User:AlecMeta/Alec's Solution to Image Filters

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

How would you do it, then?[edit]

Criticizing on the sidelines isn't the same as being in the driver's seat. If you criticize, you have a duty to try to help find a solution you wouldn't criticize. This is my attempt.

Reframe theoretically[edit]

  • Identifying the set of objectionable content is very controversial.
  • But identifying the set of "high-quality" content is routine.
    • Featured Articles, Good Articles, and other committee-based methods successfully identify quality content.
    • Voting-based methods also successfully identify high-quality content.
    • The article feedback tool is the most powerful tool of all, collecting per-reader feedback on quality.


  • Why is flagging okay for "high quality" but not "offensive"?
  • "High-Quality" is an "objective-enough" term that we can basically agree what it means in practice.
  • "Offensive" is an emotion. The proof that an image is offensive is a human sincerely saying they were offended by the image.
  • Objective measures work, Subjective measures don't work yet.

Measuring the subjective[edit]

  • Our readers will have highly subjective reactions to our content.
  • We should empower our readers to share their reactions to our content. The more feedback we get, the better.
  • Instead of asking readers for a 1-10 quality rating, we could also let them "tag" specific content with their own subjective reactions. If somethings "mindblowing" or "hilarious" or "eloquent", let them record that opinion.

A good tool with a good purpose.[edit]

  • The main purpose of a "tagging system" is to find our "BESTEST" content, where "BESTEST" is a subjective tag.
  • The main purpose of a tagging system is to find good content.
    • BUT--- if a user tags an image as "objectionable", we could shutter it for them in future.

Simple tagging[edit]

  • Every user will disagree greatly about what content is "mindblowing". The term is really vague. That's okay.
  • If readers keep tagging the same content as "mindblowing", we can assume it probably is. It's a subjective statment.
  • When I ask to see the "most mindblowing image", sort by the rate people have assigned that tag to that image.
  • Replace "mindblowing" with "offensive" and the argument still works.

Content recommendation[edit]

  • Netflix and Amazon know what I want before I do.
  • If we create a "tagging' tool/project, we can start collecting that kind of data from our editors, with an eye towards providing them with "recommended content" suggestions in the future.

In short[edit]

  • Create a whole new undertaking to let people collaborate on subjective tagging of WM content.
  • As an afterthought, let them also tag "objectionable" content.
  • The real feature is the thousands of other tags, not the tag "objectionable".