Where is social context?

In yesterday’s post on the problem with Facebook Like, I wrote that Facebook is trying to be the sole provider of social context. This got me thinking about the various places that social context may be represented in a networked software system:

  1. in the object or message (which ActivityStreams helps enable)
  2. in the context where it is created
  3. in the contexts where it is seen and used
  4. in each node of the social graph
  5. in sets of social graph elements
  6. in decentralized elements of the social graph (e.g. aggregated/syndicated profile elements)
  7. shared understanding in participants minds
  8. unshared understandings in participants minds

Facebook’s model is seeking consolidation in two places. By replacing a metadata-rich, standardized, ActivityStream based representation of the message with a proprietary API call, Facebook is foreclosing opportunities for the adding of context in creation and in viewing and utilization (items 1-3 in the list).

By acting as the sole provider of social graph and profile services, Facebook is seeking to own those aspects of context (item 4, 5, and 6 in the list). Is Facebook doing anything to enable the exchange of subsets? (item 5 in the list)

But the even the proposed more open models social graph models don’t yet support context well enough, if I understand them right. The portable social graph initiative was based on a simple model that the user wants to bring his or her friends from service to service en masse, and this just isn’t true. It would be great to see an OPML-like standard for friend lists. I don’t think this exists yet. I’m told by standards geek friends that there are standards for 6, but I haven’t seen implementations. Pointers to resources are welcome.

Decentralization, experimentation, and diversity

The future that I am rooting for uses a decentralized model that supports experimentation with many different ways of creating and supporting social contexts. I’d love to hear for people steeped in the standards world to explain which elements actually support this vision, and I’d love to see more development that uses it. Even in a world with a big Facebook, locally specialized needs can be the seeds of disruptive change.

Last, but not at all least, are the aspects of social context that aren’t represented in the nodes and lines in the graph, or the messages and objects that traverse the graph, but among the people. It is always the case that the communication and social understanding, and social misunderstandings and conflicts among people are richer than the bits transmitted by electrons. Social context is in the cultural structures and discourses that people live in, and are largely outside the scope of the communication tools we use, although the tools and use of tools are shaped by culture. Part of the reason for humility and experimentation in design is the knowledge that models and tools are never going fully represent what is happening socially, they can only augment social experience to some degree.

One of my concerns about rise of the one-true-network graph is the potential loss of diverse experiences. By displaying one’s Facebook own friends in a third party site, is this locking one into one’s existing friend network and foreclosing opportunities to meet new people in new contexts? Surely, we sometimes want to share with close friends, but also to use different contexts as opportunities to make new acquaintances.

A recent academic study at Insead video link here shows that people are more creative when they get the chance to live abroad. Cities with diverse populations tend to be more creative, interesting places. Social network concentration is comforting and not a bad thing in itself, but social network lockin is stifling.

I think that the representation of social context is hard, and diversity is healthy, and that’s why I’d like to see standardized elements that can support a lot of decentralized, competitive experimentation.

15 thoughts on “Where is social context?”

  1. Item 1 is a building block for all the rest, except maybe item 8 which is psychological and therefore not in social context. Items 2 and 3 are nonsensical since you are locating something within itself when we don’t even really know what it is to begin with. Items 4 – 6 are representations of object relations, by definition non-contextual.

    Social context, if you can pin it down anywhere, is in number 7 and it alone. JMHO

    P.S. It might not hurt to look at the decades of research on the topic in sociology, anthropology, and social psychology (not psychology) to get social context in the most meaningful perspective.

  2. Larry, thanks for the comment. I clarified the post text slightly to be more clear – the post is mostly about the different places in a software social network where elements of context might be represented, added or used. Items 1-6 in the list are all about representing context in software.

    Talking about software, #2 and #3 make more sense. A poster can, for example, add a tag when making a post, adding context. And later, a reader or software tool can use the tag to filter or otherwise affect what gets read or responded to.

    Talking about software, #4-6 represent ways of adding context within a social graph software representation. There is a set of connections for each individual. Plus, there may be lists of people (friend sets in Facebook, lists in Twitter), and there may be decentralized profile elements that are aggregated or syndicated.

    The last two items are very different from the rest. They are intended as a caution for software designers to keep from thinking that the social networking features built into software are the network and social reality, rather than tools that people use.

    And yes of course, sociology, anthropology and psychology are very important to understanding these issues, although it can’t fit all into each blog post.

  3. fwiw (!) the post being commented on was not talking about preserving original context or intent.

    The post referred to techniques that use metadata about the message to enable recontextualization for the reader/recipient, as well as metadata about the senders and recipients to faciltate the recontextualization!

    A example of recipient-based recontextualization based on a message might geographical, where incoming messages are highlighted differently depending on the (changing) location or interest of the recipient. It does not matter that the sender is in San Francisco, it matters that the recipient is in San Fransisco or considering San Francisco.

