Privacy in personalized search protects a person’s past behavior from unwanted collection, exposure, or inference. Personalization can use clicks, purchases, and other interaction patterns, but revealing those interests may embarrass users and reduce trust. section 9.5
Apply privacy safeguards when building personalized search:
- Explain which signals are collected and how they influence recommendations or results. section 9.5
- Provide an opt-out path for users who do not want personalization based on their interaction history. section 9.5
- Use personalization conservatively, such as applying it only to the top few results, and preserve user control through the current query. section 9.5
- Preserve privacy while using feedback loops to update changing personalization profiles. section 9.5 section 9.6
Sensitive recommendations
A streaming service might infer interest in violent or adult-themed movies; a bookstore might expose interest in romance or self-improvement; a grocery service might reveal preferences for junk food or alcohol. Showing these inferences unexpectedly can erode trust. section 9.5
