stodol@freja.diku.dk (David Stodolsky) (12/05/89)
Mutual moderation can eliminate the problem of authorities by distributing control to users. Say that whenever you read a post you immediately reply either worth reading (Y) or not worth reading (N). Your news reader collects review messages from others automatically and uses its database of previous review performance by these same persons to give a post waiting to be read a priority score. The database shows the correlation (in the simplest case) between your own past judgements and those of others like this (review message also shown): Name correlation review Tom .8 Y Dick .1 N Hari -.9 N Tom, who you usually agree with you, thinks the waiting message is worth reading. Dick, who agrees with you slightly more than half the time, doesn't think it worth reading. Hari, who almost always disagrees with you, thinks its not worth reading. Conclusion: read it. Your best source of review information is Hari, if he doesn't like it, most likely you will. So, you read it and like it. You reply "Y" to the net and the system updates this part of your database to look like: Name correlation Tom .85 Dick .05 Hari -.95 Of course, you also are building up reputation info. on the author and any new reviewers, whose names are added to the database. Learning is really fast, since you update the reputation of the author and all reviewers, each time you review a message. You can also develop a good reputation yourself without posting at all (except for review messages). New users can just look at which posts are getting good reviews, and on the average they will see good stuff (an Emily Postnews view) even without performance information in their databases. Astute authors could see if they are getting bad reviews from people they usually agree with and delete offending posts before they totally destroy their reputations. (A feature like this could do wonders for net bandwidth :-). If this is all done right, the quality of posts (you see) should exceed that of scientific journal articles (assuming an equal quality in authors' contributions). A database updating procedure would be enough to get started. The idea behind the followup group in the *.groupware(.f) new group proposal was to have a separate channel that eventually would be used just for machine readable review messages. This could be done immediately with the creation of *.groupware(.f) if somebody wants to write a few lines of code. More details in: Stodolsky, D. (in press). Protecting expression in teleconferencing: Pseudonym-based peer review journals. Canadian Journal of Educational Communication. ([1989, May 9] Communication Research and Theory Network [CRTNET], No. 175 [Semi-final draft available by electronic mail from LISTSERV@PSUVM.BITNET at University Park, PA: The Pennsylvania State University, Department of Speech Communication and COMSERVE@Vm.ecs.rpi.edu at Troy, NY: Rensselaer Polytechnic Institute, Department of Language, Literature, and Communication) Abstract The social environments of educational systems are less than ideal because power differentials exist that can suppress the free exchange of ideas. One solution is to strengthen personal integrity with an anonymity shield. Many text-based conferencing systems permit anonymous contributions, but this often leads to irresponsible behavior. If people are limited to one and only one pseudonym, however, responsible behavior can be expected. This reputation preserving anonymity overcomes the problems with traditional systems. A reputation is developed through peer evaluation which is based on routinely elicited judgments. Evaluative judgments of a message by one person can be available to all other potential receivers of that message immediately. Evaluations can then be used to automatically select messages worth reading. This approach deals effectively with the problems of both information overload and irresponsible behavior while providing the highest possible protection of expression. -------- Much more detail in: Stodolsky, D. (1988, October). Scientific publication through electronic media. Appendix to a research proposal, Selekterende elektronisk publikation, submitted to the Danish Natural Sciences Research Council. -- David S. Stodolsky, PhD Routing: <@uunet.uu.net:stodol@diku.dk> Department of Psychology Internet: <stodol@diku.dk> Copenhagen Univ., Njalsg. 88 Voice + 45 31 58 48 86 DK-2300 Copenhagen S, Denmark Fax. + 45 31 54 32 11
rayan@cs.toronto.edu (Rayan Zachariassen) (12/13/89)
I wrote some shell scripts and an rn macro to facilitate this; my scheme was to have a duplicate hierarchy under a new toplevel which mirrored the existing hierarchy (e.g. digest.comp.unix.wizards). It would append the name of the reviewer to a file etc. I didn't pursue it mostly because it got to the point where someone had to write code (along the lines you mention) to make it really useful, and I was also worried about the immense added volume in just passing "I like this article" type control messages around the net. Somebody (Chuqui?) later said they (actually Erik Fair, I think) had also thought of this kind of system and termed it "accolades", which I think is a very appropriate name. This kind of system would even be useful in a local context, where scores of semi-trustworthy grad student types read every newsgroup there is, and some specialize in various groups due to domain knowledge in the area of the group. The people