On likes and liking

Encouraging users to ‘thumbs up’ or ‘thumbs down’ items is a great way to get some sentiment-based feedback on what can be an unmanageably large amount of data. But how reliable is it?

Both FriendFeed and Socialmedian have a binary way of saying you found a particular news item or post interesting – a quiet nod of approval, if you will. I like this. I don’t like this. As commenters have pointed out, the word ‘like’ isn’t always appropriate (I “like” the story about a celebrity suicide?) but that’s purely semantics.

What’s the point, though? By ‘liking’ items on FriendFeed you can help populate ‘best of’ lists, and aid uses in seeing at a glance what’s worth looking at. On the other hand, why do I care if Joe Bloggs, friend of Robert Scoble, likes an item? He might find entirely different things interesting to me. When I only know one of the people who likes a story, is there real value in pulling out ‘most liked’ items?

FriendFeed doesn’t have the option to dislike, but Socialmedian’s mood indicator includes this. In a way, this can be more valuable, as instead of pulling signal from noise you’re also dampening the noise itself. On the other hand, I couldn’t really see a way to use the mood indicator, other than as a visual key to a story’s potential. It’s a new feature, so rankings and algorithms that take advantage of mood might show up a bit later on.

It just seems to me that there’s a lot missing with this metric. Firstly, it’s solicited approval. I was recently talking to a local entrepreneur hoping to use a similar concept in a product he’s developing, and my main concern is simply “why?”. If you can take advantage of the information you gain, knowing that it’s not necessarily representative, in line with every user’s tastes, and can be extremely sparse.. then it’s worth doing, but I would never use solicited thumbs up as a single point of recommendation. The other “why” is “why would users bother?” – you need some sort of incentive scheme, especially outside the fluffy world of online social media. People using social sites like the idea of wisdom of crowds; they favourite and add things because they get the whole “hive mind” concept. But I’d wager your average person – who likes and dislikes a lot of things – doesn’t see any reason to go out of their way to tell you about them.

So your problems are: barrier to entry (has to be easy to provide a thumb rating), incentive to rate (points mean prizes!), plus the social incentive (if you rate this then you, as a service user, get better information thanks to karma).

The thing is that there’s a lot more information about the objects being rated than simply soliciting one set of users’ feedback will cover. If I know that ten people liked an item on Friendfeed, what about Friendfeed users as a whole? How does the reaction to this link today differ from yesterday’s, or from the other popular links of the day? Are the comments positive or negative in tone? And what about people who crosspost the same link, or who talk about it on Socialmedian, or Facebook, or Twitter… what about the people who have conversations on blogs outside the radar?

Taking that offline, I was recently looking for two things: a good local garage, and the best high street optician. Garages aren’t exactly social network fodder, but I found various listings, some forum threads and several of them had websites. It was quite a lot of manual slog to figure out a place that did what I needed, was within a sensible distance, and had some ‘reliability’ indicators (either user feedback, or propaganda on their website!). The Internet knew all this already, but I had to spend a few hours going around in Google circles mostly comprising identical directory sites (hurrah for adding value).

As for high street opticians, there are several reviews of different chains out there, and specific mentions of local branches too. They’re spread around though; finding Sally on Livejournal’s comments about the Accrington branch having great service isn’t that helpful to me, but knowing that most customers ‘like’ a certain chain better than another, weighted against special offers and location, takes a lot of the work out of decision-making. Yet installing a phone or online app that encouraged people to “thumbs up” individual chains simply wouldn’t fly – why would people bother? Plus, generally people have one set of eyes and use one optician, so the valuable feedback (people who have experience from multiple chains and reflect that in their voting) would be drowned amongst all the customers just voting for the chain they use.

Let’s take this back online. Generally, it’s a lot easier to do the sort of data processing and top-down views mentioned above for online sources, simply because there’s more information available, and the power of a social-savvy audience means participation is easy to encourage. The scale of processing we’re talking about here is potentially huge, though; confining user opinions to individual sites means you can put limits on your data, and how happy would FriendFeed and Socialmedian be to find a third-party ‘borrowing’ their user data? Yet that seems to be the best way to aggregate across all these sites, possibly giving the output back in the form of a popularity API – you tell us what your users think, and we’ll tell you how that fits in with the bigger picture.

User profiling and taste matching is a different kettle of fish, and a lot harder to do if  you don’t know anything about the user at all (although it’s not impossible; hello New User Joe, would you say you were more like Robert Scoble or Britney Spears?). If you’ve got access to the social services a user subscribes to; what links they bookmark online, what items in a RSS reader they click through to, where they leave comments, what they say on their personal blog… We leave a frighteningly large preference footprint online, and with access to that, a clever algorithm could conceivably come up with a set of multidimensional user models that really helped users see “who else like me liked this item?” – or, more usefully, “what did people like me find interesting today?”.

Oh, but that’s not everything. The Internet isn’t a series of users operating alone in the dark, after all. There’s a web overlaying everyone, connecting us with people we read, comment on, bookmark, reply to, tweet about; there’s stuff flowing through this network, bits, bytes and data, affecting our own behaviour. To what extent do others influence us, and how can we tap into this influence network to figure out who the real movers and shakers are? Imagine a darkened screen; suddenly, far left, a light switches on. (Scoble made a post). Let’s watch that light propagate throughout the network… suddenly you get something that looks like this.

That’s what I see when I think of social media and influence networks; of course, it merely begs the question “how do we model it?”,  how to find out where the movers and shakers are, where the hubs and authorities lie, and how the interplay of linking, tweeting, commenting and liking reflects what we think about the people behind the actions.

There’s also a darker side; companies wishing to gain good standing online might want to find out who they need to talk to first. If TechCrunch writes about someone, it’s probably a better indicator of their awesomeness than if Mario on Greatestjournal does, and far more people are going to read it and come to some opinion about the subject. How many times have we felt more comfortable with a product or company that we’ve “heard of”, even if we can’t really remember why or where we’d heard of them? Familiarity breeds customers. But there is such a thing as bad press, and sensationalists love to jump on bandwagons, so what about companies that have a reputation to fix, can’t get covered by the biggest fish, but want to find a few smaller points of goodwill injection throughout the blogosphere that will start spreading warm fuzzies about their products in different sub-networks?

The amount of possiblity in this space astounds me, and I’m really looking forward to getting my teeth into some of the things I mentioned oh-so-briefly above. The work I’m doing right now will evolve into a form of trendwatching using some of the principles above, but the sheer scale of the “wouldn’t it be cool if..” end results means they’re some way off for now. Still, it’s going to be a fascinating ride.