End-user image suppression

This page deals with the technical implications of implementing end-user controlled suppression of potentially offensive images on Wikimedia projects.

For discussion of whether it would be desirable, see desirability of end-user image suppression.

This page originated from a discussion on the WikiEN-L mailing list where some level of consensus was reached between the free-speech and content suppression parties that a technical solution is possible that will go a long way towards satisfying most mainstream positions in the debate. However, it should be stressed that this page is not intended to imply that this is a fait accompli, but merely an initial exploration of the specification and feasibility of a possible solution. The features described below are not currently under development, and will certainly not be made live until their desirability has been firmly established.

The are generally four facets to this discussion:

End-user image suppression: the technical implications of implemention (this page)
Potentially offensive images: the identification of categories that would be included as part of implementation
w:Wikipedia:Descriptive image tagging: the useful categorising of images
Desirability of end-user image suppression: discussion about whether it's all desirable

Goal

The primary goal would be to give end-users the choice as to whether or not they see certain potentially offensive images by default, thus avoiding their offence and hence minimising the urge of editors to participate in self-censorship, while also placating those who would seek to discredit Wikimedia projects based on the adult nature of some content (disclaimers aside).

Basic idea

It would primarily involve ensuring that potentially offensive images are categorised, coupled with some combination of the following:

a site-based preference option allowing users to remove, hide or show potentially offensive images by default
a cookie-based preference option allowing users to remove, hide or show potentially offensive images by default
browser-based filtration (when supported in the future)

Initial default setting

An initial default setting for newly registered users would need to be chosen.

For anonymous users we could use a cookie. This cookie would be triggered when an anonymous user first attempts to view a page containing potentially offensive images, whereby they would first be presented with a page asking them to choose a default setting for their session.

One drawback of the cookie approach is that for those who already have content filtering tools (AOL child accounts, possibly schools and homes), there may be a question which even if answered "yes, want to see all types of content" doesn't result in display of the content, because the end user filter may block it anyway. However, the benefits for those without filters installed, perhaps because they aren't in their usual location, seem to be sufficient to overcome this objection.

User preference system

An Image Suppression preference section would be created. In this would be a choice to show all images by default. Under that would be the ability to choose amongst categories based on potentially offensive images and then choose to have one of the following happen:

Hide images (this would create a place-holder where images would normally be which the user could then choose to display on a case-by-case basis)
Remove images (this would remove any link or placeholder (but not the image from the server of course))

Categorising images

Tagging images would simply involve ensuring that potentially offensive images have all been given categories as per normal procedures.

Using categories directly in this way is unlikely to be a good technical solution, since it adds overhead for all category processing. However, using tags as part of templates which include categories in addition to markers for the standard content tagging system(s) may be suitable. Note that an image is likely to require several different tags, possibly several from each of many different tagging chemes popular in different countries.

Technical issues

The key technical issue for such content is the widespread availability of approaches to removing content which is present but not desired (e.g. many client side filters, the largest ISP in the U.S. defaults to filtered for child accounts) and the inability to add content which is missing but desired. In other words, the nature of the technology available favours including the content but facilitating its suppression if that is desired by a particular viewer.

A subsidiary technical issue is the desirability of using the tags supported by the existing wide range of client-side content filtering systems, so that those work automatically. These standards support a wide range of descriptive tagging types.