Twitter has recently proved its usefulness in a variety of disasters and emergency situations (see my blog post on this). But what about as a backup to the 911 system? In most cases, a person in distress who has access to Twitter on their mobile device would also be able to make a 911 call (or whatever the local emergency number is). However, there are common circumstances where this might not be so: a person is in a poor reception area and can only send text messages; they are under threat and need to ask for help covertly; they are in odd locations (e.g. international waters) and it is not clear how to get assistance; or they are in a disaster area and 911 infrastructure is down or is overloaded. Some of these may be addressed by the enhanced 911 system ability to receive text messages, but it is not clear yet how this will pan out. Twitter is already there and it works.
In these kinds of circumstance, anecdotal media reports (see for example this USA Today article) indicate success in that if a person has enough followers (how many is enough?) one of them will likely see a distress message in time, and call 911 (assuming there is enough information to locate the individual). Of course having geotagged tweets can help with this (although if a person is in poor cellphone coverage area assisted GPS will likely not work). But it seems like it would be useful for there to be some way to "elevate" your tweets in an emergency, to stand out against the background chatter. Some way to say "I really need help now" that stands out, like a CB channel 9 or Marine Channel 16.
The obvious way to do this is with a hashtag. Let's say we had a dedicated "I need emergency help" hashtag that could be monitored by agencies (even constrained to a geographic area). Well there are problems: first almost every available easy-to-remember hashtag is already widely used for non-emergency messages - just try a search for #911, #helpme, #call911, #needhelpnow, etc. Even if you devised a more unusual (and harder to remember) hashtag like #ineed911help, you'd get overloaded with retweets, commentary on the "does the #ineed911help system work?" etc very quickly. Second, you would likely get too many responses - imagine a 911 center suddenly getting 10000 calls.
Another way would be natural language processing - identifying patterns, particular phrases, etc, that likely constitute a genuine emergency call. Maybe a predictive model based on these terms. But is this really likely to work? And what about the lawsuits when it doesn't work?
So I think that for right now, the best system is no system. But I think this is something we should be talking about. If you have thoughts on this, please leave comments!