    Another example of recipient-based recontextualization based on metadata about people (nodes in the graph) – a tool to aggregate musical references (works and events) from lists of posters that the recipient has pre-selected.

    Metadata can be supplied by people anywhere in the exchange, and used by recipients in manners different from intended by senders. The examples you give regarding action streams are more good ways to code messages, which can then be acted upon differently by different people in the chain.

    The idea about games and rules, I think, complement the these ideas about coding mesages and nodes at different points, by providing more explanatory suggestions about what people might choose to do or might be observed to do.

    This comment was originally posted on A Social Interaction Design (SxD) blog on Web 2.0 & Social Media

  4. I like this — and it helps to have these examples. We often talk in sxd/ixd/ux about loss of context, or preserving context, and of course matters of context include creating context also, as you describe here.

    We might differentiate between context of action and intent, as it pertains to what authors/contributors/participants mean to do when they engage in social media; and context as supplied by display, including navigation, links, resources of other kinds, and of course the object context that constrains usability and object-related interactions.

    I wouldn’t want us to dispense with questions that relate to what the user intended, as those are important to user experience (is the user just checking in at a cafe on foursquare, or is she indicating that she’s up for company?). These questions have to do with social action and communication.

    Meta data relevant to context will include content-related information (maps, geo, places, people), as you describe here. Contextual relevance then touches both on connections between bits of information that are coherent; and actionable connections that permit interaction or communication. The former are easier to delineate; the latter necessarily involve interpersonal, social, and public matters (hence the privacy conundrum around streams and geo, for example).

    We can more easily say that a view provided of a user within a geoloc application supplies relevant context on the basis of the associations and connections it surfaces and renders (chat, placenames, people locations on a map).

    But it’s more difficult to determine the best mix of social constraints (eg show friends only on the map, or exclude people buzzing from home, as they probably aren’t interested in meeting up, hence their position is not a check in and I don’t need to know where they are) and interaction constraints (only show me people on the map who indicated they are available to meetup). Those are matters of social use, which certainly belong as much to context as creating context around display of content and interaction with content.

    Context of action and context of content?

    This comment was originally posted on A Social Interaction Design (SxD) blog on Web 2.0 & Social Media

  5. Interesting post, Adina.

    I definitely agree that “social context” — and perhaps more importantly “sharing context” — is critical to distributed social networking.

    Understanding both the context in which your posts will be viewed and consumed, and also being able to intentionally communicate something about the context that you’re coming from are critical to clear communication and creating shared meaning.

    How you translate your original context (social, physical, mental or otherwise) into a digital representation, I imagine, is very hard — not the least of which because of the asynchronous mode of most social media (i.e. real-time helps only if there’s an audience listening to you at the same time as you post).

    So, one question to noodle on: should the original context or tool that you post from always be targeted to the same downstream audience context? Furthermore, when you intend to interact with a different downstream context, should enter into a different context altogether? In other words, can a tool be used for posting to several contexts successfully? Or, will it simply confuse people by trying to do too much?

    For example: Twitter is largely conceived of as a public channel. You switch to the direct messaging mode in order to send private, one-to-one messages. There is no way, currently, to post to a subset of your public audience, nor to post discreetly to certain people. Is that a success or failure of the tool? Should Twitter private the ability to send to a select subset of your friends, or would that overcomplicate Twitter even more?

    Since the original context of Twitter is public, they offer a secret, off-the-record “cloaked” context to send private messages, but most people still think of Twitter as a public channel. Does that consistency aid in the appropriate use of the tool, or inhibit more nuanced uses?

  6. Good questions, Chris! I strongly suspect that the answer to the question about “cross-posting” is “it depends,” and the use of metadata in the message and the graph will make it easier to experiment in a variety of situations.

    Some reflections based on observation as a user – on the one hand, brute-force cross-posting is annoying – a large portion of my Facebook stream is posted retweets. On the other hand cross-posting from specialized tools is handy – for example, subscribing to people who post from Blip to Twitter. BUT the utility of that depends on good user-controlled filtering and grouping – that blip cross-poster is part of a “music” list I created, which I can dip into when I choose.

    With regard to user subsets on Twitter and social messaging, a few thoughts. First of all, groups don’t need to be private – and can be very useful if discoverable. FriendFeed groups weren’t mass-market by any means but were well-used in some scientific and research communities, for example. I think many of these weren’t private. Second, based on experience at Socialtext, groups are absolutely critical for enabling people to focus attention productively. These groups can be private or discoverable.

    I’m not sure I have an opinion about Twitter itself having private groups, but I think that discoverable groups would be a really good idea.

    Fundamentally, I think that experimentation is needed to figure out what patterns will work in what situations, and standards will facilitate experimentation and competition.

Leave a Reply

Your email address will not be published.