The Noisy Channel

 

Paul Adams’s Presentation on Social Networking

July 8th, 2010 · 23 Comments · Uncategorized

This presentation by Paul Adams, lead for User Research for Social at Google, has been making the rounds in the blogosphere. It’s long (over 200 slides!) but well worth the time to read it, even if you’re already familiar with the ethnography of online social behavior. It touches on all things online and social, from the theory of strong and weak ties to social influence to privacy. Enjoy!

23 responses so far ↓

  • 1 Daniel Lemire // Jul 8, 2010 at 8:19 am

    Great slides indeed.

    Two points:

    *) What’s up with the slider bars on the slideshare gadget?

    *) Is there a video of his talk available?

  • 2 Daniel Tunkelang // Jul 8, 2010 at 8:58 am

    I think the slider bars may be a result of uploading to SlideShare as a pdf rather than as a PowerPoint or equivalent. Not sure.

    Don’t believe there’s a video. Wish I’d seen the talk live myself.

  • 3 jeremy // Jul 8, 2010 at 8:06 pm

    The overview of the social research (Dunbar number, strong vs. weak ties) was interesting. But the fundamental premise of the slideshow kinda rubs me the wrong way. In particular, I have a real problem with slide #14:

    “The problem here is that these are different parts of Debbie’s life that would never have been exposed to each other offline were linked online.”

    My primary issue is this:

    The same problem (online linkage of things that would have remained separate offline) exists all over the web, not just with social networking. Take Amazon. Purchasing behaviors of mine that would have never been exposed to each other offline are now linked online. i.e. offline I buy my books from a different place than I buy my lawn furniture. But now Amazon sells both and links both purchases and begins to build a profile of me based on that online linkage… and having them build a profile of me is something that I don’t particularly want them to do.

    Or take Google as another example. Same issue. Offline, my search for a particular book in the library doesn’t get tied to my search for some fast food restaurant, which doesn’t get tied to my Yellow Pages search for health care providers. But online, Google makes those linkages. Without my control, or necessarily desire.

    Sure, I have the power to see all the queries that I’ve run, maybe even delete some of them. But that’s still an all-or-nothing proposition. I.e. just like I have the power to add or remove friends on Facebook, I can add or remove queries from my history. But I don’t really have any good tools to delink my various queries from each other, to prevent Google from making links and connections that I don’t want Google to make.

    So slide #14 (well, the whole intro up to that point, really) rankles me because it seems to me that if you want to get better at providing tools and interaction designs to delink subsets of one’s social network, you probably should have been working for years on giving the user the same sort of transparency and delinkage capability in one’s own search activity. And yet this has not been done. So I attach (link?) less credibility to these slides, saying that it now needs to be done for social networking.

    If you’re going to work on the problem of “unintended online linkages”, then work on that problem. The chance to do this sort of user-facing, multi-faceted information management research has been there for years. My search activities are just as varied as my social activities.

  • 4 jeremy // Jul 8, 2010 at 8:44 pm

    And at the risk of more self promotion (two days in a row now in fact — I feel dirty), Gene and I just wrote an IIiX paper about one possible way you could allow the user to start to transparently link and de-link their own search activity.

    http://palblog.fxpal.com/?p=4192

    Different groups of friends are to social networking activity, as sessions (especially explicitly created and managed sessions) are to search activity.

  • 5 Daniel Tunkelang // Jul 9, 2010 at 1:39 am

    Jeremy, it sounds like what you want is to be able to make anonymous transactions with merchants–the equivalent of always using cash in the offline world. If people get serious enough about privacy, I can imagine generalizing one-time credit cards to one-time identities. There’s still the matter of your shipping address giving away your identity, but there are solutions for that too. But it’s hard for me to imagine mainstream consumers trading off that much convenience for privacy.

    As for your desire to manage the linkage of queries in your search history, I wonder if there’s a significant demand for it (as opposed to the simpler demand to search without leaving a history). I can’t really imagine users investing effort into annotating their search histories.

  • 6 Kuldeep // Jul 9, 2010 at 6:14 am

    I think more social networking makes you sociophobiatic as people think online avatars as real

  • 7 Dave Fauth // Jul 9, 2010 at 8:32 am

    Excellent presentation.

    Credit card companies combine our purchases and begin to build that profile of what we buy, where, when, etc. I’m not sure why we are surprised when on-line retailers do the same.

    As far as search histories, I would agree that anonymous searching is an easier answer implementation than an end-user (my mom) annotating search terms.

  • 8 Daniel Tunkelang // Jul 9, 2010 at 9:04 am

    Kuldeep, I can’t even tell if you’re joking! These people don’t seem to be.

