MENTAL MODEL · M6200

Cold Start

SystemsHigh supportSystems Theory

Included

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

Refers to the challenge of how a system can begin operating without any historical data (user profiles, vehicle preheating). For example, when a new user registers for an app, the system doesn't know what to recommend.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Breaking through from zero to one. In any new beginning (new job, new relationship), don't expect the system to automatically adapt to you. You must proactively provide 'initial fuel' (self-introduction, proactive display, preheating) to help the system get through the cold start period.

anchor

Anchor fast decisions

New users or items lack historical interaction data, so recommendation algorithms cannot make accurate predictions based on collaborative filtering; the system is in a 'zero-data' state. It must start from external information (attributes, content features), otherwise it cannot function.

MINIMUM ACTION

In progress 0/4

Practice this model in one real situation:

Collect attributes proactively provided during registration (interest tags, self-introduction, etc.).
Use content-based recommendations (item features) for initial matching.
Introduce popular lists or random exploration to mitigate the exploration-exploitation trade-off.
After quickly accumulating interactions, switch back to collaborative filtering.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Cold_startZH · Explicit
verified

RELATED MODELS