# Learning a person > How a system can help someone it barely knows yet, and why fresh answers beat old patterns. - Page: https://tre-up.com/research/topics/learning-a-person - Topic 06 of 7 in [Research Notes](https://tre-up.com/research) · Last reviewed 2026-10-07 A system that doesn't know someone yet does badly if it guesses. Asking briefly is one of the few approaches that works from the first day. Preferences drift, so fresh answers should outweigh old patterns, and what a person ignores is a weak signal of what they don't want. ## Questions in this topic - [How can an app help me if it doesn't know me yet?](https://tre-up.com/research/how-can-an-app-help-me-if-it-doesnt-know-me-yet): By asking, briefly, instead of guessing. A 2023 systematic review (Yuan and Hernandez) treats the 'cold start', a new user with no history, as its own problem, and asking the person directly is one approach that works from the first day. Later, fresh answers should outweigh old patterns, because preferences drift (Koren, 2009). Markdown: https://tre-up.com/research/how-can-an-app-help-me-if-it-doesnt-know-me-yet.md ## What this topic rests on (5 sources) - **User Cold Start Problem in Recommendation Systems: A Systematic Review**. Yuan H, Hernandez AA (2023). IEEE Access 11:136958–136977. https://doi.org/10.1109/ACCESS.2023.3338705 - Institutions: Affiliation not yet confirmed - Design: Systematic review - What it found: Recommendation quality drops sharply for a user with no history, the 'cold start'. The review treats it as its own problem and maps the ways systems have tried to solve it, including simply asking the user. - Where the evidence stops: Recommender systems, not personal guidance. We could not reopen the abstract on 7 October 2026; this summary rests on the lab's earlier reading and its DOI check. - Our summary: https://tre-up.com/research/sources/yuan-hernandez-2023-cold-start - **Collaborative Filtering with Temporal Dynamics**. Koren Y (2009). Proceedings of KDD 2009, 447–456. https://doi.org/10.1145/1557019.1557072 - Institutions: Yahoo! Research (Israel) - Design: Method with large-scale evaluation - What it found: People's preferences drift over time. Models that track that drift, instead of treating old data like new, predicted ratings on a large Netflix dataset better than anything reported before. - Where the evidence stops: Movie ratings, not daily behaviour. No decay rate carries over. - Our summary: https://tre-up.com/research/sources/koren-2009-temporal-dynamics - **Negative Sampling in Recommendation: A Survey and Future Directions**. Ma H, Xie R, Meng L, Feng F, Du X, Sun X, Kang Z, Meng X (2026). ACM Transactions on Information Systems 44:1–44. https://doi.org/10.1145/3793855 - Institutions: Shandong University (China); Tencent (China); University of Science and Technology of China (China) - Design: Survey - What it found: Surveys how recommender systems learn from what people do not choose. Positive signals get most of the attention; negative signals are noisier and easy to misread, and something a person did not see cannot be counted as a 'no'. - Where the evidence stops: A survey of recommender methods. Affiliations are taken from the preprint. - Our summary: https://tre-up.com/research/sources/ma-2026-negative-sampling - **Personalized HeartSteps: A Reinforcement Learning Algorithm for Optimizing Physical Activity**. Liao P, Greenewald K, Klasnja P, Murphy S (2020). Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 4(1):18. https://doi.org/10.1145/3381007 - Institutions: University of Michigan (United States); IBM Research (United States); Harvard University (United States) - Design: Learning algorithm with a pilot deployment - What it found: An algorithm that learns, for each person, when a suggestion is worth sending, improving as their data comes in. In the HeartSteps app it decided up to five times a day whether to send an activity suggestion. - Where the evidence stops: An algorithm paper with a small pilot; not a trial of outcomes. - Our summary: https://tre-up.com/research/sources/liao-2020-personalized-heartsteps - **Using Personalized Intervention Criteria in a Mobile Just-in-Time Adaptive Intervention for Increasing Physical Activity in University Students: Pilot Study**. Ikegaya M, Foo JC, Murata T, Oshima K, Kim J (2025). JMIR Human Factors 12:e66750. https://doi.org/10.2196/66750 - Institutions: Shizuoka University (Japan); University of Alberta (Canada); Heidelberg University (Germany) - Design: Pilot study, two groups; sample: 28 university students, 2 weeks - What it found: Prompts to move were triggered either by each student's own first-week baseline or by a population average. Both raised activity in the hour after a prompt, and the personal threshold raised it more. Over the whole week, neither raised activity. - Where the evidence stops: Twenty-eight healthy students, a pilot; the two groups' thresholds came from different samples. - Our summary: https://tre-up.com/research/sources/ikegaya-2025-personalized-criteria ## Other topics - [Timing and receptivity](https://tre-up.com/research/topics/timing-and-receptivity): When help lands, and when the same help is just noise. - [Interruptions and attention](https://tre-up.com/research/topics/interruptions-and-attention): What a ping, a switch or an unfinished task costs the mind. - [Engagement is not benefit](https://tre-up.com/research/topics/engagement-is-not-benefit): Why opening an app more often is not proof that it helps. - [Autonomy and wording](https://tre-up.com/research/topics/autonomy-and-wording): Why people push back on advice, and the words that make it easier to hear. - [Trust and uncertainty](https://tre-up.com/research/topics/trust-and-uncertainty): How people come to rely on automated advice, and how a system should admit what it doesn't know. - [Privacy and deletion](https://tre-up.com/research/topics/privacy-and-deletion): What deleting your data really means, from the law to the backup tape. --- © 2026 Tre-Up. You may read, link to, quote and summarise Research Notes, and search engines and AI assistants may index, summarise and cite them, as long as the source is credited with a link to the page on tre-up.com. Please do not present our notes as your own, or republish whole pages, without permission. Terms: https://tre-up.com/terms#use · About Tre-Up: https://tre-up.com/about