Wikipedia talk:Flagged revisions/Trial/Archive 5

We're going to need a new Userbox

This user immediately approves without review all flagged revisions encountered.

Seriously, has anyone given a thought to the issue of how many people will be doing just that, and how much more difficult it will be to fight off vandals without killing legitimate edits made since? — Hiddekel (talk) 20:02, 29 January 2009 (UTC)

I imagine after the first few people get blocked for disruption, not many. Mr.Z-man 20:18, 29 January 2009 (UTC)

They won't even have to be blocked if we make it easy to turn off someone's sighting power if they keep sighting bad articles. --Arctic Gnome (talk • contribs) 22:36, 29 January 2009 (UTC)

No, if we install FlaggedRevs and I see someone willingly sighting revisions that contain vandalism, copyright violations, or BLP violations, I'm just going to block them for disruption, though I guess removing their ability to sight articles should be removed for good measure as well. Mr.Z-man 23:15, 29 January 2009 (UTC)

If the editor doesn't understand the content of the edit, he may accept it without question. Can we allow ignorance to rule the roost on this? doktorb _words^deeds 00:52, 30 January 2009 (UTC)

I think common sense should be the rule here. Ideally, an editor would only flag revisions that they understood to be 100% true and correct. In reality, users are probably going to have to flag edits through that they are 99% sure are okay. Inevitably, some poor edits are going to find their way through. So long as these are isolated genuine mistakes, it's not really a big problem. If editors are either letting larger amounts of bad stuff through due to carelessness or maliciousness, then yes, either blocks or revocation of their sighting privileges should be in order, and I won't have a problem doing that myself. Lankiveil ^{(speak to me)} 02:57, 31 January 2009 (UTC).

Explanation of "Sighting"

Please thoroughly explain the Wikipedia term of art "sighting" on this Wikipedia page: Sighting. I have seen so many references to the term in the discussions and it is bandied about as if it were instantly understandable to all, which it is not, since the usage is as a term of art. Clearly it means more than simply a revision is "seen" by somebody. Hu (talk) 20:54, 29 January 2009 (UTC)

I have attempted a stub definition of Sighting. There is more information about Sighting at Wikipedia:Flagged revisions/Sighted versions. (I am not involved in specifying or promoting Sighting or flagged revisions, but I agree that we need precise definitions.) Certes (talk) 02:38, 30 January 2009 (UTC)

Should we redirect Wikipedia:Sighting to Wikipedia:Flagged revisions/Sighted versions? I didn't find it in my search before I started the page. Hu (talk) 02:51, 30 January 2009 (UTC)

My suggestion: A collection of trusted and registered wikipedia users that review articles. A link on each article to last reviewed version if the article has since been changed by anonymous user (what made wikipedia)!? ---Sorry if for this probably being totally misplaces, first time using talk pages, I trust in an elder to locate this suggestion properly :)--- —Preceding unsigned comment added by 85.220.17.95 (talk) 04:03, 31 January 2009 (UTC)

Support and Opposition by editor experience: a study

Full study is complete and in last section.

Beginnings

I wanted to know how experienced editors were who were for or against the flagged revisions proposal, so I researched total edit counts of a small sample. I started at #301 for support and #201 for oppose (opposes hadn't hit 300 yet).

Update: now not so small, n=120. My results, supports #301-360 and opposes #201-260):

Edit Count	Support	Oppose
>10000	24	12
1000-9999	21	29
100-999	9	14
<100	6	5

If anyone wants, continue consecutively from where I left off (support #326, and oppose #226) and report your results here. Just make a copy the table for your results and I'll integrate the data. Diderot's dreams (talk) 03:35, 19 January 2009 (UTC)

Comments

Too small of a sample size to mean anything, IMO. Try researching editing time instead of edit counts. After all, when people like j.delanoy support it, everything gets messed up. NuclearWarfare _(Talk) 04:01, 19 January 2009 (UTC)

I was editing it as you wrote just because of edit count average skewing as you said. I'm redoing it, dividing the votes by edit count range, not using an average. As for sample size, I know that's why I put so far. The title's a bit of a teaser to get people to contribute, but I'll change that too! Diderot's dreams (talk) 04:12, 19 January 2009 (UTC)

Sounds like work for a bot. If you want, I can offer to write one by tomorrow (I don't believe a read-only bot requires approval to run). Melissa 4.0 (talk) 09:34, 19 January 2009 (UTC)

A bot certainly would be ideal to get the job done. All bots require approval to run, though, and must be written by an "editor in good standing". Your account has only made a few edits. If you can get it approved, then it would be great, but I don't see how that could happen. Maybe you have another account?

Another approach would be to save the entire voting page to a home computer, then extract the user names and votes, query SQL tools (an off-wiki site) for the edit count, and tabulate the data. This wouldn't require running a bot on Wikipedia.org. Can you write something like that? If you can, please go right ahead. And could you include the source code? Diderot's dreams (talk) 15:19, 19 January 2009 (UTC)

That's what I had in mind. Is Wannabe Kate good for getting the edit count, or is there something more efficient? I'm still undecided about publishing the source code. Melissa 4.0 (talk) 15:59, 19 January 2009 (UTC)

X!'s edit counter is a bit faster. _(Talk) 17:10, 19 January 2009 (UTC)

My main concern now is mass automated use of these tools a violation of Wikipedia/Wikimedia policy without bot permission, etc.? I am going to ask someone in the bot approval process about this now.

