User:Ilmari Karonen/Do it by hand

Wikipedia policies and guidelines often contain statements of the form:

For all X, you may/should do Y.

For technically savvy users, it may often be tempting to enforce such guidelines using an automated process. This is not necessarily a good idea. Such rules have usually been written with human editors in mind, and may have unintended consequences or side effects if applied strictly to the letter on a large scale.

Examples

For example, a guideline might say that all professors or all musicians with a published album are notable. This does not necessarily make it a good idea to compile a database of musicians or professors and run a bot to create stubs for all of them. Rather, such rules should be read as allowing interested human editors to create individual articles on those topics, with the unwritten assumption that, should their numbers grow unmanageable, the rules might later be reconsidered.

Similarly, when a change is made to the Manual of Style, it is often tempting to run a bot over all articles to enforce the new style. This may, again, be a bad idea. True, letting human editors implement the style change will take a long time, during which Wikipedia's style will be inconsistent in that respect, and a few missed pages will probably always remain. Even so, there are a number of reasons to prefer this method over doing it with a bot.

First of all, human editors possess a degree of contextual awareness far beyond anything that can ever be programmed into a bot. They can see where legitimate exceptions might need to be made or where other changes might be needed to make the new style look good. Furthermore, the slowness of the implementation can actually be a good thing — it allows any problems with the new rule to be spotted early in the process and minimizes the effort needed to go back to the old style if it should turn out that the new style isn't so good after all.

Wikipedia is a big place. No matter how much community participation is sought, it's unlikely that the authors of any guideline can be familiar with all the places it might potentially affect, especially if enforced mechanically to the letter. Thus, almost any rules we make will have to be ignored in some cases in order to obtain a sensible outcome. Indeed, a longstanding principle on Wikipedia has been that our rules are descriptive, not prescriptive. This means that they often lag behind actual practice, have unwritten exceptions and, in those occasional cases where a rule is written with the intent to prescribe a new practice, may never get fully implemented if editors decide that they don't really make sense.

That is, of course, unless someone comes along with a bot and rides roughshod over the objections, saying that he's "only following policy".

Of course, after sufficient debate, it may indeed be decided that having a bot, say, create stubs for all U.S. census-designated places is a good idea. In that case, of course, it's okay to do it. But one should not draw that conclusion merely from a vague guideline stating that "all geographic locations are notable", or whatever.