Comments welcome! Brynosaurus 12:41, 16 Aug 2004 (UTC)
Separate branches for small groups to collaborate on isn't a bad idea to allow groups some freedom from harassment while working on common text. But if branches remain separated--and even worse, mutually exclusive--then you will miss makes Wikipedia so good. "Adversarial editing" where users who disagree with each other push each other, check each other's references, and challenge each other's assumptions is an essential part of the wiki (and academic) process. Don't make it too easy for users who disagree to take their ball and go to their own revision. 184.108.40.206 17:08, 17 Aug 2004 (UTC)
I like it. For the votes, I was thinking that votes propagate backwards, not forward. Each user will carry a vote with them. When they move between branches, their vote will follow, and percolate down to the trunk. (This doesn't have to be done explicitly all the way down, because the branches generally converge pretty quickly.)
certainly an advangtage is that people who disrespect consensus will quickly recognize the futility of their way of thinking, without people having to explain the obvious to them over and over again. People will realize more directly what will fly and what won't, by the sheer math, and that will encourage more fair and practical editing, without providing them with the opportunity to project their chagrin on others, and engage in futile but annoying fights. Which leads to another advantage: the reduction of person-centered conflict will diminish aggression and distrust that tend to reinforcement idealogical schisms.
Regarding "adversarial editing", perhaps the problem can be thought of in terms of incentives and social gravitating centers, such as information sources: I think users would be more inclined to compare sources and cooperate if there were some features for sources built into the wiki.
Some analogies I thought of for the branching idea:
- optician, switching thru lenses with the big lens-machine thing to give you the perfect eyeglasses.
- Darwinian evolution (the tree of life) (this one is obvious, and i imagine that it is a common analogy)
- morphospace (the hypothetical(probabalistic) space of potential paths of structural development for a developing embryo)
- stochastic markov chains
- emergent collective behavior of ants (by themselves, they walk randomly-the branches. votes are pherome.)
- diffusion limited aggregation (growth of fractal structures, such as snowflakes, the diffusion is limited by the cost of creating new branches and the path-reinforcing positive feedback of votes)
- weiner process, sub-gaussian. (norbert weiner is one of the founders of the field of cybernetics) (sub-gaussian because of the trajectory auto-reinforcement effect of votes)
etc. I think such a medium; such a flow-constraint dynamic, is well-suited to the constituitive forces that will utilize it (the users). That is, if the medium is a channel, and the users' input to the system are liquid being poured into the channel, I think the combination of the two will result in a high-quality encyclopedia being poured out. Or, to put it another way, the relaxations of potential energies - the non-equilibrium thermodynamic processes of heat(energy/disorder) dissipation - will favor, through this material, the formation of convection cells with the form and content of an encyclopedia.
Okay, I got a little carried away. I hope that was comprehensible. To put it in plain english, I'm excited about your ideas. Kevin baas 18:35, 17 Aug 2004 (UTC)
not at all comfortable with this system, seems to defeat the consensus basis for wikipedia. consensus isn't reached by democratic process, we are not really trying to make an 'encyclopedia of what the majority wants' we are trying to make something much more potent, an 'encyclopedia of undisputed facts' . by voting or by the suggested branch system, we diverge from consensus at each branch or each vote. I appreciate the enormous amount of work involved in creating the idea, but I don't feel it is best for us all. My suggestion for edit wars is: if someone reverts your page, point it out to 2 other editors, and drop out of the argument.. if you can't stand to drop out of the argument, it's a sure sign of bias... I'll discuss this at length another time if you like, just drop me a line. And thanks for all the effort you've put into this, you obviously have the unselfish diligence that I see as essential to this project. drop me a line if you like.Pedant 11:08, 2004 Aug 18 (UTC)
- I received a similar reaction against the basic idea of "voting" from Jimbo Wales on the wikitech-l list, and I suspect it will be a common objection. But my perspective is that voting is not necessarily at odds with consensus-building; in fact it can be a vital tool without which in many cases consensus will never happen. Witness all the informal uses of voting people already engage in all over wikipedia, for the purpose of resolving contentious issues. In order for sensible decisions to be made, sensible alternatives must be considered - but if none of the participants' alternative visions can actually be realized and presented without continual interference from the others (or from downright incivil trolls or vandals), then all but the most dedicated (or most ideologically motivated) participants will likely drop out out of frustration, and the resulting content is just an impoverished mess of conflicting visions mashed together.
