|This is a failed proposal.|
|This page in a nutshell: A delegable proxy system could help answer the question, "If more people got involved in this discussion, what would be the likely outcome?"|
Delegable proxy is a proposal to allow each user to designate a trusted user to speak on his behalf in debates in which the user does not participate personally. It would enable an analysis that reflects what the outcome might have been had more users participated. Its implementation would be rather simple, requiring only a proxy table. Delegable proxy does not contemplate forcing matters to a vote. Rather, it would provide an additional source of information for analyzing the results of our existing decision-making processes. In those processes, objective facts and policy take precedence over opinions; alternatives can be suggested and debated simultaneously as the original proposal; and discussion is used to present relevant facts and persuade others in an effort to reach a consensus. Certain debates more closely resemble a poll than others because the nature of the subject matter precludes creative alternatives, limiting the available options to a small set of choices (e.g. support or oppose; keep or delete; etc.) discussed in one central location. In such cases, delegable proxy's ability to analyze representativeness of the results may be particularly useful to closing admins.
Delegable proxy would initially be an experimental, optional, advisory tool, with no changes to current guidelines or policies needed to implement it. The community could examine the results of the experiment to help decide whether it would be a good idea to turn delegable proxy into a binding aspect of Wikipedia's control mechanisms, and some proponents of delegable proxy oppose making it binding even if it is successful, considering that advice is actually better than control.
Users designate a trusted user they generally agree with by adding him to the Wikipedia:Delegable proxy/Table and obtaining the proxy's acceptance. If a debate arises in which the user does not participate, but his proxy does, then a closing admin may look at the table (presumably with the help of an automated proxy expansion tool, analogously to the RfA analysis tool used by bureaucrats) and note that, had the user participated, he probably would have shared the same opinion as his proxy. If the user himself participates, then the proxy is of no effect in that particular debate. A user can also revoke the proxy at any time.
The proxy is delegable, which allows proxy chains to form. For instance, if A appoints B, and B appoints C, but C is the only one of those three who participates in a debate, then C's opinion counts for all three of them. This allows for a backup; in the event that, for instance, both A and B go on vacation, their general views are still represented while they are gone. It also allows for B to delegate some of the workload of participation to C.
It is possible for users to form proxy loops in which, for instance, A appoints B, B appoints C, and C appoints A. In this case, if neither B nor C participate, but A does, then the proxy expansion will show A as being the proxy for the other two. Proxy loops are inevitable if everyone has a proxy, and desirable in that they increase the robustness of the system by allowing users to prevent a scenario where someone doesn't get represented because they are at the head of a proxy chain. Users in a proxy loop can take turns participating on behalf of the others, if they so desire.
The information provided by the proxy system becomes more reliable as more users participate and as the set of participating users becomes more representative of the overall user population. A participation bias in the proxy system does not necessarily negate its usefulness, as long as closing admins are able to glean some insight from proxy expansions after taking the bias into account.
The main rationale for the proxy system is to counter the effects of participation bias in debates on Wikipedia. In any given debate, only a very small percentage of Wikipedians participate. Especially when groups smaller than about 30 are involved, the small sample size creates the possibility for random factors to cause the outcome to deviate from what the larger community would have decided. The participants are self-selected, which creates further potential for biased results. An example might be if most of the people who take an interest in Wikipedia:Miscellany for deletion discussions are of the deletionist persuasion, or if a post is made to a WikiProject page about a pending deletion debate on an article under their purview, which draws a sufficiently large group to sway the result to Keep. Under a delegable proxy system in which all the pro-Keep WikiProject members were joined together into a chain or loop, the proxy expansion would show the same result whether one member participated or all of them. This would eliminate the advantage generated by one group of editors being more organized, or devoting more time to deletion debates, than another.
By providing a mechanism for gaining insight into the views of infrequent debate participants, delegable proxy may facilitate and reward more productive allocation of editor time. The current system encourages users to devote time to, for instance, deletion or policy debates if they want to influence the overall patterns of deletion. However, this is only a good thing if a particular user's time is most productively employed in such debates. Some users may have relatively rare and useful skills, such as bot code creation or scientific article writing, that they can contribute to the encyclopedia. If they also feel passionately about certain issues (e.g. deletion policy), then given a limited amount of time in which to participate in the encyclopedia, they have to choose between forgoing either some of the work they are best at, or some influence they could have exerted over discussion outcomes. It seems ironic that users who choose to help the encyclopedia primarily through contributions to the mainspace or technical aspects of Wikipedia should be penalized with diminished clout. Delegable proxy can provide an efficient avenue for users to make their opinions and preferences known.
Delegable proxy may also improve the quality of debates. The current system gives users an incentive to quickly review and voice opinions in many different deletion debates if they want to maximize the breadth of their influence over deletion policy application. There is no penalty for an ill-considered comment, other than being disregarded by the closing admin. Someone designated as proxy by many users might well be more careful, as poorly-considered comments not only could be disregarded, but might spur users to drop him as proxy.
Admins could use proxy expansions in many different ways. Unusual patterns of any type can provide an impetus for further investigation. For instance, if the users on one side of a debate are speaking on behalf many users, and the votes on the other side are speaking on behalf of few, it could reflect votestacking by the latter. In general, the fact that many users have entrusted someone to speak on behalf of them suggests that the user is known and respected by at least that group of users, which may be useful information to the admin. Some things can also probably be inferred by looking into which users have designated a person as proxy. Irrelevant data can be filtered out; an admin may choose to disregard, for instance, proxies of users who have made no edits for several months.