    Dave, long time! And yes, I absolutely agree with you. There’s nothing unique about online retail, only a matter of degree–and I’m not sure how Amazon compares to Wal-Mart even in that regard.

  • 9 jeremy // Jul 9, 2010 at 11:25 am

    Re: Credit cards. At least offline, people have a choice about whether they want to use a credit card (which itself is a form of “interface” into the online world) or not. Online, you do not have that same choice.

    And if you’re worried about credit card companies doing linkage, i.e. aggregating and mining your purchase behavior, there is an option: One can get multiple credit cards from multiple companies, and always use one card for one kind of purchase, and a different card for a different kind of purchase.

    Actually, that brings us to a related point with social network friend circles and identities. If people really wanted to do this whole social delinkage thing, they can already do it. Just like using multiple credit cards, it is possible for a person to create multiple identities on Facebook. Then they can reject any friend requests that come to the “wrong” identity, and initiate the “right” friend requests from the “right” identity. Yes?

    As for your desire to manage the linkage of queries in your search history, I wonder if there’s a significant demand for it (as opposed to the simpler demand to search without leaving a history).

    So by analogy, given that people can already set up multiple accounts on Facebook, one has to wonder if there is significant demand for what these slides are proposing. Sure, when you interview people offline, they talk big talk about having multiple circles of friends. But are they motivated enough (do they care enough) to really set up multiple accounts?

    This is only anecdotal evidence, but from the 435 connections that I have on FB, I have one friend.. just one.. that has set up multiple accounts for himself. The accounts are named “Firstname Lastname (work)” and “Firstname Lastname (personal)”. 1 person, out of 435. And he has only set up two accounts, not the 4-5 accounts/groups that are being proposed in the above slides.

    So doesn’t that also demonstrate that people may talk about having different groups of friends, but only 0.23% of them (less than 1%) care enough to actually do something about it. And if that’s a big enough percentage of the user base, I have to imagine that they are at least 0.23% of Google search users that also want to be able to manage their query linkages in some form.

    Because remember, query linkage is not about privacy. It’s about personalization. Google does personalization based on your search history, correct? And right now, because it gives you no way to give feedback on which queries go with which, it has to make inferences and guesses about how to do that personalization, which of your prior queries to apply to the current query. I can easily imagine that 0.23% of Google users are willing to give feedback on their query histories, to make that process better.

    But the all-or-nothing query linkage solution just doesn’t work or help. 100% query linkage is just as bad as 0% (anonymous searching) query linkage, because in both cases it messes with your personalization, which messes with the relevance/quality of your search results.

    I can’t really imagine users investing effort into annotating their search histories.

    Perhaps because annotation isn’t the right mode/interface. I’m not talking about having the user go through every query they’ve ever done and hand label it. Perhaps there is something more transparently contextual that can be done. Inline linkage feedback or something like that.

    One of the fundamental issues in HCIR is how best to create interfaces that elicit some of this delinkaging information. Google bought Kaltix (personalization engine) in 2003.. that means they’ve had 7 years now to start practicing with transparent query linkage and delinkage, on that 0.23% of their user base. To create interfaces that, even if they’re only used by that small subset, at least allow Google to know how to create HCIR techniques that more easily allow users to manage their linkage information. With that knowledge in hand, they then could be in a much better position now, and know how to apply it to social network linkages. Then, because your HCIR interfaces around query linkage have improved over 7 years, and you’ve raised the numbers of users who use them from 0.23% to 11%, you can also have much more success in the social network arena, and get more users to want to “annotate” their circles of friends, instead of the 1 in 435 that do it now.

  • 10 jeremy // Jul 9, 2010 at 11:36 am

    Quick correction/clarification on one my paragraphs above:

    So doesn’t that also demonstrate that people may talk about having different groups of friends, but only 0.23% of them (less than 1%) care enough to actually do something about it. And if that’s a big enough percentage of the user base for Google to invest all this time, effort, and money in creating yet another entire social network (GoogleMe), I have to imagine that they are at least 0.23% of Google search users that also want to be able to manage their query linkages in some form, and that Google would be willing to spend time, effort, and money coming up with the HCIR techniques for dealing with those users.

  • 11 Dave Fauth // Jul 9, 2010 at 9:33 pm

    Jeremy,

    I’m one of the 0.23% of the people who haven’t set up multiple social media personas (multiple personality disorder). I initially tried by using Twitter for business type interactions and Facebook for more personal interactions. Over time, the two groups of people have crossed over and the lines have blurred. It’s too much effort right now to start over and recreate a new work persona or a new personal account.

