Karma Initiative: project history

Alexandre Solleiro and Glenn Rolland, at BarCampParis3, designed a global model for karma.

Theses pictures show the resulting work of BarCampParis3:

What is the Karma Initiative

Karma Initiative is:

Karma Initiative is a way of establishing karma. Karma established this way is:

• Global: karma can be impacted by inputs from a single platform, or from multiple platforms (possibly all the Web).
• Persistent: valid throughout time.
• Ponderated: karma data is put into perspective thanks to contextual variables, at all times (it takes subjectivity into account, and encourages it)
• Direct: it is explicitly and transparently created, and is explicit and transparent to the reader (be it human or machine).

As such, the Karma Initiative model can be made into a plugin for existing platforms that ensure user identity (from Wikipedia to any online forum). Also, it can be applied to any proven (or unproven, for that matter) "global" ID-management system, such as OpenID or ClaimID (to name only two).

It's an Initiative for that reason. And because of the bad bad Dharma Initiative (yes, from Lost, yes, the TV series).

Karma Initiative model

To this first founding drawing, only a tagging rule has ben added. It's called 3fold. A textual description of the model contained in the drawing follows, mainly because Alex's writing sucked that day.

The second and third drawings are Glenn's drafts of query structures and grammar (this has considerably evolved since then).

The Karma Initiative model can be applied to anything that is made of an unique identity, an URI, objects and tags: for now, we find it very interesting for humans and projects.

We'll first describe the human model.

The Karma Initiative model for humans

1. Every user has a profile.
2. Every profile is made of one Karma Box , an array of trackbacks and an array of tags (in addition to the typical identity and contact data).
3. The profile is pinged by every object the user creates or modifies inside a platform (the platform can be all the Web). When the object is seen in a profile, we call it a trackback . Every trackback has at least an unique ID and a link to the object itself. Objects can be anything: a blog post, a wiki page, a blog comment, an image, a video, a sound file, a forum post, an email...
4. Profiles, objects and trackbacks can be tagged. Trackbacks automatically inherit tags from objects (but not the other way around). Profiles automatically inherit tags from trackbacks (but not the other way around).
5. There are three levels of tags for every profile (you cannot see this in the drawings), each level containing an unlimited number of tags: one level is considered pseudo-universal and is typically reserved to tags attributed by a platform, a community or a society (such as Expert, Founder, ...) and is called Universal; the second level is considered pseudo-personal and is typically reserved to tags attributed by the user himself (such as skills, interests or subjects like PHP or Web2.0) and is called Personal; the third level is considered pseudo-inherited and is typically reserved to tags attributed by other users (that's where you embrace chaos) and is called Social. We call this 3fold tagging.
6. Remember that the profile regroups all tags attributed to objects, trackbacks, or directly to the user.
7. The Karma Box is actually a table with two columns: one positive and one negative. Every trackback and tag in a profile is a line of this table. Objects and tags can be positively or negatively rated by users. So, to every trackback and tag in the Karma Box we associate the number of positive ratings and the number of negative ratings.
8. By default, every user has already rated every other user positively and negatively. There's no global score an user can give another user: + 1 + - 1 = 0. All that happens in the Karma Box is the enrichement (detailing) of this rating.
9. Indeed, when User A creates an object, User B can give it a thumbs up (thus can positively rate it) or a thumbs down (same here). That object becomes a trackback in User A's profile, and in User A's Karma Box you read 1 in the positive column, and you read 1 in that trackback's line. The system works the same for every tag. User B can only change the trackback's (or object's, or tag's) rating from positive to negative, and back again and again and again. But User B cannot rate it both positively and negatively simultaneously.

What this model achieves

• This model turns feedback into rating
• This model turns subjectivity into contextual objectivity (the bigger the context, the more objective)
• This model provides reliable information and can be a visualization of reputation
• This model can be applied to any platform
• This model allows transparent traceability of ratings
• This model allows social and semantic analysis of ratings and reputation
• This model can be applied to objects themselves

Where this model is tricky

• This model does not (this makes us happy) provide universally-acceptable score of a person
• This model can nonetheless be manipulated as so (mean human race)
• This model requires a whole lot of data to be manipulated and printed on a screen, therefore challenging user interface design and social/community design
• This huge amount of data may also require the model to be carefully studied for every implementation

Why we believe in this model

• Because not only does this model favour the creation of quality contribution, it also allows users to receive feedback on the quality of their work
• Feedback can be outputed and presented as part of a resume
• Because this model's strength lies where most systems fail: it depends on humans
• Because it gives power to the people, wooooooooooooooo!
• Because it's the result of years of unconscious search and it's born at a Barcamp, which makes us happy because we love the brain and Barcamp.
• Because we feel it's so rich that we'd never be able to make a complete list of how you could use it.

==== Problemes ? ====

==== Cas d'utilisation ====

==== Discussions ====