Designing incentives for collaborative groups

The raw material for Designing Social Interfaces does a lovely job at outlining a range of patterns to incent participation for groups. The patterns reflects a powerful insight – communities vary in their level of competitiveness, and patterns should be used appropriately for the level of competitiveness in the community. “Haphazardly introducing competitive incentives into non-competitive contexts can create problems and may cause a schism within the community.”

Game communities have been leaders in using leaderboards ranking, named and numbered levels, and related techniques to incent participation by triggering the desire to beat the other guy. But for some people these techniques are disincentives: “In user-testing, we’ve seen some strong reactions to Numbered Levels from folks who make associations with ‘being graded’ or assessed. Others noted that numbers just “seem impersonal and kinda cold.”

However, techniques for more collaborative communities have been less well described. This post sketches a few I’ve observed. Reflecting on these patterns, they seem to fall into three groups:
* praise contribution
* make opportunities for participation visible
* make progress toward shared goals visible

Praise contribution

One pattern already described in the manuscript that can be used in a more cooperative setting is collectable acheivement. People can earn badges for completing tasks and post them to their page as marks of personal acheivement. There are also “gift” badges that can be given to others for “those who’ve distinguished themselves in some way: perhaps they’ve excelled at one particular skill that the community values; perhaps they are official representatives for the community or an affiliated organization; perhaps they have volunteered to be a helpful resource for others in the community. ”

On, you can give someone else “props” for sharing a good song. Props give the receiver credit to use with others. This is a non-competitive form of Karma points – Karma points show appreciation and make it easy to compare people’s relative karma.

Improvisational sharing

Shared collections
A common collaborative pattern is the creation of a shared collection. Using a tags to collect material, a group to define contributors, or both, people define something to collect. On Flickr, for example, there are groups that collect pictures of Japanese bento boxes, makes models and serial numbers of airplanes, photos with certain color combinations. On Twitter in recent weeks, people have been sharing links to country/americana songs under the #twangthursday tag.

All that’s needed for this pattern to take hold is the ability to assemble content and people with a tag the group defines, to see the collection, and see new items.

Make opportunities for contribution visible

Gardening tools
In a large collaborative project such as a wiki, gardening tools reveal content that needs to be worked on; pages that are un-linked or un-tagged, content areas that need to be filled in and improved, new contributions that need review. These tools can be used with and without explicit rewards – one principle of “cognitive surplus” communities is that people will contribute and be creative if given a chance – simply making it easy for someone to see what needs to be done and do it is a motivator.

One example is the WikiHow editor’s toolbar which allows WikiHow editors to view edits that need review on new pages, featured pages, and more.

Visible progress toward shared goals

Fundraising widgets
To raise money for a cause, the person or group wanting to raise the money provides widgets that others can embed in their own sites. The widgets allow viewers to donate, and see progress toward the shared fundraising goal.

Information radiators
An Information radiator is a pattern described by agile software development proponent Alistair Cockburn. This is a shared bulletin board with information on the state of a shared project. For example, the “information radiator” for the Socialtext agile development process shows a burndown chart of completed stories for the iteration, whether there are failing automated tests, and the state of stories in progress. The metrics displayed in the radiator reflect items the team wants to track or improve; as the problems to solve change, the metrics may change.

There are probably many more patterns, and ways to organize them too. I’m going to put this to others to help with the brainstorming.

Credit to Peter Kaminski for ideas on Flickr photo sharing as a nomic game.

The death of the hit narrative

Bios and reviews of musicians often talk about careers in terms of hits; when a musician is good but not super-popular, the narrative is about whether this new record might be the one where the artist makes it big. But this story is starting to seem quaint.

Recorded music evolved in an economic context of mass marketing and mass distribution. To make money in this model, where there are high upfront costs and venture risk, you need to sell a ton of stuff. Mass media and mass distribution enable you to sell a lot of stuff. Thus, the evolution of music as a “hit-making” business. This story isn’t just inside baseball; some genius realized that the emergence of hits was a simple, dramatic story for the audience. “Greatest hits” shows told the story of which record was at number 12 on the charts and climbing.

