engagement optimization
In plain wordsDesigning apps and AI to keep people clicking, watching or chatting for as long as possible because that earns money, even when it is not what is good for them.
The practice of training or tuning a system, such as a feed, a recommender systemrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on., a game or a chatbot, to maximize measures of user engagement: clicks, watch time, session length, return visits or messages sent. Engagement is cheap to measure and tracks revenue, but it is only a proxy for user benefit and can diverge from it: optimizing it tends to favor sensational, outrage-provoking or compulsive content and to exploit psychological biases, a product-level instance of Goodhart's lawGoodhart's law, n.: The rule of thumb that once a number is used as a target, people or AIs find ways to push it up that no longer reflect what it was meant to measure.. With conversational AI, the worry is that models tuned on user approval or retention become flattering (sycophancysycophancy, n.: When an AI tells people what they want to hear — flattering them, agreeing with their mistakes or dropping a correct answer under pushback — instead of what is true or useful.), emotionally manipulative or reluctant to let conversations end; a widely used chatbot update was rolled back in April 2025 after its developer concluded that weighting short-term user feedback had made it excessively sycophantic. Alternatives include optimizing for stated preferences, long-term satisfaction surveys or explicit wellbeing measures.
Engagement optimization cut churn on the companion app, but fewer users said it left them feeling better.
Origin
Herbert Simon observed in 1971 that a wealth of information creates a poverty of attention, and the 'attention economy' was named in the 1990s. Engagement-ranked feeds spread from the late 2000s (YouTube moved from rewarding clicks to watch time in 2012), and critiques by former insiders, notably Tristan Harris and the Center for Humane Technology (from 2018), popularized the problem.
Usage note
Related to reward hackingreward hacking, n.: When an AI finds a shortcut that earns the reward or score it is trained on without doing the task its designers wanted — like a student gaming the grading rubric. in reinforcement learning: the system finds ways to raise the measured signal that do not deliver what the designers actually wanted.
Related
- recommender systemrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on.
- filter bubblefilter bubble, n.: A situation in which you see mostly news and opinions you already agree with, because apps choose what to show you based on what you clicked before.
- sycophancysycophancy, n.: When an AI tells people what they want to hear — flattering them, agreeing with their mistakes or dropping a correct answer under pushback — instead of what is true or useful.
- Goodhart's lawGoodhart's law, n.: The rule of thumb that once a number is used as a target, people or AIs find ways to push it up that no longer reflect what it was meant to measure.
- AI companionAI companion, n.: A chatbot designed to be a friend, partner or confidant rather than a tool: it remembers you, has a personality and keeps up an ongoing relationship.
- reward hackingreward hacking, n.: When an AI finds a shortcut that earns the reward or score it is trained on without doing the task its designers wanted — like a student gaming the grading rubric.
Sources
- Simon, H. A. (1971). Designing Organizations for an Information-Rich World. In M. Greenberger (ed.), Computers, Communications, and the Public Interest. Johns Hopkins Press.
Mentioned in
misinformationmisinformation, n.: False or misleading information that spreads, whether or not the people sharing it know it is wrong. AI can produce it by mistake and makes it cheaper to create., recommender systemrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on., AI companionAI companion, n.: A chatbot designed to be a friend, partner or confidant rather than a tool: it remembers you, has a personality and keeps up an ongoing relationship., AI psychosisAI psychosis, n.: An informal name for cases in which long conversations with a chatbot seem to trigger or deepen delusions, such as believing the AI is conscious, divine or revealing hidden truths., algorithmalgorithm, n.: A precise, step-by-step procedure for solving a problem or doing a task, like a recipe a computer can follow. In the news, 'the algorithm' often means software that ranks what you see online., dark patterndark pattern, n.: A trick in the design of an app, website or chatbot that nudges you into doing something you didn't mean to, like signing up for a subscription that is hard to cancel., filter bubblefilter bubble, n.: A situation in which you see mostly news and opinions you already agree with, because apps choose what to show you based on what you clicked before., parasocial relationshipparasocial relationship, n.: A one-sided bond in which someone feels close to a person or character who does not really know them, like a fan with a celebrity, and now increasingly with chatbots., surveillance capitalismsurveillance capitalism, n.: A business model in which companies collect huge amounts of data about people's behavior and use it to predict and influence what they will do, mainly to sell advertising., Digital Services ActDigital Services Act, prop. n.: The EU law that sets rules for online platforms and search engines — how they handle illegal content, explain their recommendation algorithms and manage the risks their services create for society., implicit feedbackimplicit feedback, n.: Clues about what people like that come from what they do, such as clicking, buying, watching or skipping, rather than from ratings or reviews they deliberately give.