OK, I've looked at bot policy and if does't edit anything, it's not a bot. It's just a analysis tool and Orange Mellon's comment below applies. Diderot's dreams (talk) 18:45, 19 January 2009 (UTC)

Oh noes, an unapproved read-only bot!1!! What people don't know will not hurt them (plus search engines do this continuously). — CharlotteWebb 21:19, 19 January 2009 (UTC)

Which tool to use? Wannabe Kate is OK but slow and undercounts a bit (I tried it on my account), we should avoid it if possible to avoid criticism. I couldn't get X!'s edit count to work. Can you guys? I have been using | SQL's tools.

Thanks so much for offering to write a program, Melissa. Before I run it on my computer, I need to see the source as you are effectively an anonymous editor. It doesn't have to be released publically, but can be emailed to me. Diderot's dreams (talk) 18:14, 19 January 2009 (UTC)

If you only want the total edit count, just use the API [1]. This is super-fast because the number is stored and incremented every time you edit, rather than counting the edits each time you check it. Deleted edits are not subtracted from it, if that matters. — CharlotteWebb 21:19, 19 January 2009 (UTC)

What wasn't working about it? I could check it fine. CharlotteWebb, my edit counter does use the API. Xclamation point 02:11, 20 January 2009 (UTC)

Now that's just silly to use the API for one step of the program when you already have database access. I guess the best way to do it would be to use the API as a backup for everything, but only when it cannot connect to the toolserver database. That's besides the point. If people only want the total edit count they are not going to want to put names in your tool one at a time. That would take too long and contain a bunch of information they don't need. — CharlotteWebb 04:20, 20 January 2009 (UTC)

It wouldn't give any results when I tried yesterday afternoon (more than once). It worked fine just now. Diderot's dreams (talk) 13:01, 20 January 2009 (UTC)

Progress statistics from the poll

I have been doing some analysis of my own, although my focus is more on the changes over time than the background of the voters. Some results are shown in the graph to the right. I note several things. Firstly, the percent support has been largely static for about ten days now, at almost dead on 60%. This notwithstanding the considerable fluctuation in net support over the same period. This confirms to me that the poll is reaching the stage where significant further variation is extremely unlikely, given the very large weight of the existing contributions compared to new votes. In addition, the rate of contributions is declining steadily, and the time between votes is rising correspondingly.

I have not yet made any attempt to extract the user associated with each vote or any corresponding background data on them. I may do so at a later date. Happy‑melon 17:33, 19 January 2009 (UTC)

Thanks for your comment. The poll does seem to be winding down, but the question of how differently experienced voters responded is still valid even if the poll is over. The issue of flagged revisions doesn't end with this poll, so the question is still out there: do more experienced users favor flagged revisions more, and if so then why? My study is to determine the answer to the first question. And that is enough for now, so let's please not discuss about why at this time. Diderot's dreams (talk) 17:50, 19 January 2009 (UTC)

Automated read only activity works on the out-of-sight-out-of-mind basis. Wikimedia serves over six billion page views a month. If you do something automated that's on a sufficiently epic scale to be at all noticeable, you deserve to be mauled for it. But this will be merely a drop in the ocean. Using a messy combination of excel, notepad and a quick python script, I now have the raw data required (editcounts, group membership and registration date for all participating users) for a full analysis. Give me some time to analyse it. Happy‑melon 18:22, 19 January 2009 (UTC)

Yeah I checked bot policy, and what we are talking about isn't a bot by definition. My concern was violating policy, I realise that 1000 automated queries of SQL's Tools or whatever isn't going to impact Wikipedia servers that much. Since you've gone ahead on your own, could I please have a copy of the raw data so I can finish my study? Diderot's dreams (talk) 18:47, 19 January 2009 (UTC)

Study results

	Sup	Opp	Neut	Net+	%+
By user group
Administrators	129	48	4	81	71.3%
Rollbackers	117	66	3	51	62.9%
Checkusers	7	0	0	7	100.0%
Oversighters	7	0	0	7	100.0%
Anons	1	2	1	-1	25.0%
Non-admins	253	216	4	37	53.5%
All	382	264	8	118	58.4%
One edit one vote
All	5,949,184	2,857,468	115,979	3,091,716	66.7%
By edit count
0<edits<100	11	10	0	0	52.4%
100<edits<1000	35	54	0	0	39.3%
1000<edits<5000	93	66	2	0	57.8%
5000<edits<10000	70	40	0	0	63.6%
10000<edits<50000	150	86	5	0	62.2%
50000<edits	21	6	0	0	77.8%

edit count	votes	pct.
Participation by Edit Count
0 edits (anons)	4	0.6%
0<edits<100	21	3.2%
100<edits<1000	89	13.6%
1000<edits<5000	161	24.6%
5000<edits<10000	110	16.8%
10000<edits<50000	241	36.9%
50000<edits	27	4.1%
Total	654	100.0%
Participation by Sysop Status
Administrators	181	27.7%
Non-Admins	473	72.3%

I have completed my analysis (apogies for not seeing your request, Diderot, I haven't checked this page since my previous post). The summaries are in the table to the right; they make interesting reading.

It seems that FlaggedRevisions is in general more popular amongst users with higher permissions: all analyses show an increase in support over the baseline. The admittedly very limited samples of checkusers and oversighters are unanimous, which is notable in itself (for reference, there are currently 37 checkusers and 34 oversighters on en.wiki).

