On Thu, Jun 11, 2009 at 11:28 AM, Dossy Shiobara <[email protected]> wrote:
> > following you" email for people that look >> like they are abusing the system. >> > > Absolutely. There should be a silent rate limit around following - normal > human activity shouldn't really be >1 follow per second, and no more than > 100 in a 5 minute period, reasonably. (We can argue about the fine-tuning > of these numbers, but lets agree that we need both of these metrics.) I think you may not be considering legitimate automated systems that can quickly find a number of people who are discussing n emerging topic. My social network analysis does that - when it sees a topic becoming hot, it does some searches to see who is talking about it. If the topic reaches a threshold of significance, the system's Twitter user will immediately follow all the people it has found who have recently talked about it. In reality, those are often people who it is already following, but as Twitter grows, it tends to include more and more new people. The system periodically un-follows people who no longer seem to be talking about hot topics. Somebody could easily interpret my system's behavior as spammy, since there's a fair bit of churn in who it follows. But that's not what is about at all. It is about identifying emerging communities of shared interest. How fast it tends to follow and un-follow depends on how fast those communities (or cliques, perhaps is a better word) arise and dissolve. I imagine that some day, everybody will have agents along these lines running on their desktops... and it will be a good thing, especially as software becomes better at summarization of this kind of behavior. It makes platforms like Twitter into something that starts to look more like a neural net than just a many-to-many network. I'm not saying there should be no limits, just offering a perspective on a kind of analysis that excessive limits could stifle. Nick