The hit leaderboard takes attention off of the music itself and puts the focus on popularity. Music, within this narrative, lacks context, other than the genre in which it’s being ranked (r&b top 40, country top 40). Just as political coverage using the “horse race” narrative focuses on who’s up, who’s down, who’s stumbled, and ignores substantive coverage of the politicans’ records and campaign content, hit-centric coverage focuses on what’s climbing the charts at the expense of other aspects of the music.

A spot at the top of the charts has become a lot less relevant as tool for becoming aware of popular music, given the tools available for music discovery, surfing your friends’ collections on, trading earworms on blip, googling influences and band members. Out of factors that make music interesting and appealing, raw popularity seems like one, somewhat interesting factor, among others – what’s that sound? Who are they playing with? Where do they come from? What are their influences? What do they want to do next? Where are they playing live?

Of course, people in the know have always had access to their friends’ collections; the recommendations of the crotchety person at the record store; references in liner notes. And there were always subcultures that engaged with the music, not just the hit leaderboard. To take advantage of these things, you needed a strong social context and a fair amount of diligence. Many more people consumed what was provided on the radio and tv. The methods available to the cogniscenti have been democratized.

And this is changing the underlying story that is told about music. The dominant story was one of high peaks and low valleys. The change in music distribution seems to be enabling a tier of musicians who musicians who can make a living, more or less, playing live, selling indy, and exploring their music. Articles that ponder when an artist might hit the big time, and wondering why they haven’t seem irrelevant, as long as the artist is able to eat and stay out of the rain.

The new story, as it is emerging, is more picaresque – where did you come from, where are you now, where are you going?

Distributed social networking and distributed profiles

The nascent vision of distributed social networking imagines transparent ways of sharing profiles and friends. Is there anything in the OpenStack yet that allows profile fields to be distributed?

Here’s the model. If I’m involved in online communities relating to, say, environment, politics, music, and sports, then right now I have a different profile per site. So I want one profile. Right? Wrong. Each context finds different pieces of information relevant. And in each context, the information from other context might be more or less relevant. So what I actually want to do is to link each new site to my core identity, have a uri for each profile field or section, and be able to choose which ones get exposed in which other contexts.

This is one of the places where the corporate world may be a little ahead of the open web. When Socialtext started working on social networking, it was clear to us that we weren’t going to be the core source of basic data like a user’s name, address, and phone number – that information was managed by the corporate directory. Instead, what we would do would be to allow organizations to mirror data from their directory and other sources into Socialtext profiles, and then allow administrators to add new fields, too. The model is that the profile is a shell; you can pull things from elsewhere or create them here. And the profile you create here can be pulled elsewhere using a REST API, also.

In order to have social networking that reflects the irreducably distributed nature of our identities, the elements of the profile need to be pulled apart and reassembled, not just mirrored from place to place.

What Twitter won’t kill – RSS and cultural depth

At the Peninsula Dim Sum Tweetup, Chris Heuer expressed some concern that the shift toward realtime status and realtime search would usher in a dark new age of cultural amnesia in which people’s attention is riveted to the trivially new and we lose the ability to understand in context. This concern is a cultural expression of the technical prediction that in the age of Twitter, Google is dead, RSS is dead, non-realtime content is dead.

I’m less concerned. Sure, the shiny realtime is taking an increasing amount of attention share. And sure, the decentralized digg of realtime meme- and link-sharing is creating a powerful new toolset for discovery. But I strongly suspect that the internet will continue to foster interest in the deep as well as the shiny, and that they are connected. Also, the improved understanding of the value of streams will increase the use of of non-real time streams.

When twitter or some other realtime stream exposes a meme or a link, you can do a few things – pass it on, “like” it up, engage in current conversation, save it for later, and dive into the topic. Realtime helps you participate in the current conversation. Google search helps with the deeper dive.

There reason I’m not worried about amnesia is Clay Shirky’s math about the creative surplus. When hundreds of millions of people are networked, it only takes a very small number of more deeply engaged people to create Linux and Wikipedia. Sure, many more people may see and spread a meme than the number of people who dive deeply into it and organize around it. But math and accessibility favor the deep dive. Not many people will do it, but enough people will, and enough people will connect with each other, that we’ll get the depth and the action.

Let me give just one tiny example. Twitter, and specialized tools like and make it easy to share immediate music updates. And then Google, Youtube, Myspace,, Amazon and many more sites make it super-easy to find more about the music and get further into it. The stream is the tip of of the iceberg.