The "one edit one vote" row shows how the votes stack up when the voice of each editor is weighted according to how many edits they have made; so a user with ten thousand edits effectively casts ten thousand votes, while an editor with only a hundred edits casts 100 votes. Obviously this is not a particularly equable way to judge a poll, but it is very interesting to note that the balance of net and percentage supports closely mirror the one-editor-one-vote tallies.

The breakdown by edit count is perhaps the most interesting result. Support appears to be stronger amongst those with very few edits than amongst those with slightly more; indeed there is a region where there is net opposition to the proposal. For improved gradation I have created and uploaded another graph, visible to the right, which orders voters by their edit count rather than by the time they submitted. The height of each line is essentially the number of voters who took a particular position and had less than x edits. The blue line is net support, which briefly dips below the x axis before rising to its final value of +118 or so. net support count hovers very close to zero for editors with less than a hundred editors, reaching a high of +2 and a low of -3 (this section of the graph contains 25 editors). It dips sharply below the axis to reach a low of -22, before rising back to zero and beyond. The point at which the line crosses back into the positive is at around 4,250 edits, there are 254 contributors below this point and 400 editors above it. Very interesting.

My source data files are extremely messy but I can publish data from them if anyone still wants the raw values, although I can't upload them here as .xls files are blacklisted. If anyone wants more or different statistics (I still have a set of account registration dates that are thus far unused), just say the word. Happy‑melon 21:09, 19 January 2009 (UTC)

A completely random fact: the editors who have commented on this poll are responsible for over three percent of all edits to en.wiki. Happy‑melon 22:36, 19 January 2009 (UTC)

I have added results for non-admins, and a table of participation by edit count. Interesting results. Diderot's dreams (talk) 23:33, 19 January 2009 (UTC)

Here's my interpretation of the study: The voting results show divisions by experience about flagged reviews. Those with "some experience" with Wikipedia are opposed to FRs. (100-1000 edits, by a 20% margin). Those with less experience seem evenly divided, although the sample is maybe too small to tell. Those with more experience (>1000 edits) increasingly favor FRs. This verifies what has been said before by the opposition: that FRs are one more way to make WP more complicated and difficult for the less experienced to contribute, and easier for the more experienced to get their way in editing conflicts.

The results say something about Wikipedia democracy too. The "some experience editors" are underrepresented in the voting, only 13.6% of voters. They must be a much higher percentage of editors than that. Certainly "expert editors" (>5000 edits) are not nearly 57.8% of editors, which was their pct. of voters. The "inexperienced" and anons are hardly represented at all. So the decisions at WP are made by the "expert editors".

Why don't the less experienced vote? Three factors, I think. They are (1) unaware of the procedures or its importance, or (2) are too busy learning the ins and outs of editing the encyclopedia, or (3) are busy making edits in their initial enthusiasm. Their ignorance is perpetuated by the prohibition about canvassing and the way voting is done-- nobody is sent a ballot, there are no set election dates, there is just an informal notice when they log in, if they log in.

For flagged revisions, this study shows that consensus does not really exist for its implementation. Only 53% of non-admins agree, and a large class of editors, for whom the effects will be different are opposed. Maybe that's OK for a trial run, but for a final implementation the support must be greater and more uniform.

Diderot's dreams (talk) 20:25, 20 January 2009 (UTC)

I don't think I can agree either with your analysis or the majority of the conclusions you draw from it. You make the extremely common statistical mistake of failing to consider possible systemic errors, and estimate their maximum possible impact on the results. For instance, you correctly note both that a smaller percentage of the population of users in the 100-1000 edit range contributed than did other populations, and that the percentage of those users who did contribute that supported was also lower. It is statistically equally plausible that the sample taken from that population was biased in opposition to the proposal (for whatever reason) as it is that the 100-1000 edit population is overall less supportive of the proposal. Since the sample was self-selecting (rather than randomly chosen), the presence of such a bias is in no way inconsistent with a completely fair ballot; it merely means that users in that population who were opposed to the proposal were more likely (for whatever reason) to contribute than users who supported. I suspect that you are at least partially correct in that the overall level of support in the 100-1000 edit population is lower, but to assert that it is a full twenty percent lower is not supportable. Statistics are dangerous tools.

Your second analysis (of the relative balance of involvement in the decision-making process) is framed in such a way as to reduce the impact of such statistical uncertainties; you are probably correct to conclude that highly-active users make up a larger proportion of the sample who voted than the active editor population as a whole. However, remember that said population is itself biased: an active user is more likely to have a high edit count (as only active users have the opportunity to increase their count), and a user with a high edit count is proably also more likely to be active (as they are more likely to be committed to the project). Taking the simplest possible situation in which edits to a wiki are distributed completely randomly across pages, and hence the edits to a voting page are a random sample, you would still expect to see a higher proportion of those edits being made by the more active users, as each highly-active user has mor edits to be potentially selected. Imposing the condition that only one edit from each editor can be counted (a feature of a ballot) and the proportion of highly-active users' edits included will rise further.

I agree with some of your conclusions, but others appear entirely arbitrary: to conclude from an apparent reduced support amongst users with lower edit counts that "FRs are one more way to make WP more complicated and difficult for the less experienced to contribute, and easier for the more experienced to get their way in editing conflicts" is entirely unsupportable. The implication, although I know you are merely repeating others' comments, is that the statistics 'prove' that the proposal is a diabolical scheme to suppress users with lower edit counts, which those insightful users have seen through and oppose. There is simply no way you can support such a sophisticated assertion with these simple figures.

