We have had a lot of sites throughout the history of Stack Exchange that have suffered problems dealing with "recommendation" questions. That is, questions which are intending to gather a collection of recommended items, from which the author selects their preferred answer out of. Here, this would be questions like the following:
I enjoyed Lucky Number Slevin, what movies out there are like it that I'd enjoy?
Are there any good Steampunk movies out there?
However, these kinds of questions have very poor performance on the sites. This is because their functionality conflicts with the intent of our sites. Our goal is to have expert solutions that are identified by the depth of the knowledge of our users. The accepted answer system is founded on the principle that an answer serves as an understood closure to the question.
- They require extensive maintenance (which never happens for long) to be accurate
- The voting on such lists conflicts with the scope of voting on the site (people vote as if it was a poll, not based on quality)
- No single item in the list can conclude the question, which conflicts with the nature of Q&A.
- Many problems result in trying to reduce their scope, due to the subjectivity of how "narrow" criteria can be. So if any are allowed, generally all are allowed.
- Compounding on that, the ease and spread of itemized list recommendations gives a very poor image of the site as it does not showcase the expertise that many are looking for.
Now, at the moment, there isn't necessarily an intent to invite or allow recommendation questions. Recommendations were in fact noted as off-topic during the definition phase, and there's no trace of them in the top voted on-topic questions. Nevertheless, I'm writing this to hopefully get people in the mindset to understand why recommendations are hazardous, and consequently help setup a "no-rec" policy on the site before it opens to public beta.