As for realtime displacing slower streams, the key is a better shared understanding of attention. Yes, realtime feeds replace RSS for immediate information, because it’s more immediate. But there are other, less-real-time uses for information feeds. As Lee Bryant and Shell Israel observe, RSS is a glue technology for exposing streams of information in relevant context, when there isn’t a need for immediate notice.

The first question is why is a stream needed at all. Is a source of information visible enough to the people closest to it, and are they the only people who want the information? No need for a stream. Do you want the information to be more visible, and visible to more people, get a stream. Do you want to share the information in a different context? Embed a feed. Does it not matter if the information is an hour old? Use an RSS feed. These stream management practices will become better understood, and the popularity of live streams will accelerate learning about how to use streams overall.

Belated kudos to CNN and congrats to imajes

After so much the well-deserved criticism of companies that reward their fans and customers with legal hassles, it’s good to see an example of a company that did the right thing. About a month ago, CNN acquired the Twitter account that posted CNN breaking headlines and hired James Cox, who had created the feed as a personal project, as a consultant. The Twitter feed, which James had originally set up to run headlines to his cellphone, was pushing a million followers. It might be within CNN’s legal rights to yank the name and kill the project – instead they realized that James set up a valuable service, so they took it over and are rewarding James for his good work that benefited CNN. Great example of loving your lead customers.

Making FriendFeed useful for Twitterheads

FriendFeed is cool, because it lets you have conversations about status updates, links, photos, and other objects posted on many services, such as delicious, flickr, and more. FriendFeed doesn’t have the “Facebook black hole” problem — content sent into Facebook becomes unfindable, since updates and conversations are visible only to the friends of commenters — because Friendfeed makes content is public by default, and you can “follow” without friending.

But there’s a big drawback for people who are avid Twitter users. Many Twitterites feed their Tweets into FriendFeed. The comment threads following Twitter posts are cool, but the cost is cluttering FriendFeed with a huge stream of redundant Twitter posts.

Here’s how to solve the problem: Create two saved searches in FriendFeed. First, create a search that is everything BUT Twitter. This is what I type into the search box: -service:twitter friends:aslevin Then save the search. Once you do a search, FriendFeed gives you a link to “save this search.” Second, create a saved search for only Twitter updates with comments. This is what I type into the search box: service:twitter comments:1

Voila! Now, you can see the non-redundant Friendfeed stream aggregating posts from other services. And you can see only the Twitter posts that have generated comment threads on FriendFeed.

This is making my FriendFeed experience more useful already. Once Seesmic and Tweetdeck add support for FriendFeed, you can have the searches running in standing desktop columns, too.

Designing for Community – variables and constants

Anthropology: The Art of Building a Successful Social Site is an good ReadWriteWeb articl about how Joel Spolsky & co built a successful social site for technical Q&A. One distinctive feature of the site the answers are wiki-editable, so they can improve over time.

The thought and care required to create a thriving community is a best practice in general. The specific items, however, vary by community. The sorts of rankings that appeal to competitive male programmers might be the wrong affordance for a more gender-mixed or female-predominant audience. Facebook, which is more mixed, has “like” as its only core rating. LiveJournal has lots of ways to show emotion. The need to design the affordances for community interaction is a constant requirement; the design itself will vary by community.

Social network nomads

There’s a migration in progress. Spurred by increasingly restrictive content rules, which barred content in categories ranging from breastfeeding to fan fiction, members of the LiveJournal community are flocking to Dreamwidth, a new service based on a fork of the LiveJournal code running on new servers. The first 200 seed accounts sold out in hours minutes. The Dreamwidth community is in a promising place because the LiveJournal code was open source, and the community had skilled techies who could port the code and run a new server.

Migrations for liberty are not uncommon. For example, the community of self-hosted blogging tool users migrated from MovableType to WordPress when WordPress tools got better for installation and extensibility, and MovableType moved to more restrictive licenses. The availability of self-hosted social network tools like BuddyPress will make it easier for communities to achieve self-rule with a modest cost and moderate level of technical skill.