The conclusion "So the decisions at WP are made by the 'expert editors'" is more defensible, and appears superficially correct. My issue here is to challenge the underlying assumption that this is necessarily a Bad Thing. Wikipedia is explicitly not a democracy, and implicitly is somewhere between a noocracy and a timocracy: the weight of users' 'votes' is dependent primarily on their level of respect within the community, and on how coherent and persuasive their arguments are. Most importantly, users gain "respect" primarily by demonstrating a commitment to, and constructive record in, building and maintaining the encyclopedia. This is why we give vandals, trolls and sockpuppets zero weight in discussions, and give more weight to more respected users. As a high edit count is correlated with (although not directly related to) increased 'respect', as is higher user rights, it is probably neither surprising nor detrimental to see greater influence being wielded by more 'respected' users, provided that this is an entirely natural situation and is not being encouraged or enforced. By summarily discounting the opinions of the 181 administrators who participated to note that "Only 53% of non-admins agree", you discarded fifty five percent (4,878,294) of the edits involved. That is, the group of administrators, who make up just a quarter of the contributors, have made over half the edits. Should they be accorded a correspondingly larger voice? No, of course not. Should their voice be artificially suppressed because of a perceived imbalance? Equally not. Statistics are a dangerous tool. Happy‑melon 18:34, 21 January 2009 (UTC)

Thank you for the reply. First let me say I do not think there is a conscious scheme to thwart newer editors. There is a phenomenon occuring, but it is not a conscious scheme. I would have said so if I thought it was. Now some replies to your specific points.

I haven't failed to consider systemic bias in the results for "some experienced" editors. One doesn't discount results because it is theorectically possible to have systemic error, such error is always theoretically possible. You need to present a compelling argument or evidence of a specific systemic error. This you haven't done. 89 people voted, and voted "NO" by a 20 point margin. You must accept the fact, or show this systemic error.

One thing I will say. Eighty-nine may be too small a sample size. There is some possibility that the actual percentages of the entire group could be substantially less or more. Not likely, but possible.

I see no way to discount the conclusion that higher count editors have participated at a much higher rate. If you just count active editors, I'm sure the result would be the same.

I did not conclude that "FRs are one more way to make WP more complicated and difficult for the less experienced to contribute, and easier for the more experienced to get their way in editing conflicts" because of the statistics presented. The statistics simply support that conclusion. My analysis is not just a scientific analysis of the data, since this page is a discussion the issue of FRs. The conclusion is my belief based on my own experience, and the experiences of others.

I am not disregarding the opinions of "expert editors", I am just saying the total vote does not represent consensus, because of edit-count class support differences and razor thin support amongst all non-admins.

Thank you for correcting me that Wikipedia is not a democracy. I always understood that for article content debates. I knew also for policy making, but these straw polls and talk of consensus really fool you into thinking it is one. After all, what is consensus but "agreement by some number above a majority" coupled with "amongst varied groups or interests."? The former is democracy. I thought we had an advisory democracy when it comes to making policy. But more experienced editors have more power in the decision making process, because voting and advising is more known and accessible to them for the reasons I stated in my interpretation, and a lack of potentially correcting mechanisms, such as campaigning. The participation rates make that clear.

This study has opened my eyes that we don't have an advisory democracy, but an advisory aristocracy. So let me comment on our monarchy with an advisory aristocracy. Expert editors have much advisory power in creating and shaping policy. This naturally leads to policies which favor them, and the needs and interests of underrepresented groups are likely to be underconsidered, and over time degraded. Further, I have noticed a certain tendency towards contempt and arbitrary dismissal of lower count editors. We have fewer new editors who stick around these days. This leads to stagnation. All of these things are characteristics of aristocracy, and reasons not to have it.

All types of the editors are important to the success of the encyclopedia: anons, "inexperienced", "some experience", "expert editors": all should have equal access to voting. We recognize the superiority of democracy in some organizations, like countries. We need to recognize it here.

I am also against the idea of monarchy, here or anywhere where it wields power. Even a benevolent, sagacious, and successful monarch. But lets leave that for another time.

Diderot's dreams (talk) 15:02, 23 January 2009 (UTC)

Thanks for the response, I think you have a better grasp of what is being done with this analysis than many who try to present "statistics". I've tried to avoid drawing many conclusions of my own from these data because of the inherent danger: it is extremely difficult to use statistics to "prove" anything. So you're right that data should not be discounted because it has potentially large errors, it just means that we must take even greater care in using it to justify positions. The deeper you try to read into a set of data, the greater your errors grow, until you cannot say with any certainty that the statistics actually say anything at all. If you read into data just far enough, you really can use statistics to prove literally anything.

As such, I think you have the correct approach to say that your analysis was not merely a scientific analysis of the data; you are correct in this environment to add your own conclusions (although you should be careful to note where you do so). The inherent weakness of pure democracy is precisely that: to reduce everything to number counting - to statistics, in fact - so that large numbers of opinions can be counted quickly and fairly, the system discards all consideration of the merits of the arguments presented. In a democracy there is no fundamental difference between a vote placed after hours of careful thought, and a vote placed entirely at random by someone who really doesn't give a damn. That such votes are counted equally is a travesty, but in a system such as a nation where quantifying and considering every actual argument presented is logistically impossible, there is no acceptable alternative: since not all arguments can be considered, all must be discarded, otherwise you have a disenfranchised majority ruled by those who are allowed to present arguments.

