dsstodol@daimi.aau.dk (David S. Stodolsky) (10/25/90)
markabel@uswat.uswest.com (Mark Abel) in Message-ID: <5938@uswat.UUCP> writes: >There has been some discussion in this group recently regarding a "self >moderating posting scheme." The idea was to have a system that tallied >reader response in some weighted fashion to provide a writer with feedback. There is a big difference between just giving a writer feedback and selecting a writer for publication. ======================================================== My "Report card" resulting from the original posting: Report 1: 1:BRAVO ------------------------------ Report 2: A:Agree B:Relevant C:Understand D:BFD (Big F* Deal = Trivial) E:Let's do it F:Amuzing X:.... For you: 2ABCE ------------------------------- Thanks for the feedback folks. Actually doing this would mean that the reports go to a mediator before being transmitted to the author. There are two reasons for this. First, we want reports to be protected. With journals, referees are normally anonymous. Using pseudonyms is the right way to do this. Second, a mediator would combine the reports in some way. Thus the author would get a summary report, and each referee would also get a report on how their judgements fit the overall pattern of judgments. Some of these referees would be invited to submit a follow-up to the article they had evaluated. Doing this on Usenet means setting up a moderated group. Anything posted to this group would be automatically forwarded to the mediator by mail. For example, you read an article in comp.groupware.d (for digest) and give your feedback as a follow-up. This could be done manually, or somebody could develop a macro which asked a sequence of questions and which could be easily invoked after reading an article (I think this could be done easily with the nn newsreader package). After the time for responses had elapsed (say a week), the mediator performs a set of calculations and issues reports by mail to authors and referees. Referees selected to be new authors are expected to mail their articles to the mediator within a week. These are issued to start a new cycle. Now, how to insure that the mediator isn't cheating? Each report received would be posted to comp.groupware, under a pseudonym. Thus, anybody could check the calculations or even try their own. Better yet, this would allow readers to independently decide who was worth reading, regardless of dominate opinion. This is what I have termed "mutual moderation". Actually, it is simpler since it does not require central mediation. But it does require each person to maintain a reputation database on their own machine. Actually, we could start doing this without creating a new group. The minimum requirement is that we agree on a "report form". Then the reports could be picked out of the stream of articles in comp.groupware by some patch or macro in the newsreading software and used to rank the "human readable" articles waiting (see the reposted material on "mutual moderation"). I am just guessing about the best way to do this on Usenet. Earlier I had proposed that the group comp.groupware.f be created to handle the feedback traffic, but this was rejected by the people on the "newgroup" committee. One argument put forward was that since neither group would be moderated there was no point to their both being created, since the difference would be ignored. This plan overcomes the objection. (I really think the successful vote should have been respected [see companion post "Hidden history of comp.groupware"]). I did not propose a moderated group originally because I feel that moderation is seriously abused on Usenet, and the power imbalance created degrades the quality of interaction. I originally preferred the "mutual moderation" approach since this avoids centralized power, but there is a role for some coordination mechanism which indicates the main lines of agreement in opinion. The more experienced reader can always search for minority opinion, based on their own reputation database. With this approach, there would be a couple of articles every cycle on each of the discussion threads in comp.groupware.d. Comp.groupware would carry in addition to the current traffic, referee reports (under the author's pseudonym) that had been processed by the mediator. Weekly statistical reports and copies of mailed invitations issued to new authors by the mediator would also appear in comp.groupware as would any minority opinions. I would prefer to have a separate group for articles meant to be machine readable, but until there is significant traffic in reports, I don't think it is worth pushing for a new group. An alternative would be to have all readable material in comp.groupware.d and people could select articles based upon how they were sent (articles sent as a result of invitations from the mediator would be expected to have a consistent high quality). This approach would make it easier to follow threads (reference or topic connections). As of the moment, I do not know of a way that can be used to distinguish between invited articles and others. This could be indicated in the headers, and then be processed automatically. Such an approach would require that the reader to have some "smarts" in their news reading software. Since this can not be depended upon, a separate group for invited articles might be best. The more sophisticated reader would be able to integrate articles in the invitational group with those in other groups which referenced them. In summary, the following channels of information can be identified: 1) Invited articles 2) Spontaneously contributed articles 3) Reports from referees (machine readable) 4) Mediation statistics and invitations The last (4) would only have information on a weekly basis so it would not be necessary to have a separate group. This information could go into (1) or (2). A precondition for the creation of (1) is that somebody write the mediation script or simulate it for a trial period (see companion posting on "Consensus Journals"). A precondition for (3) is that people start giving ratings in response to articles. When there started to be a lot of these, a separate group could be created. -- David S. Stodolsky Office: + 45 46 75 77 11 x 21 38 Department of Computer Science Home: + 45 31 55 53 50 Bldg. 20.2, Roskilde University Center Internet: david@ruc.dk Post Box 260, DK-4000 Roskilde, Denmark Fax: + 45 46 75 74 01