In a 2004 Corante article, Clay Shirky argued that ownership of server space would be key to community self-determination. This pointed post about LiveJournal TOS issues observes that lack of server ownership may be a particular problem for largely-female communities, since the interests of women are under-represented. Come to think of it, not sure I agree with that, since more men pursue porn and porn is disapproved of under common terms of service.

As noted in follow up to Clay’s original post, owning the server isn’t the only way to achieve online liberty. In the early days of the US republic, only property-owners could vote. But that changed. Liberty, political thinkers concluded in the physical world, is not an attribute of property ownership — it’s an attribute of being human. Someday online residents are likely to organize methods to secure voting rights over their terms of service even when they rent the server and don’t own it.

The migration to a land where the community can be free to practice its own norms, of course, has a long history prior to the internet. Puritans moved from England to be able to practice their unpopular religion and be free to oppress folk they didn’t like. European peasants flocked to the US for economic opportunity in the 1800s and early 1900s. The human species has a nomadic history. Give us something to move from, or something to move to, and our tribe will pull up our tents and pitch them somewhere else. Social networks that pretend to lockin will find that the tribe has gone.

The Network is the Infomediary – 10 years later

Ten years ago, John Hagel and Mark Singer published Net Worth, a visionary book describing an opportunity for companies to take advantage the internet-fueled trend toward customer power by managing customer data on behalf of customers. At the time, I wrote a critique, The Network is the Infomediary predicting that the “infomediaries” would arise from social networks that participants trust to manage their data for them.

Hegel and Singer predicted that the Infomediary business would be a single application handling household purchases — something like a cross between a butler and a corporate procurement application. The SAP of household purchasing would cover every aspect of household commerce, soap and travel, housing and entertainment, requiring a vast customer base, and huge information processing infrastructure. The infomediary would will have “lockin” over customers because they have gathered so much customer data.

At the time I thought that the “monolith” scenario was unlikely because consumers would be unlikely to trust the behemoth. It seemed more likely that infomediaries will grow out of networks and communities. At the time, though, the leaders in the online community business were far from being worthy of community member trust. Providers like AOL and Geocities, Amazon and Yahoo were doing a poor job of protecting participants’ data.

Where are we now

The 1999 essay has held up remarkably well. I still believe networks are the seeds of infomediaries – but the universe described hasn’t quite arrived yet. Social networking has grown and flourished, and is developing more tools and practices for customers to express preferences to vendors. Twitter is showing some dramatic examples of online expression of customer feedback (think #amazonfail). Yelp has product reviews and rather embryonic social networking. Social media monitoring helps aggregate consumer feedback from individual viewpoints across social networking services.

The biggest social networking service, Facebook, still hasn’t proven itself trustworthy as a steward of data for customers. However, the power of their networked customers has helped pull them away from practices that customers perceived as abusive, including using browsing behavior without consent, and terms of service that take ownership of customers’ contributions.

And there still aren’t good ways for people online to weigh in, and to have feedback be cumulative, and to have that feedback be weighted by data about what people actually do. And there really aren’t good ways for people to express affiliation with a community to have their voices count more.

Most of recent action these days is around the realtime. Aardvark and Hunch let people ask the community for advice. But we’re not yet seeing tools that enable customers to organize to request service from vendors. But realtime is just the tip of the iceberg. Realtime can help fix a problem but can’t create alternative solutions. For example – I recently organized a Drive Less Challenge for Menlo Park, CA and surrounding communities. We had a network of people participating in a contest to do less driving alone for the week after earth day. People did what they could, and also observed barriers to driving less. Where aren’t there enough bike racks? Where do people need shuttles to the train? How many people are scared of putting their bike on a bus? It would be great to aggregate that feedback and make it available to service providers.

What did I miss
In the 1999 essay, I probably underestimated the ability of analytics to scale to the level needed to run an infomediary application. But I’m still skeptical of the “strong AI” required for an agent to make purchases on personal buyer’s behalf. Browsing and asking for human recommendations is a learning process – I often change my mind based on browsing – and I don’t want to cede browsing to a bot.


Paul Saffo says never to mistake a clear view for a short distance. The vision that the internet will reverse the power balance between marketers and customers, popularized by Esther Dyson in the 90s, is now getting a step closer to realization. Social networks and social media are a key part of realizing this trend, but the tools and practices to make it happen are still in an early stage.