I think this is a key consideration in Wikipedia's governance: we are perhaps the largest organisation in current existence (possibly in history) which still enforces the principle of weighing arguments rather than votes. Being essentially unique in that aspect, it is not surprising that the principle is easily forgotten, and you are correct that when people see a "straw poll" they conclude that it is a straight ballot which can be counted in the 'traditional' democratic way. That is not the case; we hold only a handful of truly democratic votes here, even the ArbCom elections have scope for powerful arguments to outweight powerful voting blocs. A straw poll really is not a vote, and we confuse ourselves to think that it is. Here, as everywhere, it is the quality of argument that counts, although as we see the huge number of people involved makes it very difficult to see what those arguments were. You can see just by reading a handful of !votes that there are many people whose arguments are simply irrelevant to the matter at hand; they provide no insight whatsoever in to why the user supports or opposes the actual proposal under consideration. If the straw poll is evaluated in the "wiki way", with emphasis placed on arguments rather than raw votes, these contributions would be deweighted, just like in an RfA or XfD. Unfortunately there is no one on en.wiki who is both impartial enough to judge the arguments objectively, and who has the authority and integrity such that both sides would actually abide by their decision. This is beyond anything we have attempted to handle before.

I think the other point I would make is in response to your thoughts on the balance of power. I think we tend to easily forget the other reason Wikipedia is a unique organisation, and what sets it apart from nations or corporations. Pretty much every organisation that has a governance, particularly nations, have little purpose other than their own sustenance. A nation is created to provide for the masses what they cannot provide for themselves; fundamentally protection on all levels from things that would do them harm. No matter who governs it and how, its purpose is fundamentally to sustain and protect the governors, be that the entire people in a democracy, or the rulers from the people in an autocracy. Wikipedia is not (supposed to be) like that. We don't exist to serve ourselves, we exist to sustain and protect the 6,827,930 pages that form one of the greatest bodies of knowledge ever assembled, and the millions of people who read it every day. "Imagine a world in which every single human being can freely share in the sum of all knowledge. That's our commitment". All of us.

Now, it must be borne in mind, and it is a hugely important consideration, that that vision can never be satisfied without the communities, indeed the foundation pledges to "listen and take into account our communities in any decisions taken to achieve our mission". The communities are an essential component, but that does not diminish the fact that they exist to serve the encyclopedia, not the other way around. We create a community so that we can write an encyclopedia; by creating an environment that is best-suited to writing a professional encyclopedia, we create and sustain the community around us, so long as people continue to share that same vision. I don't believe that the community has necessarily lost sight of that vision with this latest twist, but I believe it is easy for individuals to do so. The majority of users who contributed to the poll did have the best interests of the encyclopedia at heart, but some didn't. And they are the people whose votes should carry no weight. Determining which votes fall into that category is, of course, the impossible question. Happy‑melon 22:50, 23 January 2009 (UTC)

Independent results

By user group

Category	% Support	Support	Oppose	Neutral	Total	Category %
ip	25.0%	1	2	1	4	0.6%
no-group	49.8%	156	157	0	313	43.5%
rollbacker	63.4%	128	70	4	202	28.1%
sysop	72.0%	144	52	4	200	27.8%
bureaucrat	85.7%	6	1	0	7	1.0%
checkuser	100.0%	9	0	0	9	1.3%
founder	100.0%	1	0	0	1	0.1%
oversight	100.0%	10	0	0	10	1.4%
accountcreator	65.0%	13	7	0	20	2.8%
ipblock-exempt	80.0%	4	1	0	5	0.7%

By edit count

Category	% Support	Support	Oppose	Neutral	Total	Category %
0–100	48.3%	14	14	1	29	4.0%
101–1000	43.2%	41	54	0	95	13.2%
1001–5000	59.2%	106	71	2	179	24.9%
5001–10000	63.9%	78	44	0	122	17.0%
10001–50000	62.9%	166	92	6	264	36.7%
50001—	80.0%	24	6	0	30	4.2%
Total	59.7%	429	281	9	719	100.0%

One edit one vote

% Support	Support	Oppose	Neutral	Total
67.4%	6,601,021	3,060,065	132,115	9,793,201

I went ahead and built a tool to perform the statistical analysis proposed by Diderot's dreams, mostly as a means of double-checking the results produced by Happy-melon and Diderot's dreams above. I got all the data through SQL's Tools (caching it for reruns). In my user groups table, the groups are not mutually exclusive (if a user belongs to both checkuser and oversight, his/her vote would contribute toward the figures in both categories), which explains why the percentages add up to more than 100%.

The figures in my results are higher due to the new votes which were placed after Happy-melon gathered his data (I ran the tool yesterday). The percentages agree with those presented above within a margin of ±2% in most categories. The main exceptions are the 0–100 and 101–1000 ranges; in my results, % Support has decreased in the former and increased in the latter. This might have been caused by users who voted support crossing the 100-edit threshold, although the differences are too small to be statistically significant.

My completely random fact: There are two editors who voted twice each (one voted support twice, the other oppose twice). Since both editors are fairly new (within the 101–1000 range), this could have been an unintentional mistake on their parts. However, it would help if we had some mechanism (such as a bot) to avoid such accidental or deliberate vote tampering from going unnoticed.

