Search Manipulation

From wiki.measurementlab.net
Jump to: navigation, search


Interference Methods

  • Filter bubble “personalization” so that some clients/users don’t see results.
  • Profile pollution that affects personalization.
  • Manipulation of search results (e.g., removing certain results)
  • Google bombing: other forms of manipulating indexes and making censored info harder to find
  • Google search results customization?
  • Facebook newsfeed filtering
  • Filtering search results
  • Gmail blocks/deletes/files-into-spam-folder letters that it thinks are spam


Detection

Search-measurement and storage.png

Search Results Rewriting

  • Measurement manipulation of “searcher” experience, considering taxonomy of what is unique based on potential variables of: geography, browsing or search history, whether logged in, cookies and font cache, time of day, browser type, et al;
  • Quantification of modifications may include: order of results, whether any results are returned, intermediaries that may modify the results (such as the provider itself or third parties injecting content);
  • Input for testing can be gathered from Wikipedia page titles;

Best Practices:

  • Longitudinal component,
  • Different search engines;

Need to control for:

  • General quality of the search engine,
  • Decided set of “normal” results for a search,
  • All search types and providers,
  • Changes to underlying algorithms that may change results naturally;
  • Only ascribe motivation of manipulation afterward, just measure the differences at the time of testing;
  • Need a low-bar search term that will expose the manipulation occurring in the search and indicate the technique;

Potential non-measurement outcomes:

  • Urge people to use a search provider that is more controllable by the user,
  • Convince search providers to publish a blacklist of terms or URLs,
  • Create pressure through international market forces for less censorship,
  • Inform users that there are options and informs public choice,
  • Potential site akin to ‘is this site censoring this term for me’ .com