A common objection to delegable proxy on Wikipedia is the potential for users with sockpuppets or meatpuppets to game the system. Users could start multiple accounts, each designating the main account as proxy. This would be a risky proposition, however, as discovery could not only result in blocks and bans, but in harm to the user's reputation that would make legitimate users less likely to designate him as proxy. Automated analysis of statistics such as account creation date, edit count, date of proxy execution, date of last edit, voting patterns, and possibly IP addresses, can detect suspicious patterns. The proxy system may help fight sockpuppets by providing another source of data that can be used to unmask them. Moreover, as use of the proxy system becomes more common, it would become more difficult to influence a debate through direct participation by a small number of sockpuppets; it would be necessary to use more of them, or link them together into proxy chains. Unsophisticated attempts to do either would leave suspicious clues. And once one sockpuppet in a proxy chain is detected through checkuser analysis or other evidence, the other members of the chain can be scrutinized as well. A lack of on-wiki communication among members of a proxy chain would be a suspicious sign, as presumably users would choose someone they know. In short, it would be hard for a sockpuppetmaster to convincingly emulate the formation and use of a legitimate proxy chain; and such activity, once discovered, would unmask the whole chain and lead to its dismantling.
Some users object that delegable proxy will encourage cliques. It is worth asking, what, exactly, is objectionable about cliques? One definition of clique is "an exclusive circle of people with a common purpose." Presumably the exclusivity is the problematic part. It is true that members of a proxy chain or loop can decline to choose a person as their proxy, and in that sense, that person is left outside of that loop. In some cases, that might be desirable, as when no one is willing to name a troll or crank as their proxy; that could actually improve the quality of decisions. The proxy table would be public, so anyone could see the membership and structure of each proxy chain. Moreover, anything that members of a proxy chain do on Wikipedia will still be a matter of public record, and other users can review their contributions and participate alongside them. Lastly, the proxy chains will not affect outcomes by coordinating mass participation by their members (e.g. through off-wiki communication such as IRC), since participation by any one members counts as participation by them all in the proxy expansion. This makes it different from some cliques that have appeared on Wikipedia in the past.
Another pitfall is that users might leave the project, yet their proxies would remain in effect, distorting the proxy expansions' representativeness of currently participating Wikipedians. There are many possible approaches to this issue, such as expiration dates on proxies. The simplest solution is for the person running the analysis to factor in the length of time since a user actively contributed, and how substantial their contributions to date have been. Someone whose only edit was to appoint a proxy a few days ago would likely have much less influence than a Wikipedian who accumulated thousands of edits over the past three years and left three weeks ago. In the latter case, the length of time they were gone would be only 3 weeks out of 156, introducing a 2% distortion for that individual. It is up to the person using the analysis whether that margin of error is acceptable.
Another objection is that Wikipedia discussions are made by consensus reached from a discussion on the merits of the issue at hand, not votes, and therefore it violates the guideline that polling is not a substitute for discussion. While delegable proxy can be used to obtain a more accurate numerical representation of what the outcome would have been had a larger group participated, it is not intended that decisions be based solely on proxy expansions. If the facts and policy are clearly on the side of the minority, then the outcome should reflect that. Not all cases are clear-cut, however. Subjects such as notability, especially in borderline cases, often involve judgment calls. The quality of a judgment call presumably is influenced by the judgment of the person making it, and in those cases, the proxy expansion's ability to show which users have been entrusted by others to weigh in on their behalf may be useful.
Moreover, some decisions do take on characteristics of polling. There is substantial disagreement between inclusionist and deletionist factions about what types of articles the encyclopedia should keep. Under the consensus-based process by which policy and guidelines are made, widely agreed-upon standards becoming binding while contentious ones are excluded from policy pages, leaving a large grey area in which the rules' application must be decided on a case-by-case basis. The policymaking process itself is guided by few particular superordinate goals or standards other than a general mission to improve the encyclopedia, which means different things to different people. When debates take place without commonly-agreed upon standards by which to objectively prove the merits of one argument or another, the results are prone to come down to subjective opinions, preferences and wishes of the users who participate. In those cases, an admin has little choice but to close based on whatever is the rough consensus of participants is, or declare there is no consensus and maintain the status quo. So, we do vote sometimes, and the results are biased by who contributes. This is evident in the need for a rules against forum shopping and canvassing; if such attempts to manipulate the composition of the group of decisionmakers had no potential to influence the result, people would not try it.
One objection to delegable proxy (and polling in general) is that it is contrary to the normal practice of coming up with creative ideas in the course of discussion and reaching consensus through persuasion and modifications to the original proposal. In some discussions, such as policy debates, creative alternatives are proposed that branch off into separate discussions which take place simultaneously. This process allows opportunity for better proposals to emerge that the community can agree on. Other debates, such as AfD and RfA, do not lend themselves as well to creative solutions because the process imposes limits on the possible outcomes. An RfA, for instance, will ultimately result in a decision to promote or not promote someone to sysop. An AfD has a limited set of outcomes, such as keep, delete, or merge. And certain policy discussions eventually reach an impasse in which continued debate is not generating any new alternatives, and the discussion focuses on a decision to enact or not enact a specific proposed change. In those cases, discussion may still be useful in that as objections are raised and addressed, users can become better-informed about aspects they hadn't considered, and possibly switch sides, may result in better decisions and a less divisive final outcome. Sometimes a poll is in order, though, when it is evident that more debate will not significantly change opinion or result in new solutions, or when the decision needs to be made in a timely manner (e.g. we would not want RfAs to drag on for very long). When we do poll, delegable proxy can be a useful tool for analyzing the representativeness of the results.