who would like to follow all these things but find the thought of wading through all this cruft depressing (and therefore don't), would be enticed to read news again if the good stuff had already been filtered out. rayan ps: here's the posting I'm referring to: From news.misc Sat Jun 24 16:08:11 1989 Path: jarvis.csri.toronto.edu!rutgers!apple!chuq From: chuq@Apple.COM (Chuq Von Rospach) Newsgroups: news.misc Subject: Re: In Moderation: A Moderator's Response Message-ID: <32547@apple.Apple.COM> Date: 20 Jun 89 15:44:13 GMT References: <632@biar.UUCP> <2213@qiclab.UUCP> <2258@drilex.UUCP> Organization: Life is just a Fantasy novel played for keeps Lines: 58 [ ... ] >1. You could get your feed from anywhere, and transmit it anywhere. >In Moderation Network would supply you with a list of cruft article IDs; Gads. Erik Fair wrote an article for 'login:' a few years back that covered something very similar to this concept. Called the "accolade" named by yours truly, it had people sending control messages around the net 'congratulating' a specific article -- and you could set your system to only read messages with a specific minimum number of accolades. Alternatively, you could look for accolades by a given net-address who's judgement you trusted. Good ideas always seem to come back, although I don't believe it's practical large numbers of accoladers. A few specific editors, though... >2. You could get your feed from anywhere, and transmit it anywhere. >In Moderation Network would supply you with a feed of cancel messages. >Each of these would be frobbed up to ensure that they will be accepted >by your news software Unfortunately, both of these ideas carries one serious negative (at least, from the point of IMN). Once a site trashes an article, it won't retransmit downstream. This means that a site like mcvax could buy the IMN service and all of Europe would live on its coattails. Any group of sites willing to funnel through a single point could then accept the services of IMN passively without paying for them. Besides, one of the main advantages of IMN that I see is getting the trash off my modem. Doing the deletion after the fact actually increases traffic load rather than decreases. It also has timing problems -- if you read your news between the time the article comes in and IMN pops in it's batch of deletion messages, you see all the articles you're paying to not see. chuq Chuq Von Rospach =|= Editor,OtherRealms =|= Member SFWA/ASFA chuq@apple.com =|= CI$: 73317,635 =|= AppleLink: CHUQ [This is myself speaking. No company can control my thoughts.] You are false data. Therefore I shall ignore you.
shoulson@cunixf.cc.columbia.edu (Mark E. Shoulson) (12/13/89)
[ mutual moderation by y/n votes of readers...]
GAK! that's just too messy for words. I'm not sure about most readers,
but I, at leasty, don't want to be troubled by such a cumbersome procedure
just to read mail. Worse, there's the enormous decision-making process
involved. Most people read news for recreation, and don't feel the need to
weigh alternatives so carefully just to find out the latest news on Donkey
Kong (or whatever, even serious topics). Sure, there's always the decision
of whether or not to read a given article, and this proposal only tries to
make that decision easier, but it seems that it's just adding too much
confusion.
For another situation, this is a pretty neat idea, but I don't think it has
much value on the net. If someone wants to write some code to try it out
locally somewhere we could get an idea of how people feel about it.
~mark
DISLAIMER: If you don't like my opinions, wait 'til you see Columbia's.
o o o o o o o o o o o o
o o o o o o o
o o o o o o o o o o o
Mark Shoulson: shoulson@cunixf.cc.columbia.edu shoulson@cunixf.bitnet
{...}!rutgers!columbia!cunixf!shoulsondavidbe@sco.COM (The Cat in the Hat) (12/14/89)
She said that he said that she said that stodol@freja.diku.dk (David Stodolsky) said:
< correlation moderation info removed >
-New users can just look at which posts are getting good reviews, and on the
-average they will see good stuff (an Emily Postnews view) even without
-performance information in their databases. Astute authors could see if they
-are getting bad reviews from people they usually agree with and delete
-offending posts before they totally destroy their reputations. (A feature like
-this could do wonders for net bandwidth :-).
A good idea, but it doesn't deal at all with subjects; only authors. I may
disagree with someone's opinion, but I may still want to read what they have
to say on the subject, your groupware posts as an example (1/2 :-) ). And
subject headers occasionally (often) bear little resemblance to the message
content.
-The idea behind the followup group in the *.groupware(.f) new group proposal
-was to have a separate channel that eventually would be used just for machine
-readable review messages. This could be done immediately with the creation of
-*.groupware(.f) if somebody wants to write a few lines of code.
Go for it. Write the "few lines of code". Personally I think it would take
more than that; then you have to convince people to install it. Which is,
coincidentally enough, one of the things the groupware group was designed
to discuss.
--
David Bedno, Systems Administrator, The Santa Cruz Operation, Inc.
Email: davidbe@sco.COM / ..!{uunet,sun,ucbvax!ucscc,gorn}!sco!davidbe
Phone: 408-425-7222 x5123 Disclaimer: Speaking from SCO but not for SCO.