Since my tool is almost completely automated (one just needs to check for any missed votes), it might be useful for anyone wishing to run a similar study, even on other polls. I am hoping to be able to make the source code available soon. Melissa 4.0 (talk) 21:32, 23 January 2009 (UTC)

Why are you separating the results by user group? It gives the false impression that the opinion of, say, an administrator is worth more than that of a non-administrator -- Gurch (talk) 21:40, 23 January 2009 (UTC)

The analysis was intended to study the participation levels and voting tendencies of the different user groups, without any prejudice on the merit of their votes. As may be seen from the detailed interpretations offered by Happy-melon and Diderot's dreams above, our goal is not to judge the voters, but to understand why they voted that way. Melissa 4.0 (talk) 21:50, 23 January 2009 (UTC)

Thankyou for running this parallel analysis, Melissa, it is always beneficial to corroborate data independently. And I think you are quite correct in your approach: with both imperfect data, and imperfect knowledge of the statistics required to analyse them correctly, the most constructive approach is to present the data in the most open and accessible way possible, and leave each reader the opportunity to draw their own conclusions. Happy‑melon 22:53, 23 January 2009 (UTC)

{{editprotected}} (outdent) Precisely. I was impressed that you managed to complete your analysis in less than a day, but since I was already halfway through mine, I decided to complete it. I intend to update the statistics now that the poll is officially closed, but there are some corrections which should be made first to the votes page (I can't edit it myself due to the full protection):

1. Strike out the vote by 24.83.204.61. The vote was later signed by Trevor MacInnis (who had voted while logged out), but the original signature was restored in a malfunction of SineBot and retained when the bot was reverted (although Trevor MacInnis's vote was added again). The vote was later simply indented on grounds of it belonging to a single-purpose account.
2. Add the signature for the vote by Matty, which is still unsigned.
3. Strike out one of the duplicate votes by Xenus ([2], [3]) and Rockstone35 ([4], [5]).
4. Some of the sub-section headers are placed incorrectly. Opposes 1–100 contains 101 votes, Opposes 101–200 contains 99.
5. Update the final result of the poll.

Melissa 4.0 (talk) 10:47, 24 January 2009 (UTC)

Between me and User:Neurolysis, we seem to have covered all of these. There are two votes that I very strongly suspect to be sockpuppets; I have asked for a quiet yes/no to "should we indent?" at RfChU. Happy‑melon 12:16, 24 January 2009 (UTC)

Thanks! I am updating my statistics with the latest results, which indicate an increase in support in almost all categories (1.2% overall). Currently, one of the suspected cases is still retained as a vote while the other isn't. I feel that dismissing SPA votes might be dangerous in this poll, since there was a lot of emphasis on involving new users (as you said yourself). I will follow the outcome of the RfChU and update the statistics if there are any changes. Melissa 4.0 (talk) 15:05, 24 January 2009 (UTC)

Just wanted to say thanks for the independent study. And the results for no group users is valuable. Basically a tie. And the "some experience" group support rose significantly, but still losing by 14%. I thought that sample size was a bit too small. And of course late No voters may be discouraged since the decision has already been made in favor. Diderot's dreams (talk) 15:27, 24 January 2009 (UTC)

Or, of course, the thought that the poll was being considered to demonstrate a consensus would have spurred oppose voters to contribute when they might otherwise not have, in the hope that they could reduce the percentage support. Equally the support voters may have been discouraged from voting as it appeared to have already been decided in their favour. One of these factors likely outweighs the other, but it would be foolish of us to believe that we can work out which one it was and separate its effect from any genuine shift in support. Happy‑melon 18:16, 25 January 2009 (UTC)

User group vs. edit count

User group	0– 100	101– 1000	1001– 5000	5001– 10000	10001– 50000	50001+	Total
no-group	25	91	105	40	49	3	313
rollbacker	0	4	68	59	65	6	202 (3%)
sysop	0	0	6	23	150	21	200 (23%)
bureaucrat	0	0	0	0	6	1	7 (47%)
checkuser	0	0	1	0	6	2	9 (17%)
oversight	0	0	1	0	6	3	10 (23%)
accountcreator	0	0	2	9	9	0	20 (167%)
ipblock-exempt	0	0	1	0	4	0	5 (1%)
ip	4	0	0	0	0	0	4
Total	29	95	184	131	295	36	770

Your first two points made me curious about the discrepancy between the results of the "some experience" category (101–1000 edits) and the "no group" category, so I ran some more computations (on the same data). It turns out that 63% of the "no group" users who participated in this poll have submitted more than 1000 edits (and 47% of these submitted more than 5000 edits). Given that vandal reversion is one of the main causes of high edit counts, I find it strange that so many users did not apply for at least the rollback flag, despite being eligible for it. I would guess that either they're happy with Twinkle, or they contribute a lot to article writing. Melissa 4.0 (talk) 16:47, 25 January 2009 (UTC)

"one of the main causes" is not the same as "the only cause". Nor is 1000 edits a "high edit count". -- Gurch (talk) 17:00, 25 January 2009 (UTC)

So? My post did not imply either of those two assertions. Melissa 4.0 (talk) 18:55, 25 January 2009 (UTC)

Hi. Very interesting analyses. Does anyone have access to the size of these groups (admin,...,ip) or (expert,...)? Thehalfone (talk) 13:41, 29 January 2009 (UTC)

