« Skull Job | Main | Ben Fry Redesign »

Facebook Pulls Their Thumbs

Coming out of the IA Summit this year, and following thought-trails down twitter-paths, new friends' blogs etc. It's obvious that design patterns were a huge topic at the Summit. Social design patterns, possibly even more so. Also heard (and stated myself) a couple different times: sharing anti-patterns may be even more critical right now than collecting and cataloguing all the possibilities.

My own talk featured a handful of reputation patterns (coming soon to the Yahoo! Design Pattern Library—I promise!) But, if you were listening closely, you probably picked up on a theme. The patterns themselves (Points, Levels, Trophies, Badges, etc) have incredible potential to do harm to your community dynamic. So, taken another way, these patterns can also stand as effective cautionary anti-patterns.

Anyhoo… all of this made me vow (to myself, and now to you) to try and share more after-the-fact analysis, dissection and anti-pattern identification of social media trends. Have you ever noticed how good our industry is at noticing (and praising, and downright trumpeting) “the new” in social media? New product launches get extensive TechCrunch coverage. New Facebook features put the analysts in overdrive. But what about the quiet re-jiggerings? Or outright feature failures?

This, for example, was passed around at work recently. Inside Facebook has noted that FB recently pulled the thumbs-up/thumbs-down voting mechanisms from the News Feed interface. IF notes:

While Facebook has always allowed more general Facebook preferences (like More About These Friends/Less About These Friends), this feedback was much more granular and potentially powerful.
And also:
It's too bad Facebook wasn’t able to get more value out of the explicit preference data users were generating. For now, Facebook will need to rely on more implicit data.
I think potentially powerful is the key phrase here. What follows is a somewhat-cleaned-up version of my take on the removal.

(It's worth noting, too, that I spent a certain amount of time last year trying to document and rationalize a series of 'Best Practices' for Voting & Rating mechanisms at Yahoo! Thumb-ratings as inputs to recommender systems is a design problem that I've thought a lot about, and that thinking informs this response.)

I hadn't noticed the removal of the thumbs from FB News Feed until the Inside Facebook article appeared. I strongly suspect that they were removed for one specific reason: underwhelming performance. They were using thumbs up/down in the manner closest to what we prescribe (as explicit input into a recommender system) but even then the value proposition has to be clear for users, and it's got to be apparent that the value you'll get out will, at some point, justify the work you're putting in. I bet no-one was using them.

Or at least not in the manner that FB expected. As evidence, I offer up the comments on the Inside Facebook article: the only people bemoaning the loss of the feature seem to be doing so on the grounds that it was useful as a place-keeping feature—delineating read from unread entries. (Beware ambiguous inputs!)

Users probably didn't entirely know that the thumbs were feeding personalized recommendations. It could also be related to this (from those afore-mentioned Ratings Best Practices):

What type of items are rateable?

Items that should be rateable by the community share some common traits. Here are some of them:

Rateable items should have some intrinsic value.

We should never ask users to provide meta-data (basically 'add value') to an item who's own apparent value is low. Or, more specifically, we should be careful to only ask for user participation in a way that acknowledges an item's intrinsic value: it might be okay to ask someone to give a Thumbs-up rating to someone else's blog-comment (because the 'cost' in doing so is low—basically, a click) but it would be inappropriate to ask for a full-blown review of the comment. There would be more effort and thought involved in writing the review than there was in the initial comment!

Rateable items should persist for some length of time.

Rateable items must remain in the 'community pool' long enough for all members of the community to cast their vote. There's also little use in asking for a bunch of meta-data for an item if others cannot come along afterward and enjoy the benefit of that meta-data.

A Rateable Item's Lifespan

Highly ephemeral items, such as News articles that disappear after 48 or 72 hours, probably aren't good candidates for rating. Items with a great deal of persistence (on the extreme end are real-world establishments such as restaurants or businesses) make excellent candidates for ratability—furthermore, the type of ratings we can ask may be more involved. Because these establishments will persist, we can be reasonably sure that others will always come along afterward and benefit from the work that the community has put into the item.

Unless FB users understand the longer-term benefits to rating News Feed events (ie. that they're training the system) then they may just look at those thumbs as a long, undifferentiated (and endless!) stream of tiny rateable 'things.' Unfortunately, any one of those things probably doesn't have either of the qualities I describe above: most of them have very low perceived value (“Tom left a comment on Jody's wall!” practically has no enduring value beyond the time it takes to read and digest the information.); and—almost by definition—the items don't persist for any appreciable amount of time. They're highly ephemeral.

Juxtapose this with Tivo (which is kind of the canonical example) where thumbs are used to rate TV shows for recommendation purposes. The programs being rated have high value (ie, they cost money to produce; we get entertainment value out of them) and they're highly persistent. (They'll recur at the same time next week, and in reruns, and in syndication, etc etc. For all intents and purposes, tv episodes are immortal.) And Tivo's whole user experience (including user-education movies that ship w/the unit, their printed manual, heck the dang remote has 'em hard-coded on there) are oriented around the Thumbs up and down-voting. I'd venture that thumb-voting, and the recommender system are a huge part of why many people buy Tivo in the first place. (Okay, that plus 'pause live TV.')

When it comes to explicitly-input recommender-systems we should acknowledge the limitations of folks' interest in 'feeding the machine.' If they understand the benefit… and they think that the work they'll put in will at some point be worth something to them, then folks will play along. Unfortunately, there's a much easier strategy for 'tuning out' facebook updates that I don't like: I just start to ignore the facebook feed altogether.



This page contains a single entry from the blog posted on April 18, 2008 4:36 PM.

The previous post in this blog was Skull Job.

The next post in this blog is Ben Fry Redesign.

Many more can be found on the main index page or by looking through the archives.


Subscribe to feed Subscribe to my feed
Powered by FeedBurner
Creative Commons License
This weblog is licensed under a Creative Commons License.
Powered by
Movable Type 3.33