On 18 September 2011 14:38, Yann Forget yannfo@gmail.com wrote:
At the beginning, I was quite neutral about a filter: I had no idea how it would work, and I wouldn't use it, but what if somebody else wants it? But after reading nearly all comments on this list, I think that the arguments for a filter do not hold water. The pratical implemention would be a nightmare, and the purpose not really within Wikimedia mission. The thread above on how to create categories for a filter is full of irrational assumptions, impracticable propositions, and impossible solutions. It seems it is time to drop the whole idea...
The problem is that "offensive image" is a magical category.
The concept of a "magical category" is useful and important. It's something that sounds simple in ordinary language, but turns out to be a nightmare to implement and result in stupidity.
I got the phrase from http://lesswrong.com/lw/td/magical_categories/ , which discusses the problem of magical categories in artificial intelligence. But it's vastly useful in general.
A magical category will be something that is put forward as objective, and thus reducible to computer instructions, but everyone seems to have a different subjective interpretation. Worse yet, there is not even a reducible way to process as-yet-unknown examples, other than "I know it when I see it".
Magical categories are common in rhetoric and interpersonal human politics. An example is when someone appears to solve a problem in words, but you go "what!" because you know the details are subjective, squishy and not actually reducible in any way at all.
Apart from a simple "all images on/off" filter, *every* proposed offensive image category in this discussion has been a magical category: subjective, individual, argued over, and with a ton of "I know it when I see it."
In passing this resolution, the board appears to have fallen for the illusion that magical categories are possible to implement. This is unfortunate.
- d.