- Voting does not necessarily eliminate the motivation for participants to build consensus and incorporate minority viewpoints. Most sensible participants in a venture like Wikipedia usually want to get along and address the legitimate concerns of others, including the concerns of tiny minorities of one; they just don't necessarily want to be subject to constant harrassment by trolls and ideologues who will never contribute to consensus in any case. Furthermore, each of the members of a 90% majority viewpoint participating in one discussion on one page has a very good motivation to listen to and consider the 10% minority viewpoint, even if they don't have to follow it - because each one of those members of that 90% majority realizes that on a different issue being discussed on a different page at a different time, they might find themselves a 10% minority facing down a 90% majority group that happens to include the person that happened to hold the 10% minority viewpoint on the first page. ("Tyranny of the majority" is certainly a problem in certain broken voting systems, such as the two-party system of the U.S. - but that's because the two-party system is rigged to make sure the same majority gets to decide on every issue, until the public eventually gets fed up and flip-flops to the other party. But let's not go there. :) )
- FYI, my interest in this idea isn't born out of personal frustration about revert or edit wars, as I gather you're assuming from your response - I haven't even been in an edit war yet. My interest is rather born out of a desire to see the wiki idea eventually successfully scale to public collaboration systems perhaps hundreds or thousands of times the size and popularity of the current wikipedia, in which there might be hundreds of active watchers and/or editors butting heads on popular or controversial pages rather than just two or three or five, and in which the content being edited collaboratively might not always just be "cold hard facts" (as on Wikipedia) but also really controversial stuff such as public policy recommendations produced via mass popular participation. In such an environment I suspect current Wiki systems would survive about as long as an ice cube in hell. Even Wikipedia, with its emphasis on dealing only with "undisputed" facts, could in a few years easily become yet another of those nice idealistic technological experiments like Usenet that was great until the spammers took over and chased out all the intelligent contributors.
- Brynosaurus 03:50, 19 Aug 2004 (UTC)
Some ideas 
Throwing some ideas at you:
Perhaps the "leading page"; that is, the dominant/preeminent page (the page that would show in a "refined" print version), instead of being the most recent edit, would be the page w/the most votes. Thus, for a new change to become substantiated and become the "acting" page, it would have to have consensus. Until then, the precedent would be mantained.
I could envision a problem with this, thou: someone makes a decent change, but it never becomes substantiated, not because people don't like it, but simply because they don't notice. That's not too much of a problem, because only... Now wait - if votes are propogated back, the version with the most votes will be the trunk, because as soon as there is a branch, all the descendant (more recent) versions will have less votes. There must be a distinction then, between lead votes and back-propogated votes, if votes are back-propogated. (the reason they'd be back-propogated is so that people could find the main branches) Lead votes would determine the "acting" (dominant) page. If this solution is insufficient, an automatic vote-weight bias could be given to descendant (more recent) branches.
This would be an alternative way to pacify vandalism, and it would also be a way to reinforce the legitimacy of consensus. There still remains the problem that it might slow down the progress of the page, perhaps over-stabilizing it, esp. if people are overly dormant / inattentive: their failure to vote might be obstructive.
Perhaps people who have made changes that were agreed upon (initiated branches that were choosen) could tentatively forward-propagate some votes with them? (adaptive viscosity, w/logarithmic growth and decay) This, however, may be over-canalyzing, restricting the exploratory advantage of branching.
Perhaps these features could be turned on or off, and perhaps parameters for them could be likewise modified.