The number of users in each explicit group can be determined using the {{NUMBERINGROUP:}} magic word; I've added percentages to the table. Note, however, that these numbers include inactive users, of which there are many in each group; so the percentage of all active users is higher. There is no sensible figure either for all accounts or for all IPs. Happy‑melon 16:48, 29 January 2009 (UTC)

Thanks! Thehalfone (talk) 08:51, 3 February 2009 (UTC)

In the news

See the BBC [6]. DuncanHill (talk) 17:45, 26 January 2009 (UTC)

Why are people so determined to misinterpret this proposal? There are drawbacks to FLR, no doubt, but it's not the dark, radical nightmare provision I'm hearing about from detractors. The phrase "revisions must be approved before they are added" makes it sound like Wikimedia is setting up a Star Chamber or something. Estemi (talk) 17:55, 26 January 2009 (UTC)

Exactly, I just emailed feedback to the BBC about their over simplification of this issue, it is a disservice for them to paint this essentially procedural proposal about one facet of the project as a live or die moment for the community. And failing to even mention the communities moral and legal responsibilities towards all BLP subjects in favor of highlighting two specific cases is lazy and misfocused. A badly informed piece of journalism. Mfield (talk) 18:10, 26 January 2009 (UTC)

Estemi, what is inaccurate about that quote? That's exactly what flagged revision is (modulo nitpicking about how pseudonymous users can see unapproved edits). --Gwern (contribs) 18:33 26 January 2009 (GMT)

Except that it's to be used like semi protection only on articles with heavy problems, and initially as a trial, not on all articles and for all edits as suggest those news articles. Cenarium (Talk) 18:45, 26 January 2009 (UTC)

Ahh, but the eternal problem of who will decide what constitutes a heavy problem and/or good faith, and that based on which conditions, still remains to be solved. Moreover, who will "police the police" - "overlook the overlookers"? Don't worry though, philosophers have been working on it for millenia :)) 82.230.24.185 (talk) 20:39, 26 January 2009 (UTC)

I agree with the above anonymous user. It seems that under this proposal annointed POV pushers will be given more power with even less accountability than in the currenly flawed system. -- noosphere 21:21, 26 January 2009 (UTC)

Don't shift the goalposts, Cenarium. You weren't criticizing the quote "Wikipedia's founder, Jimmy Wales, is proposing a system of flagged revisions, which would mean any changes made by a new or unknown user would have to be approved by one of the site's editors, before the changes were published." (emphasis added), but the quote describing of the actual mechanism itself - which was correct.

(To say nothing of the common viewpoint that this is going to be a slippery slope policy change.) --Gwern (contribs) 21:52 26 January 2009 (GMT)

As I understand the proposal, that quote is not accurate. The proposal is to create a new class of users who will have to approve changes by anyone (not just new or unknown users) before the changes are incorporated to the version of the encyclopedia visible by everyone. -- noosphere 22:14, 26 January 2009 (UTC)

Not quite true depending on how you read the statement. All users are able to see the latest draft version, but the first version they see is the approved version, they simply need to click on the draft link (link design is still being worked on) to access the most current version. All users will edit the most current version by default which is the same as today. Dbiel ^(Talk) 01:10, 27 January 2009 (UTC)

According to this, you have to be logged-in in order to have the option to view the draft version. That means that the vast majority of Wikipedia users (who are not editors, logged in or otherwise) will only see the version that has been approved by the newly created and appointed class of "reviewers". -- noosphere 01:39, 27 January 2009 (UTC)

The description you are quoting is unfortunetaly misleading. You can find a detailed explanation here [7]. In short non registered users are still able to see the latest edits (they are just not the default display). Unfortunately many people might have voted on false premises at least reading many of the angry rejections and the portrayal in the news seems to suggest that and the botched description might have played a role in that.--Kmhkmh (talk) 03:49, 27 January 2009 (UTC)

Click here to see an example of what an unregistered (IP) user would see. Note: this is the default view which is going to be customized for en.Wikipedia so it is not a perfect example, but it will at least give some example of a link to jump between the sighted version and the current draft version. Dbiel ^(Talk) 05:22, 27 January 2009 (UTC)

In the New York Times there was also a mischaracterization similar to the BBC case: [8]. Given the amount of disinformation still floating around, I personally would prefer a new poll, running longer and with on overhauled description, that explicitly explains why that flagged revisions do not break "anyone-can-edit" and that all readers still can see all edits.--Kmhkmh (talk) 03:56, 27 January 2009 (UTC)

Given that the recent poll just turned the software on, it is probably in all our interests to just discuss how, if at all, Flagged Revisions should be used. Likely, if used, it will be in a limited way. Please see here for the list of proposed limited trials. I'd ignore what the press are saying; they basically didn't read what was said Fritzpoll (talk) 11:21, 27 January 2009 (UTC)

Has FR been turned on?? Where has that been announced? Last I saw, the developers had more requirements to be met, before they'd turn it on. - Hordaland (talk) 22:01, 29 January 2009 (UTC)

Not yet. {{Nihiltres|talk|log}} 22:20, 29 January 2009 (UTC)

a different proposal

Wikipedia is patrolled quite well. Ted Kennedy's death was not online for very long, according to news reports. A non-bureaucratic way to deal with this is to keep patrolling and to add a disclaimer.

which is far better than (and which I DON'T support)

