“There’s No Hiding on Facebook”

I was asked to write an editorial for the Guardian’s Comment website about Facebook and privacy. Here’s an excerpt:

Facebook has been repeatedly criticised on privacy grounds. While the company claims it doesn’t sell user information, details are made available to third-party application developers, who account for much of the site’s profits. And researchers have found that personal data can be “leaked” to advertisers and data aggregators, who already collect browsing and behavioural information about people as they move about the web. Just last week, Facebook announced a multi-million dollar deal with Nielsen, known for their meticulous tracking of television ratings and internet metrics.

Even without these partnerships, Facebook makes privacy advocates uneasy. University of Wisconsin professor Michael Zimmer accurately identified an “anonymised” Facebook dataset from the description that it was a private college in the northeast (spoiler alert: it was Harvard). Similarly, the “Project Gaydar” research team at MIT found that gay men’s sexual orientation could be identified based solely on their friends. It’s not just information you make explicitly available – age, partner’s name or favourite film – that identifies you on Facebook. Close analysis of a network of friends can reveal deeply personal details, even with a private profile. These studies suggest that it’s impossible to retain complete control over personal information within a detailed, publicly available network.

I’m happy with how it came out and I look forward to hearing everyone’s comments. The Guardian website right now has a majority in favor of “if you post your personal info on Facebook you deserve whatever you get,” so if your understanding of online privacy is slightly more sophisticated, feel free to leave me feedback.

American Prospect article on the death of Geocities

Phoebe Connolly quotes me in an excellent American Prospect article about the death of Geocities:

Other online platforms began to spring up, and soon GeoCities became a fond memory for most users. Blogger was introduced in 1999 (and purchased by Google in 2003), making it easy for anyone to start a blog. MetaFilter, a community blog, was launched in 1999. The social networking site My-Space was founded in 2003. These services also marked the entrance of a very public form of socializing–where, unlike email or listservs, the conversation, and content, was accessible to those not part of the conversation. In offering a platform for creating online identities, GeoCities started a trend that has been replicated by companies ever since.

But once those online identities are created, are they the property of the users or the corporations that host them? David Bollier, author of Viral Spiral: How the Commoners Built a Digital Republic of Their Own, calls corporate-controlled spaces like GeoCities and Facebook, “faux commons.” For him, true online community spaces are defined by users having control over the terms of their interaction and owning the software or infrastructure. Corporate spaces come with “terms of service” agreements that lay out the rules users must abide by and what control they agree to surrender in exchange for using the product. “Oftentimes corporate-controlled communities are benign, functional, and perfectly OK,” Bollier says. “It’s just that the terms of services those companies have or the competitive pressures of business may compel them to take steps that are not in the interest of the community.”

I really enjoy internet history and although Geocities was something we all made fun of at its peak, it was a useful free hosting solution, and it certainly has a place that should be remembered. It’s sad to think of all those Backstreet Boys fan pages and web diaries disappearing for good.

Transcription: Do not use MT-Stat

So I’m spending a lot of time reading over my transcripts from interviews and meetings. I used two companies. I would absolutely recommend CastingWords for any non-confidential transcripts; they use Mechanical Turk and do a generally excellent job for very little cost.

I would not, at all, recommend MT-Stat transcripts. Their work is very shoddy. I’ll have to redo most of the transcripts myself. It clearly hasn’t been edited, checked, or in any way proofed by anyone, because the transcripts are incoherent. Here’s an example:

Interviewer: I’m constantly finding new areas of stuff that I didn’t know with the children. I’m like, “That’s what I should be working with Mr. Jason.” In fact, massive reaming, he has uh.. he took out bull or cow—

Interviewee: Yeah.

Interviewer: One of them is like, they’re the picker of like, you know, he’s kind of sad rabbit wearing his shoe, you know, grinding some sad rabbit and it’s like.. how do we know… how to postpone the fear of not getting a job and finishing graduate school and duh, duh, duh, duh, and play, what kind of dissertation is that, read another book. [LAUGHS]. It’s totally true like you can.. you can spend.. I can spend five years doing reading just some sort of that as long.”

Useful, right? Clear and precise, well-edited, right? No. In contrast, here’s the ACTUAL transcript, which I did myself, and which didn’t take very long:

Interviewer: Oh God, I mean, I’m constantly finding new areas of stuff that I didn’t know existed, I’m like, “That needs to be worked into the dissertation.” In fact, Matt Groening, he has a– this book called School is Hell –

Interviewee: Yeah.