Kevin baas 21:03, 22 Aug 2004 (UTC)
Negative values could be put into the parameters, to reverse the dynamics.. Kevin baas 21:47, 22 Aug 2004 (UTC)
Philosophical rambling 
Here's a useful perspective:
In a postmodern interpretation, we are not concerned with "the Truth" as an absolute original, but rather with the faithfull reproduction and simulation of phenomena. In information theory terms, we want reliable communication over a noisy channel. Taking an information theory approach, what we have here is an ensemble of information channels with noise. The goal then is to develop impartial "error-correction algorithms", like those used in a computer. This is possible because the ensembles of channels are partially redundant - some people have access to the same information, and are going to say the same thing. Then, we use what a standard error correcting algorithm does - bit-voting; if 5 channels say it is "1" and 3 say it is "0", it is most likely "1". (now branch-voting) Yet this does not completely solve the problem; this is not a sufficient application of the "noisy channel coding theorem". That choice of "1" may still have been wrong, which might later be discovered by information in the future. Thus, there needs to be some temporally-manifest "bit-voting" mechanism as well as the atemporal ensemble bit-voting. An example of this might be the above-mentioned "adaptive viscosity" tentative vote. Currently existing examples of this are user rfc, mediation, and esp. arbitration, though I think that most examples are implicit, through policies, ettiquete, citing sources, people's memory, etc.
Kevin baas 21:47, 22 Aug 2004 (UTC)
degradation of proxy votes. Oh, and in light of the information theory perspective, proxy votes should have fractional value, because they represent degraded messages. (they have a lower probability of being accurate than a direct vote, and the weight of the vote should be proportional to the probability of it being an accurate prediction) - perhaps, like the adaptive viscosity, a sample can be used to determine the approx. probability, and this can be a multiplier on the vote weight. proxies of proxies would then cumulatively degrade. (there could, in addition, be a static coefficient proxy vote degradation multiplier that an adminstrator could set universally (or could be voted on - all of the parameters could be voted on, creating a self-tuning system.) )
degradation of predictive votes. Likewise, the same logic should be used to determine adaptive viscosity - the contributor carries a portion of the vote proportional to the (a posteri) probability of that branch being an accurate prediction for the vote. If two branches are made, the portion of predictive votes is split among the branches according to the rules of probability.
interaction of proxy votes w/predictive votes. The proxy votes that the branch creator carries with him are not part of the predictive vote, they are part of the branch creator's vote, and therefore are not degraded. The proxy votes that other contributes carry with them are part of the portion whose vote is yet to be determined, and thus are part of the prediction and are respectively degraded. In sum, proxy votes stay w/the user's votes, and are degraded or not degraded exactly as is the user's vote.
social effects. All this would encourage people to work towards consensus and make edits that people will agree with, as they will be rewarded in due proportion w/predictive statistical clout.
Kevin baas 01:35, 23 Aug 2004 (UTC)
decaying proxy votes. The weight of a proxy vote should decay overtime as the votes are not overridden or confirmed by the user, so as to check the accumulation of arbitrary power which may result from users' inattentiveness. This decay represents the "information decay" that the user has about their proxy's voting pattern, which naturally diverges over time.
Kevin baas 23:54, 25 Aug 2004 (UTC)
- Kevin, thanks for the great ideas and discussion! I particularly like your idea of degrading proxy and/or predictive votes via tunable parameters. The idea of degrading proxy votes had occurred to me too at one point a while ago, but to keep things simple I didn't explore it much or include it in the initial write-up. Also, your suggestion of having the weight of predictive votes be gradually attenuated as they propagate forward through a branch certainly sounds like a better solution than just having votes suddenly expire all at once. It's not obvious to me whether this attenuation should be based primarily on time (i.e., how "old" the votes are) or on "branch distance" (i.e., how many new revisions the votes are propagated through) or both, but that's a detail that can be worked out later, perhaps as another tweakable parameter. - Brynosaurus 09:33, 26 Aug 2004 (UTC)
Section-based branches 
Here's a radical idea! Notice on the George W. Bush page, the last couple of reverts were "selective reverts". Different parts of the page become disputed, and some would vote one way on a para and a different way on a different para. - but this would mean that, if there were two paragraphs with two options a piece, there would have to be four branches. If someone made a change to one branch, they'd have to propagate that change through all the competing branches in order to be sure to not lose the change. There should be a way to synthesize branches - to pick advantages from each and combine them together. What if the paragraphs branched independantly; if the page was broken up into "information chunks", each independantly operable - it might be more feasible to do this on the section level. Each section would act like a page, in that it would have it's own independant branching tree. Insertion/deletion/moving of sections might require a special interface but in any case would be likewise voted on. Insertion/deletion/moving would not interfere with the per-section wiki-branches.
