I wonder if there should not be a MathOverflow Twitter presence, somehow generated automatically, perhaps from the few HotNet Questions to which MO questions ascend to that list?

There could be some Twitter notice that, e.g., "When does a space of endomorphisms contain invertibles?" has risen to one of the top $100$ HotNet questions.

Or do we not even want to promote MO? I would argue that we should promote MO. But that is, perhaps, for another post.

• I think there used to be Twitter bots on SE sites, but that was phased out a couple of years ago. – Asaf Karagila Aug 26 '17 at 20:53
• One question is who would be responsible for tweeting out messages. I'd hate to be that person. Then again, I have a pronounced aversion to Twitter. – Todd Trimble Aug 26 '17 at 21:46
• Somewhat related older post (now deleted): Does MO have an official twitter account? The discussion under this question seems a bit related, too: Why do we link to Twitter, Facebook or Google+ below every question? – Martin Sleziak Aug 26 '17 at 21:48
• @ToddTrimble Just out of curiosity, why so Twitter-averse? – Reinstate Monica Aug 27 '17 at 4:07
• Same reasons that just about anyone will tell you if they tell you why they dislike Twitter. This seems fairly representative: rcpmag.com/blogs/lee-pender/2011/03/… But there is no shortage of people who will try to argue against such points, and I'm not much interested in going down that road, certainly not here. – Todd Trimble Aug 27 '17 at 4:57
• @ToddTrimble Looking at twitter accounts of other SE sites (such as math.SE. TeX.SE, academia.SE, ...) it seems that the questions are picked by an automated process, not by a specific person. This answer summarizes what I know about the rules how they are chosen. (It is possible that more details can be found in meta.SE questions tagged twitter.) – Martin Sleziak Aug 27 '17 at 6:00
• @MartinSleziak Interesting, thanks. So the writing of the tweet is automated as well? Also it's probably good that the criterion is based more on bounties than on page views, although I'd imagine some human oversight might be desirable in any case. – Todd Trimble Aug 27 '17 at 12:15
• If the choice of questions to tweet were based on hot network questions list, then questions containing MathJax in the title would never appear there. Still, I guess it is possible to calculate for such questions whether question is hot or not (they are just not picked in the networkwide list.) – Martin Sleziak Aug 27 '17 at 14:59
• Well, I guess that some users consider it an advantage if a question does not appear in hot questions list. My personal preference that if omitting LaTeX would lead to a less clear title, it's better to sacrifice the possibility to have a hot question than to sacrifice clarity of the title. (Anyway, it seems that we digressed from the original topic a bit.) – Martin Sleziak Aug 27 '17 at 15:06
• I think a Twitter presence would be very likely to attract a lot of low-quality posts and comments. – Steven Landsburg Aug 28 '17 at 2:51
• BTW if you search a bit in Google or on Twitter or using #mathoverflow you can find some tweets MO-related tweets. In fact, JDH tweeted about this meta question (and it seems that he has many tweets related to MO posts.) – Martin Sleziak Aug 28 '17 at 5:59
• My experience is that there is an active community of knowledgeable mathematicians on twitter, and I expect that an MO twitter presence would be very welcome for them. – Joel David Hamkins Aug 28 '17 at 14:02
• @StevenLandsburg I often share MO posts on twitter, which evidently produce traffic (frequently resulting in the announcer badge), but I haven't especially noticed low-quality answers or comments on those posts. – Joel David Hamkins Aug 28 '17 at 14:06
• @JoelDavidHamkins : Thanks. Your experience trumps my speculation. – Steven Landsburg Aug 28 '17 at 15:01
• @JoelDavidHamkins: Additional low-quality comments would be unlikely anyway, since commenting requires 50 points. – Stefan Kohl Aug 29 '17 at 9:41

I am posting this mainly to summarize some points mentioned already in the comments - in an answer they can be a bit expanded and presented in somewhat better form. (The comment section grew quite long, I am one of the people to blame for that.)

I guess that at least some of the information mentioned here might be useful for the decision whether and how MO Twitter account could be set up.

Twitter account of other SE sites

As you can notice, many sites in Stack Exchange network have twitter accounts. Tweets are posted by twitter bots which were set up by SE and post tweets about some posts from a specific site.

For example, I can mention Math.SE. TeX.SE, Academia.SE, Physics.SE, etc. You can find twitter accounts for various sites in the list of Stack Exchange sites. However, not every site has a twitter account.

All I know about the rules, based on which the posts which are tweeted are selected, can be found here, here and here. Brief summary: For 8 posts a day (once in every three hours) the posts are chosen from these categories: bounties, unanswered, often viewed, interesting questions and answers, hot on meta. I am not sure whether the exact algorithm which chooses posts to be tweeted is published somewhere.

To me this seems as the "cheapest" option. There already is existing infrastructure and an algorithm which selects what to tweet. So it would probably not be too much work to add this also for MO. (However, since these twitter bots are no longer supported - one of the main reasons might be the problems mentioned above - I am not sure whether SE staff would be willing to do this.)

EDIT: Probably it's worth adding that there are some SE sites that have a twitter account which is run by community, i.e., tweets are added manually and not by bots. There are two such sites that I am aware of: Literature Stack Exchange (twitter account and announcement on meta) and Mi Yodeya (twitter account and announcement on meta).

Which questions to tweet

Several possibilities how to choose questions which should be tweeted have been mentioned.

• As I wrote above, twitterbots for other sites already have some algorithm how to choose questions.
• Hot network questions were suggested as a possibility by the OP. I'll just point out that questions with MathJax in the title are never chosen in the HNQ. Still, I guess it is possible to calculate for such questions whether question is hot or not (they are just not picked in the networkwide list.)
• Some other possibilities would be choosing questions similarly as in the newsletter (suggested in a comment).
• You can see questions listed as greatest hits and questions which are currently hot. AFAIK the choice of these questions is related to choice of question in the newsleter and the hot network questions list.

Users who have access to site analytics can see how much of the referrals are generated by twitter. (As I mentioned in my comment, the statistics of traffic sources for math.SE t.co is listed among referring sites with 3.4%. This is aprroximately same level as reddit.com, facebook.com, m.facebook.com and googleweblight.com; but lower than mathoverflow.net with 5.9%. I do not have access to site analytics on MO.)