"You pull the trigger, and the gun fires. Dumb, dumb, dumb." - robertb@sco.COMstodol@freja.diku.dk (David Stodolsky) (12/18/89)
dce@icky.Sony.COM (David Elliott) writes: >How do you handle the case of someone being "permanently censored"? >[...] >So, is there a mechanism for dating the info so that a person that >changes their ways can "live down" their bad reputation? Its up to each reader to decide whether to "age" their database. For instance, with nn you can choose to "kill" a given author's messages either for a month or permanently. Even if all information is retained without ageing, a person could recover their reputation if a good message from them was read by one respected reader. A good rating from this reader would immediately get the message read by others. Also, new users would more likely read messages from a discredited author. Once a new user got the "Emily Postnews" view, they would, if highly interested in the group, probably investigate different points of view within a group, including minority opinions. If a good reputation was developed with many new users, it could get old users to read messages from the discredited author again. Finally, if you wanted to build up a good reputation fast, you would look for good messages from discredited or new authors. When others changed their readiness to accept such message based on your evaluation, your own reputation would become very good (everybody would read your messages and/or follow your recommendations). On the other hand, someone who added their "its good" review to a message that had received that response by many others already, would not be enhancing their reputation by much. -- David S. Stodolsky, PhD Routing: <@uunet.uu.net:stodol@diku.dk> Department of Psychology Internet: <stodol@diku.dk> Copenhagen Univ., Njalsg. 88 Voice + 45 31 58 48 86 DK-2300 Copenhagen S, Denmark Fax. + 45 31 54 32 11
stodol@freja.diku.dk (David Stodolsky) (12/18/89)
In Message-ID: <602@scorn.sco.COM> davidbe@sco.COM (The Cat in the Hat) writes: >A good idea, but it doesn't deal at all with subjects; only authors. I may >disagree with someone's opinion, but I may still want to read what they have >to say on the subject, your groupware posts as an example (1/2 :-) ). I was trying to keep it simple by limiting evaluation to a single dimension, but since you asked, here is the overall framework (Stodolsky, 1988): Stodolsky (1984) suggested a model for structuring dialogue based upon categories of criticism related to validity claims (C - F). Even cooperative responses (A & B) to statements can be critical of the manner of presentation (Grice, 1975). Priority to cooperative responses gives the following ranking: (A) Addition or correction offered; (B) Additional information needed; (C) Do not understand; (D) Not sincere; (E) False; (F) Disagree. A review message could contain one or more of the above. Generally, a person would not be expected to respond to a lower criticism if responding to a higher one. That is, it doesn't make much sense to say a message is "false" if you do not understand it. However, in general, review messages can be considered to be a vector of responses. Depending on your objectives, you could tune your reader to select referenced messages based upon the values of the vectors received. For instance, if you were in the mood for controversy, then messages receiving criticisms of "disagree" would be selected more often. The question of content is really a separate issue, but it is possible to think of each message being described on a vector of newsgroup names. Each person would also have a vector of this type indicating the groups that they read. If many people reading news.groups rejected a message (as irrelevant) that had been cross-posted to news.admin, and if news.admin (only) readers thought it was relevant, then this information could be used by pure news.groups readers to skip the message. This type of system would also generate a lot of data for name space management. That is, it could be used to measure how "close" different names were to each other. ============ Reference: Stodolsky, D. (1988, September). Self-management of criticism in dialogue. Fourth European Conference on Cognitive Ergonomics, Cambridge, UK: European Society for Cognitive Ergonomics. ============= -- David S. Stodolsky, PhD Routing: <@uunet.uu.net:stodol@diku.dk> Department of Psychology Internet: <stodol@diku.dk> Copenhagen Univ., Njalsg. 88 Voice + 45 31 58 48 86 DK-2300 Copenhagen S, Denmark Fax. + 45 31 54 32 11
stodol@freja.diku.dk (David Stodolsky) (12/18/89)
rayan@cs.toronto.edu (Rayan Zachariassen) writes in <89Dec13.033005est.2334@neat.cs.toronto.edu> >I was also >worried about the immense added volume in just passing "I like this >article" type control messages around the net. This would be the first order effect. But consider the drop in volume that would occur when certain people found out that they were getting negative reputations from posting in certain groups. Or if they couldn't take the hint, that no one was reading their postings. The overall objective of the moderation is to give people enough feedback so that they will not post things that are not worth reading. If mutual moderation was generally adopted, then review messages would get net wide distribution and referenced messages would be dynamically distributed. That is, if a message got good reviews, then a lot of systems would ask for it. Your software would always be looking at the review messages and when a referenced message passed your criteria, it would be requested. Sort of what happens manually with book reviews. Such a requested message could also go on to disk of each system it passed through, where it would be retained for a while. So if you were requesting a message that was already in demand, it would probably be on the next system (or your own) already. In _Babar: An Electronic Mail Database_ (SSL Technical Report [P88-00015], Xerox PARC, April 1988) Steve Putz said that mail management is faced with the problems of how to handle the categorizing and shared access. These problems also seem to be present in the Usenet news system. News and mail need to be integrated in some way for mediated communication to function effectively. Doing this would also probably eliminate the need for votes on newsgroup creation, since newsgroups would not function much differently than mailing lists. I can see two extremes for document distribution, and both of these reduce traffic. The first concerns documents that almost no one reads. These can fly around on the author's disk as long as the electric bill is paid :-) in the hopes of being discovered. The second concerns eternal gems of wisdom that are constantly referred to. These would just sit permanently (except for the rare update :-) on every machine's disk waiting to be read. So what is needed, for an effective solution, is an integration of news, mail, and archives that balances transmission and storage costs given the mix of messages and readers. -- David S. Stodolsky, PhD Routing: <@uunet.uu.net:stodol@diku.dk> Department of Psychology Internet: <stodol@diku.dk> Copenhagen Univ., Njalsg. 88 Voice + 45 31 58 48 86 DK-2300 Copenhagen S, Denmark Fax. + 45 31 54 32 11