This would also encourage cooperation because the page would not branch as a whole, creating independat pages, but elements from different "edit camps", would they form, could be combined - would be combined - without any special effort.
Also, if the restructuring (insert/delete/move) was not branched, but serialized, that would force the contributors to collaborate, as they'd all have to operate under a shared page structure - & it would ensure that the branches were compatible.
Kevin baas 21:52, 26 Aug 2004 (UTC)
- I think you got a point in that there will be related pieces of "branched" information, within the same page, but also between pages. Thus, related sets of branches. This is addressed in SCM systems by having branch types, which unfortunately leads to having a way to select a priori certain branches, if found of an available type.
- The only simplifying factor is that one shouldn't care about section branches, but only for page branches, maybe with a possibility to combine pages, or to expand a link in-place (which is in fact the same thing).
- Well, I am impressed that this idea is taken seriously. I'd suggest to experiment... Marc Girod 13:54, 1 Sep 2004 (UTC)
- Yes, I second Marc's suggestion about only having branches for whole pages, but adding a facility for in-line page expansion so that multiple separately-editable, separately-manageable pages can be made to appear as one big page with sections and a table of contents. It seems like a good idea to me even without branches or voting. Brynosaurus 13:38, 3 Sep 2004 (UTC)
Ranked Voting 
All these ideas are absolutely brilliant! There seems to be alot of people thinking independantly about moving the wiki idea into some sort of direct democracy platform. My thoughts: All the discussion so far has assumed that a user will cast a single vote among a variety of different options. This has all the classical draw backs of first past the post voting systems - Voters are pressured to vote for one of the two options they predict are most likely to win, even if they ideally do not want to vote for either of them, because a vote for any other option will be likely to be wasted and have no impact on the final result. A much better system would be some sort of ranked voting system, I myself prefer the condorcet method. It is a very elegant and powerful voting system. Thoughts? Kodemizer Sept 24, 2005
Oh yeah, one more thing, does any know if there is a specific project for developing the MediaWiki software into a system to be used for electronic democracy?? Kodemizer Sept 24, 2005
Here is the location of a specific proposal on the English wikipedia. Sarsaparilla 08:20, 16 February 2008 (UTC)
- Well, it is not entirely specific yet. However, the basic outlines are there. A format for a proxy table is created, to make it easy to use and analyze. A master proxy table may be created, and possibly special-use proxy tables, by editors who see a need for such. How such tables are actually used is up to the analyst who desires a more accurate measure of overall consensus. Such analysis can include many other factors besides what is in the proxy table, and deweighting from proxy depth is one such; there can be many others, such as edit counts of various kinds, days since registration, currency of edits, user status (admin?), or others. The point is that it is the discretion of the analyst what any of this means. The simple applications require no special tools, more complex ones would require tools.
- If delegable proxy is used for making binding decisions, many issues must be faced and resolved. But for advisory use, none of them need be addressed, really. Instead of designing it, top-down, let it grow, bottom-up, and see how it works.
- --Abd 18:39, 16 February 2008 (UTC)
Opening up Wiki democracy 
Hi Bryan, your page on democracy is very interesting, and it's a shame that it hasn't developed since 2004. Would you have any problem with moving or copying the current content to a new page out of the User namespace, to clarify the copyright status and invite contributions? --Chealer (talk) 20:21, 15 May 2012 (UTC)