Interviewer: –and one, it goes through every single, I mean, have you read it? like, anyway, there’s one…it goes through every single year at school, there’s one page for graduate school and every single thing on that page and I’ve had this…I got this book when I was like thirteen, and I got…I constantly think about because they’re all 100% true. One of them is like, there’s a picture of like, you know, his kind of sad rabbit wearing a suit, you know, the Groening sad rabbit and it’s like, how to, how to postpone the fear of not getting a job and finishing graduate school and da da da da da, and by working on your dissertation is that, read another book. [LAUGHS] It’s totally true like you can…you can spend…I could spend five years doing reading just on social status alone.”

I’m not saying this is particularly scintillating material. It’s chit chat. But the transcript provided by MT-Stat is just awful. And I paid them more than $400 for a series of transcriptions (I will also mention that they quoted me a specific number, and then a month later tried to charge me more than $300 more than that, and then claimed that they had never given me the original quote, which I obviously had in my email). Please do not work with this company Mt-Stat (aka MT Stat or MT-S.T.A.T). They’re terrible, especially when a company like CastingWords will do far superior work for the same price.

(I do have some issues with the use of Mechanical Turk, but I suspect that Mt-Stat also uses them and isn’t upfront about it. Either that or they’re using non-English speakers. They are definitely not using professional transcribers. CastingWords has an extensive editing process which contributes to the superior quality of the manuscripts.)

Status and Attention

I just read a good short paper on Crowdsourcing, Attention, and Productivity [pdf] by Bernardo Huberman, Daniel Romero (who’s an intern here at MS Research with us this summer) and Fang Wu at HP Labs. They used a big dataset from YouTube to measure content contribution and attention.

Thesis: People contribute more to content sites like YouTube when they receive positive attention, and a lack of attention causes people to uploading less content and, in some cases, to stop contributing altogether.

Those contributing to the digital commons perceive it as a private good, in which payment for their efforts is in the form of the attention that their content gathers in the form of media downloads or news clicked on

This isn’t an entirely surprising study. There’s lots of evidence that status is a major motivator for online participation– not just academic studies, but in general game and social software design (see my Tumblarity and FourSquare posts for recent examples). That’s why every arcade game has a leaderboard and why Yelp has an elite classification and why I’m writing a dissertation on the topic.

But Huberman et al. use “status” and “attention” synonymously, which interests me. They operationalize “attention” as “number of views.” On YouTube this makes sense, since the highest-viewed videos bubble up to the index pages, and videos that crack the top 100 in their category get “honors” that appear on the statistics part of each individual video’s page. So on YouTube, attention maps fairly neatly to status. And I think this is true for most sites that have quantifiable status metrics based on views, followers or whatever the site labels it.

On other sites, of course, status might be linked to skill (high scores, artistry, writing reviews of new restaurants), looks (clothing choices, aesthetics, makeup skills), wealth, whatever. But if I’ve created an amazing Polyvore collage, it’s only a status symbol if other people see it (and I’ll be more likely to create more if people view my existing collages). Similarly, although time doesn’t map directly to attention, having a low Slashdot number or an “oldschool” Upcoming badge is meaningless if nobody knows about it. I need to have recognition for my wealth, skill, or looks in order for them to function as status within a group.

I’d argue that attention is an important part of the status metric; but I don’t think more attention always translates to more status (the term “famewhore” comes to mind). But perhaps the attention is what encourages people like Julia Allison or Nick Starr to continue living public lives, even as they receive a great deal of negative attention at the same time. I would be interested to see if attention of any kind correlates with participation, or whether it is only positive attention; if the YouTubers had thousands of hits, but an equal number of vitriolic comments, would they continue to post videos?

Finally: We hear a lot about the “attention economy” or “publicity culture,” in which the most valued skills are those which increase attention. And many people decry this culture for bubbling-up sensational, sexual, or violent content– or just short bursts of info-nuggets– rather than meaningful, thoughtful, difficult ideas. I’d argue that what attracts attention is culturally specific and so we can’t automatically assume that an attention economy leads to lowest-common-denominator content. (Another assumption I’d like to see tested.)

It’s a beautiful day today and all I want to be doing is riding my bike around outside.

Technological Woes, Continued

… And now my Twitter account got suspended, probably because I have a link to tiara.org on my Twitter page. It’s under review, but Twitter doesn’t have the greatest appeals process for suspected spammers. I must have really pissed off some patron saint of technology out there.

We’ll see how long it takes for my iPhone to stop working or my laptops to explode.

My apologies

Somebody hacked my site yesterday and added iFrame code to two pages, start.html and lj_bib.html, which redirected users to some shady-ass spam/malware site. So the entirety of tiara.org has been blocked. I’ve fixed the problem, changed my password, and submitted the site to Google Webmaster tools for re-analysis, so everything should be back to normal in a day or two.

