id":"why-short-video-learns-you-faster","runs":[{"text":"Why short video learns you faster"}]},{"t":"p","runs":[{"text":"A fifteen-second clip gives the system a complete signal every fifteen seconds. A forty-minute video gives one signal in forty minutes. Feed the same learning machinery a hundred times more feedback per hour and it converges on your preferences far quicker."}]},{"t":"p","runs":[{"text":"That is most of why short-form feeds feel uncannily accurate within a day, while a film recommendation service still suggests things you would never watch after a year. It is not a smarter model. It is a faster loop."}]},{"t":"h2","id":"where-the-honest-uncertainty-is","runs":[{"text":"Where the honest uncertainty is"}]},{"t":"p","runs":[{"text":"The filter-bubble story (that these systems trap people in a narrowing tunnel of agreement) is far less settled than headlines suggest."}]},{"t":"p","runs":[{"text":"Several large studies, including a set of 2023 experiments run with Meta’s cooperation during a US election, found that switching people to a chronological feed changed what they saw a great deal and changed their political attitudes very little over the period measured. Other work finds real amplification effects. Researchers disagree, the experiments are hard, and the platforms control the data."}]},{"t":"p","runs":[{"text":"So the fair statement is: these systems demonstrably shape attention, and their effect on belief is contested. Anyone who states either side as established fact is going beyond the evidence."}]},{"t":"h2","id":"what-you-can-actually-do","runs":[{"text":"What you can actually do"}]},{"t":"p","runs":[{"text":"Unlike most systems in this block, this one has real controls, and they are free."}]},{"t":"ul","items":[[{"text":"Watch and search history","b":true},{"text":" is what the model uses. On Google and YouTube it lives at "},{"text":"myactivity.google.com","code":true},{"text":", where you can view it, delete items, pause collection, or set automatic deletion after three months."}],[{"text":"”Not interested” and “Don’t recommend this channel”","b":true},{"text":" feed the model a negative signal directly. They work, though slowly, because one signal competes with thousands."}],[{"text":"Logged-out or incognito viewing","b":true},{"text":" gives the system much less to work with, at the cost of worse recommendations, which is exactly the trade being made."}],[{"text":"On most platforms there is a "},{"text":"following-only or chronological view","b":true},{"text":" buried in the settings. It is usually not the default, and usually reverts."}]]},{"t":"p","runs":[{"text":"None of this is about resisting a manipulation. It is about knowing that a system is predicting your behaviour from your behaviour, and that the only input you control is the behaviour."}]}],"check":{"question":[{"text":"A user carefully likes only educational videos but keeps watching arguments to the end without liking them. Their feed fills with arguments. Why?"}],"options":[{"text":[{"text":"The ranking model weights completion and rewatching heavily, so finishing a video is a stronger signal than a like that follows a short view."}],"correct":true,"tempting":null},{"text":[{"text":"The platform deliberately promotes conflict because outrage is more profitable than education."}],"correct":false,"tempting":[{"text":"Assumes an explicit editorial intent where a behavioural prediction plus a weighting choice already explains the result."}]},{"text":[{"text":"Likes are used only for the public like count and are not a ranking input at all."}],"correct":false,"tempting":[{"text":"Overstates the case; likes are a genuine signal, merely a weaker one than sustained watching."}]},{"text":[{"text":"The model has not yet had enough data, and the feed will correct itself once enough likes accumulate."}],"correct":false,"tempting":[{"text":"Treats the mismatch as a data shortage rather than as the system correctly learning from the behaviour it was given."}]}],"order":[2,1,0,3],"why":null},"prev":{"slug":"already-around-you","title":"Where it already is, without the label","minutes":6},"next":{"slug":"spam-fraud-and-the-oldest-ai","title":"The models that have been working since 2002","minutes":9}}}