Geosocial filter failure

I have been off blogging for a couple weeks while taking time to make headway on a book on the principles of social interaction design. I don’t like being away from blogging this long. It can precipitate lifestyle changes in the offline department, which when combined with the recent stretch of fine Bay area weather we have been having, can becoming self-reinforcing. Pro-social in an offline way, anti-social in an online way.

Today’s the day that I reflect on the book, come up with examples, and look over the rough cut to get a sense for what’s still missing. I was doing this a few minutes ago when I drifted into thinking about Clay Shirky’s concept of filter failure. I had been searching for what to say about geolocation services. And something jumped out at me that I think merits a quick post.

In news of SXSW recently, geolocation services (primarily Foursquare and Gowalla) were cited as the geek habit du jour (or make that week). Several attendees blogged that Foursquare had replaced twitter as the means by which to find friends and coordinate meetups. Of course this was how twitter was used at SXSW a few years ago. Both tweets and Foursquare checkins, if they are used to signal one’s location, effectively do the same thing. Tweeting is then reduced to its signaling function; and Foursquare subordinates location checkins to user location signaling.

This is interesting if you dig into it a bit, and not only only because in each use case the primary function of the tool is subverted to its event-specific use case. (Live socializing changes uses of all social tools in that attention and audience are now both focused and directed on a social occasion. The sociality of networked audiences is replaced by a physically co-present sociality, or situtation, in whose service social tools can now simply provide communication and signaling functions. Interesting, eh?) For twitter, subversion of the application occurs when tweeting is used to signal, or in other words, when talk is subordinated to posting one’s location. And in Foursquare, the subversion of the application occurs when users, instead of checking in to a place, use places to post their location.

Twitter at SXSW = linguistic statement communicates position. The tweet, rather than being conversational, is used to coordinate meetups and to be find-able during the course of the event. The linguistic statement loses its expressive function in favor of signaling.

Foursquare at SXSW = checkin to place inverts relationship between person and place. The business or establishment is not as important as the location of the user; the user’s interest is in sharing his/her location, and the checkin serves as a somewhat arbitrarily chosen means of saying “I’m here.” The Foursquare checkin loses the its function as a means of capturing the value in where people go, in effect to be used for signaling location rather than declaring affinity for places frequently visited.

Now, clearly, geolocation services and applications are useful for different reasons and in different ways, when used during live events. As is the case often with twitter, be it to find friends nearby or to tweet during presentations, etc. But the most common use case for Foursquare is not the live event. It is, I think, in capturing local habits of its users, surfaced in where users go. We all know from checking in that doing so does not mean “find me here” or “I’m free, let’s hang out.” The checkin does not signal social availability here and now. Some have observed that Foursquare’s approach to friending is in contrast with following on twitter primarily for this reason. We neither want to be followed in real life as we are on twitter, nor mean to suggest by our checkins that we are free to meet up.

This suggests that there might be a problem ahead for Foursquare, if it wants to move from point-based checkins to some kind of higher value social utility. And here’s where I am somewhat confused about how geolocation services fit into the social media landscape.

Ideally, a social application extracts and captures some kind of value from a user’s activity, builds relations between the user’s selections and content, and makes those available for others (e.g. the majority of non-participating users) to navigate by and reuse. User checkins with Foursquare really ought to do more than locate users, or provide them with points and badges. They ought to layer taste and preference mapping onto physical maps, thus socializing the otherwise asocial world of place and location (maps). And it would seem that Foursquare, in working with businesses, sees this Yelp-like opportunity quite clearly.

But does the checkin capture any of that relation? Besides habits of frequenting particular places, and aside from the assumption made that a person checks in because she or he likes (vouches for) the place, what value can be extracted from the act of checking in? Checkins are not modified by ratings, are not accompanied by mood icons, or presence-availability signals (eg Free, Do not disturb, party time!). All checkins are equal. As a system that should be adding value to location by extracting subjective preferences and habits, the checkin model is too simple. (The whole point thing is another matter altogether, and certainly adds bias that may later look to have been a mistake. Good for viral adoption; bad to turn location services into game-like experiences. Though points may be put to use by businesses, in which case Foursquare will face other troubles ahead.)

I want to return now to the filter failure comment by Clay Shirky. For in the context of information overload, filter failure may be part of the problem (and solution). But all social web content is not just information. Much of it is communication, and some of it is relational — interpersonally, or socially. And filters applied to communication and to relational interests perform notably poorly.

Filters applied to information, as search results, subtract “unwanted” results per criteria selected or saved by the user. The filter is meant to eliminate and sort out information that’s noisy. The implication being that information overload can be solved by application standing criteria. I want more of this, less of that. But the relation between me and the information I want is only unilateral. And unless I am willing to adjust my filters regularly, I am going to have to choose standing interests and preferences. And when applied to communication and social relationships, standing interests and preferences simply fail to reflect the transactionality and dynamics of social talk and interaction.

Now it occurred to me that there might be a different way to think of filters. I take Shirky’s comment at face value and so to mean filters on content. So, terms, phrases, authors (sources), and perhaps other meta data (time period or recency, trend or popularity, etc). Criteria applied to the data format, or to the information value of the content. Filters would be one-sided, or individual. Each users sets his or her filters according to his or her own preferences.

But the problem of information overload in social is solved by capturing value in relations between users and content (places included), and of resurfacing that value for use in social interaction and communication. It’s intrinsically a two-sided problem, not a one-sided problem. Communication and social relations are each reciprocal and mutually-interested forms of action. Information filters are monological. Social filters would be dialogical. The distinguishing feature of value in social is that it is shared. Shared interest distinguishes social content from the unilateral tastes and preferences I might apply by means of filters. I want multi-polar social filtering, not just an upgrade of the types of filters I use for search.

In the action streams proposal I wrote up and posted recently, the critical difference between activity and action streams was the two-sidedness of the system. I proposed that stream posts be capable of coupling, and that with coupling, posts would be capable of conduction social action. An invitation post carrying buttons for accept/maybe/decline, for example, would be an action stream post used for invitations. This two-sidedness that comes with post coupling permits social interaction because it enables reciprocation.

I wonder if the same issue may be at hand for social tools. If, for example, no granularity in filtering will be good enough to solve the social problems surfaced by geolocation services, unless they are two-sided, reciprocating, and mutually shared. Social filters would then permit us to use relationships to filter information, and would capture real social value from shared interests (mutually recognized interests).

I have thoughts on what this would look like. But I wanted to get this out because it struck me from SXSW commentaries that when social is live and located, as it is in the case of events, social filters are in effect and systems work well. Shared and mutual interests in themes, topics, meeting places, and being found for face to face interaction are tacitly and implicitly approved by those who checkin using Foursquare during an event. Not so, when we drift about town and occasionally check in to a cafe, restaurant, bookstore, or bar. One sided filters won’t suffice here. We need to know more than what we know we like — we need to know what others like. Including, of course, whether they are available to hang out and if so, interested in doing so with us.

Interestingly, I think many of us use twitter to check each others’ availability in this Foursquare use case. Which just goes to prove the point that communication is two-sided, and that social and relational content exceeds the handling capacities of one-sided information filtering. The question “hey, want to have coffee?” is way of ascertaining mutual interest. But of course dm’ing or tweeting that is to risk a “no.” And that’s where a two-sided signaling model might come in handy.

This entry was posted in Marketing, SxD Theory, UX and tagged , . Bookmark the permalink.