I’ve traced the hack to a major Dreamhost security breach, where 3,500 account passwords were “compromised,” mine probably among them.

I was impressed that Google Safe Browsing identified the exact pages that had been hacked and the links I should look for, which made it (relatively) easy to fix.

Some celeb twitter tools

So I’m working on a project about celebrity use of Twitter. Here are a few recent tools I’ve found to be endlessly entertaining when looking at celebs, status, and social norms on Twitter:

• Who Celebs Tweet, with the tagline: Have they tweeted you? I find this the most interesting because they have a very clear demarcation between who is a celeb and who isn’t. Like, according to them, Heidi Montag is not a celeb. I don’t necessarily think she should be a celeb, but to deny that she’s famous seems odd. Maybe the proprietors never read the tabloids.
• TweetingTooHard.com - this is sort of like Texts From Last Night minus all the drunk skulduggery and adding a lot of self-aggrandizing obnoxiousness. Tops now is “fan belt light came on in the 911 so now I’m driving the Cayenne Turbo S - the backup, backup car. Trying not to think about the Tesla…” That’s pretty bad.
• Truth Tweet attempts to verify celebrity Twitter accounts, using all sorts of sources to do so. Extremely useful for my purposes (e.g. nerdily making lists of what signals celebrity “authenticity” on Twitter).

Tumblarity and Quantified Stand-ins for Social Status

So Tumblr launched its Tumblarity index last week. (Here’s Gawker’s obnoxious take.)

Tumblarity is a metric that measures one’s popularity, or degree of Tumblr-ness, depending on who you ask. It’s displayed on a nifty stats page modeled after The Feltron Report. Tumblr hasn’t revealed exactly what it uses to calculate this number, but it certainly includes number of posts, followers, likes, and reblogs. My Tumblarity is 3 (which is very, very sad, in case you’re wondering), but if it were higher, I could see where I rank in the top 50,000 Tumblogs or in my local area.

I stole this image from download squad so you can see what a slightly better Tumblarity score looks like:

Like number of Twitter followers, Tumblarity is a quantified metric: a number that stands in for more complex social phenomena, like popularity or status. Tumblr helpfully includes leaderboards to make it extra-easy to compare Tumblarity with your friends, rivals, and frenemies, causing tech dorks pundits to complain about the “popularity contest” aspect of the feature.

A few basic things about quantified metrics:

1) They are always stand-ins for more complicated status measures. A single number cannot possibly convey the nuances involved in social status and social hierarchy (e.g. Why do so many people read your Tumblr? What group/subculture/community does it appeal to? What actions do you take to maintain this status? What does your community value that your blog provides?).

2) Techie/geek/engineer types love quantified metrics precisely because they facilitate comparison. Several of my informants talked about how Silicon Valley types love talking about VC funding and valuation because they allow people to attach clear numbers to companies in order to rank them (and convey status on their CEOs, VCs, and employees). (See also those obnoxious “30 under 30,” “100 Most Influential People in SV,” ranking lists.)

Clearly, people in general also like comparative metrics — see the high score lists at arcades, the Fortune 500, the Best Dressed list, etc.– but they’re becoming increasingly prevalent in social software (built by nerds).

3) Quantified status metrics spur competition and therefore increase user action. I’m assuming Tumblr is trying to reward certain types of behavior, which in this case is pretty obvious: Tumble a lot, follow lots of people, reblog a lot = spend more time on the site = benefit to the company.

4) Social status is an under-studied, under-rated aspect of product design and motivation for user action. This is the subject of my dissertation and I’m seeing increasingly explicit aspects of this in social software (which: yay!).

But let’s not fool ourselves that an algorithmically-generated number “is” social status. I’m sure there are tons of sub-groupings and communities on Tumblr that value different things. I’m sure the top 100 Tumblr users are popular for different reasons. I’m sure there are Tumblr conventions and social mores that mark someone as an insider or outsider, a newbie or a jaded user. There are many good business reasons for the company to boil this down to a single number, but it only tells us a little bit of the overall story. Tumblrites: ideas?

In Lieu of Writing a Real Post, I Post a Video

Here’s the @100interviews video of me from SXSW. Note I look exhausted. That’s because it’s halfway through South by and I was exhausted. Good basic overview of my dissertation.

Alice Marwick 100 Interviews

Loopt, Locative Media, and Prescriptive Social Software - Part Three

This is the third part in a series of articles looking at how mobile social software prescribes certain types of social behavior. Part One discusses foursquare; Part Two discusses Brightkite. This part looks at Loopt.