    For consumption of social media data, I would like the ability to categorize my “friends” or “followers” into groups. In FB, I would like to be able to look at only a subset of the more important people and then occasionally look at the fringe friends / way outer circle. I can’t do that now.

    Twitter does allow that in a way through lists. Of course, that is a more loosely coupled connection as I don’t have to automatically follow people who follow me.

    For creation of content, the design and interfaces aren’t there to easily do what slide 214 is trying to show. Lot of room there for research and improvements.

  • 12 Dave Fauth // Jul 9, 2010 at 9:37 pm

    And as far as contacts on my phone (slide 90), I want them in alphabetical order because that’s how I’m going to either text or call them. I know their name and I know how to look them up by their name. I don’t need them grouped.

  • 13 jeremy // Jul 9, 2010 at 9:48 pm

    I’m one of the 0.23% of the people who haven’t set up multiple social media personas

    Wait, I’m confused.. you mean that you’re one of the 99.77% of people that haven’t set up multiple social media personas, correct?

    For consumption of social media data, I would like the ability to categorize my “friends” or “followers” into groups. In FB, I would like to be able to look at only a subset of the more important people and then occasionally look at the fringe friends / way outer circle. I can’t do that now.

    Right.

    So all I am saying is that for consumption of web media (via search), I would like the ability to categorize my information needs into groups.

    Furthermore, I am saying that even though only 0.23% of people set up multiple personas in social media, to manage their groups, Google is still moving forward with plans to give people tools to do this partitioning. My complaint is that I’m sure that there are at least 0.23% of Google users who want to be able to manage their information seeking behaviors, by splitting their tasks into multiple groups, or sessions.

    So given that Google has many orders of magnitude more search data (not only web pages, but user actions, user histories, etc.) why would they not use that data to start practicing their interface / UX and corresponding HCIR algorithm design on the problem of search query stream delinkage, i.e. allowing the user to explicitly do “grouping”.

    Had they been practicing/learning how to do that for the past 7 years, i.e. learning how to do UX and algo designs that allowed the user to easily delinkage a certain kind of information (search query), they would then have the experience and knowledge that could be applied to a different kind of delinkage (social network).

    You’re right.. there are not designs and interfaces to easily do what slide 214 is trying to show. But the basic concept of user-enabled delinkage is what slide 214 is all about. They’ve had 7 years (since buying Kaltix) to practice creating interfaces that get users familiar with the concept of delinkage and grouping. So why have we seen nothing? Why do I still not have the tools that I want to have, to search for web information the way I want to search for it, to use Google’s full potential?

  • 14 jeremy // Jul 9, 2010 at 9:57 pm

    That’s why I don’t attach a lot of credibility to these slides. Great, all this delinkaging social grouping sounds very nice. But there have been other opportunities to build and experiment with such delinkage interfaces, which Google has not pursued. So I’ll believe it when I see it.

  • 15 Daniel Tunkelang // Jul 9, 2010 at 10:14 pm

    Not all delinkage problems are created equal. As I said earlier, I think that most people who would be concerned about managing the representation of their interaction history have the simple desire to not have such a history stored anywhere. I personally like the idea of users taking a more active role, but it’s hardly low-hanging fruit–and it’s not what these slides are about.

    In contrast, the slides discuss the problem that happens from unexpected connections across your explicitly represented social network. It takes a lot less work to convince users that this is a problem, and that means it’s a lot more plausible to convince users to invest at least a little effort to mitigate it.

    Still, no one’s done it yet, and it remains to be seen whether Facebook, Google or anyone else will deliver a compelling solution to this fundamental problem of online social networking.

  • 16 jeremy // Jul 10, 2010 at 3:14 pm

    I’m not saying that all delinkage is 100% equal. And I’m not necessarily saying that are the exact same number of people who would engage in explicit delinkage during their exploratory search sessions as there are people who would create social friend groups.

    I’m saying that the number who would so engage with their own information seeking sessions is not zero. And even 1-2% of the hundreds of millions of Goog users is enough of a population to be able to start to learn some of the basis principles of UX and algorithm design around the space of explicit, user-driven delinkage.

    By analogy, not all faceted search is equal. Some domains are purely categorical. Others have continuous-valued price or time ranges or geographic dimensions, which need to be somehow quantized or otherwise fit into a faceted schema. But instead of doing no faceted search at all just because your current domain is categorical only, if you at least do something you start to learn how it works. You start to learn how to do it better, what the advantages and tradeoffs of drill-down vs. parallel section are. For example. So when you’re faced with a new, real-valued faceted domain, much of what you’ve learned will be transferable.

