• Print

Why isn’t social media more like real life?

You know the graph. Use it to provide a more human experience.

I finally got around to looking at my personal network graph on Linkedin Labs the other day. It was a fun exercise and I got at least one interesting insight from it.

Take a look at these two well defined and distinct clusters in my graph. These are my connections with the startup I worked for (blue) and the company that acquired us in 2008 (orange). It is fascinating to me that all these years later the clusters remain so disconnected. There are shared connections within a common customer base, but very few direct connections across the clusters. I would love to see maps from some of my other colleagues who are still there to see if theirs show the same degree of separation. This was an acquisition that never really seemed to click and whether this is a picture of cause or effect, it maps to my experiences living in it.

That’s an aside though. What this graph really puts in stark relief is what every social network out there is learning about us. And this graph doesn’t really tell the whole story because it doesn’t represent edge weights and types, which they also know. Social networks know who we connect with, who we interact with, and the form and strength of those interactions.

But this post isn’t a privacy rant. I know they know this stuff and so do you. What this image got me thinking about again is why social networks aren’t using this information to create for us a social experience that is more like our real world, and frankly more in tune with our human-ness.

Social media properties plumb this data to know which ads to show us, and sometimes they use it to target messages to us more effectively. Remember those LinkedIn messages we got with the pictures of our friends? We all clicked on them. But they just don’t seem to be making that much effort to make use of what they know to innovate on our behalf, to improve our experience.

For example, Facebook knows all of this too and yet they continue to cling to the curious fiction that our social life is one giant flat maximally-connected equi-weighted graph. A single giant room where we all stand shoulder to shoulder wondering who all of these strangers are. A place that refuses to acknowledge the nuance and complexity of our real world relationships. And Twitter, for all it’s wonderfulness, does the same thing. And Google Plus? Why are you making me curate circles? You know what they are. At least take a guess at making them for me.

They call themselves social networks, but in terms of how they express themselves to us, their users, they seem to be using the word “network” the way broadcast television does. The experience is more analogous to a vast mesh of public access television networks than with the complexity and richness of real world social connections. You say something and it is presented to everyone, no matter which of those clusters they inhabit. So 10% care and the rest of them filter it.

In the natural world of human-to-human conversation, communication travels person to person, modified and attenuated along the way. Or, in some cases, amplified into a cluster-spanning meme. I think it would be fascinating to see social media properties experiment with recreating some of these more complex dynamics. What if I could “talk” to a well-defined cluster in my graph and see the strength of the signal attenuate rapidly as the distance from that core increased? Not to make it invisible, but perhaps make its volume more appropriate to the another cluster’s contextual center of gravity.

Or, in the inverse, knowing things about my graph Twitter could give me a really nice low-pass filter that gave preference to those in my stream that are “close” to me, or share a common edge type, but who might not be tweeting at high frequency.

There are lots of possibilities along these lines. And I know that a big part of what makes these services useful is their simplicity. Fine. But ultimately, I wonder is all of this network science going to benefit me in any direct way as a user of these services, or is the whole field of data science ultimately about reverse engineering me for sake of advertisers?

I wrote a post a while back about our paleolithic roots and the way we consume media. The “diet” part aside, what I’ve been thinking about a lot since is a digital design sense that caters to our neurological reality. Instead of designing for the convenience of the machines and demand that we adapt, design for who we actually are. Buggy. Tribal. Easily distracted. Full of bias. Curious. Whatever. I’m eager to see a more ambitious approach to design that infuses our digital worlds with more of the nuance and subtlety we find in the physical realm, all while preserving the reach that makes our digital world special.

tags: , , , , , ,
  • Kathysierra

    I would like the option to choose both outgoing and incoming strength and attenuation. But for a near-opposite reason: to help compensate for the filter bubble I am already in. Given that my brain already tends to *prefer* that which comes from those closest to my core network, I would love to say “give preference to the edge”, especially for incoming messages.

    Outgoing, I would like a simple slider that lets me set just how far my message reaches (subject to each receiver’s own incoming preferences, etc.)

    Regardless of how this could and should be used, the idea of more “subtle and nuanced” ways to view and communicate in our many networks is fascinating. It’s just that I find algorithms that filter on my behalf *based on what the algorithm “thinks”* much more creepy. If there is to be subtlety and nuance, I want to do the tuning. The fact that a network “knows” my behavior does not mean it can make the right inferences. And as I said, I might want the system to do the opposite — to see the patterns in my behavior, assume subconscious biases, and take action to override them :)

    No matter what, I want full awareness and control over what the algorithms are doing. Of course it might be a cognitive burden for users to understand and tweak them, so I’d also love some preconfigured sliders from people or groups I trust. One day I might want the Clay Johnson “info diet” suggested EQ, and for another time or context, I want the slider settings that some artist uses…

    • Jim Stogdill

      That’s an excellent point about using knowledge of our graph to puncture the filter bubble. I’ve oven thought of setting up a second twitter account where I follow all the people that I might not want to read every day, but that I feel like I should at least peek at once in a while.

      I’m trying to make a little bit of a distinction here between the usual discussion of algorithms for finding interesting things (here’s a book you might like) and using knowledge of the graph to recreate more real-world-like information propagation models. In real networks signal decay from node to node, or get subtly changed as the relay adds their own spin. In our current online social constructs even a whisper is as effective as a shout, because everyone hears everything. I’m imagining something where a message I put out there has a signal strength measurement as it arrives at each node along its travels, and local UX choices made by those users can determine whether my whisper or shout is loud enough to make it. And things like that.

      I absolutely agree with you that users (both initiating and receiving) should have control. But even harder than control might be making the experience so obvious that you don’t have to think about it any harder than you do when you are deciding whether to whisper or shout in a theater.

      • joecardillo

        I’d like to see a social network develop algorithms that measure signal strength but also measure the relationship strength between people and allow users to easily add their input. For example, one of the niches I’m interested in is how libraries and public repositories archive data and make it available…but that has been largely self driven, and there isn’t much in the way of “celebrity librarians/archivists.” And what about likes, thumbs ups, etc… those could all be harnessed to create relationship values and authority that are more than just “this is someone you might know or an article you might like.”

  • joecardillo

    I think one of the reasons most social networks use the flat, equally weighted graph model is because they are highly influenced by / still think of what they do as advertising and broadcasting, both of which have a history of being one way, or at best two way streets. As you point out, that’s not really how we humans interact.

    • http://twitter.com/jstogdill Jim Stogdill

      Good point. Along those lines I think it’s also because of who they think of as their customer. Weighted graphs would enable more interesting user experience, but the user is rarely the customer in these environments. Which I guess is just another way of restating your point.

  • http://twitter.com/NoelDickover NoelDickover

    “Is the whole field of data science ultimately about reverse engineering me for sake of advertisers?” – Wonderful question! This almost looks like a buried lead here in an excellent post. Yeah, you’d think they would use the social graph to make an attempt at structuring our experience, more than just, “Hey man, have you thought about following this dude?”

    • http://twitter.com/jstogdill Jim Stogdill

      I think it really boils down to who is paying. In enterprise IT, where the IT department is the payer, users never really got good experience. IT buyers cared about features like single sign on, and easy manageability, etc. So users got crap interfaces.

      On the consumer web they can’t give users a crap interface, because that would impact time spent on site, but they don’t necessarily have to provide features that help the users in ways not directly related to revenue creation.

      Also, I think this stuff is just hard. We’ve had to do a lot of engineering work simply to get basic web pages to load at the scale of a facebook. The complexity and subtlety of this stuff is another order of magnitude up the complexity curve. I think it will come, eventually, but it might be slow.