It is a truth universally acknowledged that a Web site with a large user-base must be in need of some social networking features. After-all turning a Web readership into a community, which can share knowledge and opinion, benefits the Web site and the user-base alike. And hey, social networks are hot, so if you can make your site into one of those, you’re bound to boost its value. Right?

Not always: in Digg.com, we have a site where social networking is positively toxic to its core model. Yet the company continues down the social road.  

For those few who haven’t visited it, Digg is a user-driven news aggregator. That is users submit links to interesting news stories, which get voted up (dugg up) or down (buried) by the rest of the users. The ones that are most rapidly “dugg” make it to the site’s front page (that’s a simplification - the algorithm that dictates whether a story to make it to the front of the site is more complex and obfuscated than that).  Digg’s selling point is that the content that makes it to the top of the pile is dictated entirely by the crowd and the algorithm. There are no editors, moderators or overseers deciding what hot and what’s not.

So Digg is a classic embodiment of the “wisdom of crowds”. It turns out that the model works well when your crowd is amorphous and disconnected. But what happens when it starts to get clumpy? What happens when connections occur between parts of the cloud, when things get tribal? Bad things, that’s what.

Over the last couple of years the Digg team has looked to capitalise on all the information generated by the army of diggers, aiming to improve the algorithm so that each user sees what they are interested in.  One strategy adopted was to add social networking into the mix. The thinking is simple enough: if Alice spots that Bob tends to submit interesting stuff, give her an easy way of following Bob’s activities and give the two of them a way to tip each other off about interesting stuff. Consequently Digg introduced a way to build lists of ‘Friends’ - making it easy to follow what particular users are doing and what content they are digging. It also added the ability to send a message, or ‘Shout’ to Friends.

The result was twofold. Some users began digging stories based on who submitted it, rather than the intrinsic value of the content. Networks of friends have emerged that work to promote the submissions of particular users. Of course, such concerted teamwork could happen anyway through e-mail and IM. However with friends and shouts Digg has specifically created the tools that are detrimental to the site’s core mission.

“But hang on a second”, you may be thinking - “that sounds like a good idea, letting me follow someone smart and interesting improves my experience”. You’re right. But the problem with the Digg model is when a group mutually up-vote each other’s stories, the results are non-local to that group. In other words if a cluster of Bobs and Alices have a deep interest in moth-wrangling and digg each other’s stories, it doesn’t only effect the group’s results: Carol and Dave will also see the story rise through the ranks despite the fact that they are much more interested in the latest earwig-racing news.

The addition of a social element also exacerbates the “power user problem”. These are users who build large networks and therefore manage to drive large numbers of stories to the front page. The bête noir of many users are power users who   spot stories submitted by someone else and re-submit it as their own - usually leaving the original submission in the dust. It’s leads to this kind of pitch-fork rattling story from December - and Digg’s response which you can read here.

So Digg clearly takes the issue seriously and has spent significant time tweaking its design to get around the problem. An early response was to removing the list of ‘top posters’ to reduce competition for the title of “number one Digger”. Breaking the site up into topic sections was another move designed to give users an easy way to filter content.

It has also spent a lot time tweaking its algorithms to reduce the unwanted network effects and to make it harder for network-dugg stories to hit the front page.  But as the algorithm has become more sophisticated, so the conceptually simple link between number of diggs and the likelihood of hitting the front page has been lost. Users now puzzle over why some stories with large numbers of diggs never making it to the front page. The answer of course is that the stories may have been dugg, but not by the right kind of users.

Digg has tied itself in knots because of two fundamental problem with its model:

1. The number of diggs that a story receives is global - everyone sees the same number.
2. Everyone sees the same stories hit the front page (assuming they have all the sections turned on).  

Unfortunately network effects mean that while everyone sees a story as having the same number of diggs, the value of each dig varies for each user:  a digg from a Friend is more valuable to me than a digg by your Friend. An unprompted digg from someone is more valuable than someone who has been prompted by a friend. So the contents of my front page should be different to yours.  This is a fact recognised by Digg’s recommendation engine.

So how can Digg get around this? It can’t easily. Giving everyone their own individual front page would lead to confusion, as would displaying an ‘adjusted’ digg number for each user next to each story.    

Digg could simply remove the Friends and Shouts functionality and leave stories in the Upcoming voting queue anonymous until they had passed some threshold of popularity. But it won’t do that because it wants to increase its value and social networking is by far the easiest path when you have a large user-base.

Are there any major lessons here for other sites? Only this: In the vast majority of cases  giving your user-base the tools to communicate is a Good Thing. The main exceptions are: (a) when your customer service and products are so bad that letting users converse will create a lynch-mob (b) when your core model requires a user to act without being biased by other users’ activity.

Before retrofitting sociality to existing site, pause for a moment. How will is the site's mechanics be affected if a sub-set of the users start acting as a coordinated mass, rather than informed individuals?