Loopt

Loopt is a website and mobile app that, in the words of their website, “transforms your phone into a mobile compass. Connect with friends and get alerted when they are nearby. Share your location, photos and comments with friends and social networks. Explore places and events recommended by friends and Yelp.”

Loopt is a lot like Google Latitude: rather than voluntarily checking in to a particular venue, and broadcasting that information to a social network (the foursquare/Brightkite model), Loopt automatically tracks your location and shows that to your friends. But while Google Latitude is building a platform, Loopt is a standalone application.

My experience signing up for Loopt immediately made me uncomfortable. When you add someone on Loopt, it sends them a text message, which is much more intrusive than an easily ignorable email (the reason for this is that you don’t provide Loopt your email address right away; more on this in a bit).
While many social software applications use the term “friends” in a very generous sense (Facebook: everyone you’ve ever met), “friends” on Loopt should presumably be people you really trust rather than people you’d be comfortable with hanging out in a bar. The software doesn’t indicate this, though; it goes through your phone book/contacts list to find potential friends and suggests that you add all of them. Two people friended me back– both very close friends. And both of them, within minutes, had IMed and texted me to say that it was sort of “creepy” to see exactly where I was on a map. If all three of the people involved in this transaction found it creepy, then it’s likely to be violating some sort of social norm. (And, several days later, only one additional person has responded to my friends request.)

Friend: wait, did it check me in when I started the app?
me: i guess it must have
it says you checked in like 1 minute ago
did it send you a text ?
Friend: heh, ok. didn’t realize that would happen
me: that’s kind of annoying

While you can configure Loopt not to check you in automatically, it does so by default. Additionally, if you turn this feature off, the next time you open the application, it asks you to turn it on again (I suppose that without this auto-tracking, the app is fairly useless, but it makes it clear that the preferred behavior is auto-checkin).

There are fairly intense implications of always knowing precisely where your friends are, which is not information I would always want to have. This is particularly true when you’re talking about someone you’re dating. You wouldn’t necessarily ask your girlfriend where she is all the time, but once that information becomes available, there’s a temptation to use it. Of course, you can turn off Loopt (or Google Latitude) whenever you want, but if you usually use it, turning it off implies that you have something to hide.

Loopt also has an intriguing feature called “Loopt Mix” which connects you with people you don’t know in your neighborhood. To use it, you provide Loopt with an email address. The email address you provide will be used by random Loopt Mixers to send you messages, so Loopt says, “Make sure to remove any contact info that may have been added in your email signature!” This is an annoying user experience which signals something about possible privacy/safety violations.

The next screen gives you various options for filling out your profile: name, picture, “About Me,” age, gender, interested in networking, friendship, dating women, dating men, tags, and “featured communities” - TechCrunch, Imeem, Rock the Vote, and the National Resource Defense Council. Once you’ve filled this all out, you can view people in your area. Here’s what this looks like in practice:

 

In other words, this is a great example of people using social media to hook up. Loopt Mix displays your picture and info to people near you, who can then anonymously email you using the tool. The vast majority of people using this app in my area, as you can see from the second picture, are gay men (I live in San Francisco — this may differ in other cities). I’m not surprised about this; the safety issues involved in people putting their locations and pictures of themselves online are gendered, and other hook-up sites like Craigslist Casual Encounters, Adult Friend Finder, and Manhunt are overwhelmingly used by men. One Loopt Mix user’s profile recommended Grindr, which is an iPhone app specifically designed for M4M hookups — a clear sign of the audience Loopt Mix is appealing to in SF.

On the website, there’s a “journal” feature which lets you view where your friends have been over time. There is also some recommendation component, although none of my friends have shared any information yet, so I can’t evaluate this feature. If it’s entirely dependent on information provided solely by my friends– people who I’m close enough to feel comfortable with them knowing my location all the time– it won’t have comprehensiveness like Yelp or any other venue-based review database.

So what does Loopt value?

• Persistent knowledge of location of friends (always-on location tracking)
• Persistent self-disclosure of location
• Meeting people based on location (for whatever, although in practice seems to be for sex)

The overwhelming value here is that location is a useful piece of personal information that should be revealed. I suppose this an obvious feature of locative social media, but whereas Brightkite and foursquare voluntarily ask for this information, and tie it to a specific location, Loopt’s ideal seems to be seamless location tracking. While this is certainly valuable, as with Brightkite, it’s not necessary actionable. The automatic nature of the app means that you never know whether someone wants to be “found” or not. Generally, while Loopt doesn’t prescribe social behavior, it seems likely to give rise to all sorts of etiquette/social problems if widely used. The potentially negative social implications of the technology seem to outweigh the (non-obvious) advantages of using it.