    And Goog *has* those hundreds of millions of users, i.e. they have a base from which they could already have started to practice and learn how to do user-driven delinkage. They didn’t. Why not? I think it’s because it’s not in the Goog DNA. Giving users more explicit control about what is happening and what they can make happen is just not Googly. They’re also not really into research, i.e. looking forward a few years beyond the current user gradient and developing something for only 1-2% of the current population, with the goal of understanding how that idea works, so as to apply it to a large population later. Google just doesn’t seem to think that way.

    It seems like that’s starting to change and I keep looking forward to the day, Daniel, when you can tell us more about some of those changes. Because I welcome and applaud them.

    But it’s not like the opportunity wasn’t there, for years, to start developing and researching and figuring out some of the basic principles of transparent, user-facing delinkage. That’s the jist of what I’m saying. The opportunity to do core research, and figure out how to solve some of the fundamental issues, so that you can then apply that knowledge to a wide variety of problems, including social networking, has been there for at least 7 years now. Your points about exact equality of the domains, or relative numbers of users that would engage in the practice, are all true. I don’t necessarily disagree with those as raw facts. But I see them as orthogonal to the larger picture.

    But I’m starting to repeat myself, so I should stop. Continue this via fisticuffs in Geneva? :-)

  • 17 jeremy // Jul 10, 2010 at 3:46 pm

    Sorry, one more quickie:

    In contrast, the slides discuss the problem that happens from unexpected connections across your explicitly represented social network. It takes a lot less work to convince users that this is a problem, and that means it’s a lot more plausible to convince users to invest at least a little effort to mitigate it…Still, no one’s done it yet

    If you’re interested, this point is one that I’d like to fisticuff (or buy ya a drink) over. Because some of these options are already there. For example, if users are concerned about their gay bar coyote ugly photos being seen by the 10-year olds in their classroom, as was mentioned in Paul’s slides, then there is already a way to control that:

    http://blog.facebook.com/blog.php?post=11519877130

    So if users really want this sort of option, why is it that not only do they (probably) not use it that often, but that they (probably) don’t even know about it. Did you know about it? You can even do the same thing with your status updates, on a per-status-update basis.. keep certain people from seeing them.

    Would be interesting to chat about this.

  • 18 Daniel Tunkelang // Jul 10, 2010 at 4:13 pm

    First off, I am thrilled to be achieving a such a high comment-to-post ratio again–I was afraid I’d lost my touch!

    Back to delinkage. I like what you and Gene have to say about supporting exploratory search, and I feel transparency buys a lot of return there. And I’ll readily concede that Google has a lot of work to do in order to better support exploration–it’s an area where I’m hoping I can help. Certainly Google has done a lot more historically to prioritize p@1, and average precision in general, and most efforts to support exploration are of more recent vintage.

    As for my “no one’s done it yet”, my point is not that no one has offered the flexibility, but rather that no one has made it feel effortless enough that users actually take advantage of it.

    Anyway, I’ll see you in Geneva in two weeks. Happy to continue this discussion over chocolate, cheese, and fisticuffs. :-)

  • 19 jeremy // Jul 10, 2010 at 4:22 pm

    As for my “no one’s done it yet”, my point is not that no one has offered the flexibility, but rather that no one has made it feel effortless enough that users actually take advantage of it.

    Which is all the more reason Goog should have been practicing on personal query history and personalization delinkage for the past 7 years, so that by now they’d have a decent understanding of the issues and the space and the UX concerns. :-)

    Chocolate, cheese, and fisticuffs it is! See ya there.

  • 20 Weekend Reading // Oct 8, 2010 at 8:00 pm

    […] Social Networking Presentation: by Paul Adams. A great read.Deep insights into how the offline and online works meld together. […]

  • 21 Why the future is all about groups « Foxfly – Free groups on the fly // Nov 9, 2010 at 2:51 am

    […] a great job at highlighting the way we manage our social groups offline in this presentation on “The Real Life Social Network.” […]

  • 22 Why the future is all about groups // Jan 29, 2011 at 10:10 am

    […] a great job at highlighting the way we manage our social groups offline in this presentation on “The Real Life Social Network.” […]

  • 23 Attention vs. Privacy // Jul 24, 2011 at 11:22 pm

    […] Privacy is a big deal, especially for Google — and particularly in the context of rolling out a new social network. Still, I’m not persuaded that privacy is the only or even the primary concern motivating the concept of social circles. […]

Clicky Web Analytics