The other possibility is to require that every sentence must have a citation, no exceptions. And add the warning. We shouldn't be opposed to a warning because it looks "ugly". Chergles (talk) 23:17, 26 January 2009 (UTC)

Chergles (talk) 23:17, 26 January 2009 (UTC)

You're proposing such a banner for every single article? And requiring citations for every single edit would likely push away a hundred times more editors than FRs ever would. ♫ Melodia Chaconne ♫ (talk) 23:26, 26 January 2009 (UTC)

The banner (top one) is easy and an honest answer. The bottom banner is poorly worded. Citations would vastly improve the quality of articles but make it harder to edit. The citation issue is whether we want quality or quantity. I, myself, use a lot of citations but sometimes I don't. There is a question of whether we want to "push editors away" but get a very high quality product or encourage mass participation and have a product that this sometimes of low quality. This is not an easy question to deal with. Chergles (talk) 23:33, 26 January 2009 (UTC)

The top banner is great - I'd just add the words 'please help make it better'.Riversider2008 (talk) 00:17, 27 January 2009 (UTC)

Facts are subject to correction? Ahem. I sincerely hope not. You surely mean inaccuracies. Kaiwhakahaere (talk) 01:07, 27 January 2009 (UTC)

Replaced "facts" with the word "inaccuracies" Chergles (talk) 16:15, 27 January 2009 (UTC)

I could go for this. Yes, on every article. Perhaps "continued" should be changed to "continuing"? - Hordaland (talk) 22:01, 29 January 2009 (UTC)

this sounds like a more obvious answer than a delayed, internally moderated list of changes to an article. It actually is a correction to the problem (information may not be accurate), not a technical solution to a percieved problem (even with flagged revisions, information STILL may not be accurate) 81.187.12.206 (talk) 18:47, 1 February 2009 (UTC)

Electoral articles and flagged revisions

I am no fan of the proposal to block editors from working on Wiki by this "flagged revision" charade, BUT, as I am far too eager and willing to continue working on projects within Wiki, I am willing to continue as long as i understand some of the practical consequences of this plan.

In June this year, the European Parliament election, 2009 (United Kingdom) will be one election taking place within the European Parliament election, 2009. As many of you may be aware, the updating of candidates and results for these elections will be a major piece of work for those editors who enjoy taking part in what I call "electoral administration". We are the editors who have been filling in election results, constituency boundaries, and political party information for as long as Wiki has been getting going.

My problem with flagged revisions in relation to these "electoral administrative" edits has two questions, which I hope can be answered here. If an editor with a lot of experience wishes to update these articles (and others like them) with results or candidate details, will there be a major delay to this? Will there be editors experienced enough to know that the results being entered are not vandalism?

Flagged revision seems, to me, a total slap in the face of the Wiki project. But my willingness to continue remains strong. I would like assurances, therefore, that in this practical example, the "fr" proposal will not forbid me and others from continuing the work we have done for so long

Cheers, doktorb _words^deeds 17:59, 29 January 2009 (UTC)

If there is no need for protection on that article, flagged revs won't be turned on for it. If there is need of protection, then flagged revs will let it be updated faster because with regular protection like we have now the users would have to first go to the talk page to request the change, and then another user would have to go and make the change. --Arctic Gnome (talk • contribs) 18:27, 31 January 2009 (UTC)

There are 10 possibilities being considered for trials over at Wikipedia:Flagged revisions/Trial/Proposed trials. Only 2 of these is for flagged revisions restricted to those articles that would otherwise be protected (2 and 10). The other proposals would certainly flag articles which would not otherwise be protected. This is the problem with this debate. Everyone has their own idea of what flagged revisions would be, but we do not have a single proposal. I believe that, regardless of what guidelines we gave, many more articles would be flagged than would ever be protected. I think this because it seems rather softer to many people. Thehalfone (talk) 09:59, 4 February 2009 (UTC)

I so heartily agree with thehalfone in this, which bears repeating: "This is the problem with this debate. Everyone has their own idea of what flagged revisions would be, but we do not have a single proposal. I believe that, regardless of what guidelines we gave, many more articles would be flagged than would ever be protected." I can see the comments now: "Well, my idea of what should require flagging was mentioned here in the poll and/or discussions and wasn't definitively refuted. Therefore, there's consensus for it."

What was refuted time and time again, was the idea that we need (or can) define beforehand what FR is to be used for or tried out on. The poll was only about turning on - or not - the capability. Everyone who says what will or won't be done with it is expressing a personal opinion about "our" intentions, no matter how authoritative s/he makes it sound. - Hordaland (talk) 15:03, 4 February 2009 (UTC)

Forget 'flagged revisions' -- just turn anon editing off and be done wit' it

Seriously. Anonymous editors are more trouble than they're worth! Granted, some of them (1%?) actually do make useful edits, but those that do should get accounts. The vast majority of anonymous editors are troublemakers and vandals, and disallowing completely open-editing without accountability is just a recipe for disaster. Furthermore, this middle of the road 'flagged revisions' bullcrap is just going to drag Wikipedia into 'bureaucratic hell', and turn off a lot of truly good editors. You saw what happened when the German Wikipedia turned this cr*p on? Wait times of THREE WEEKS! That's totally unacceptable! Mark my words: FLAGGED REVISIONS WILL ULTIMATELY FAIL.

The only solution for true accountability is to require all users to open an account, which is free, so we're still, "the encyclopedia that anyone can edit." Ok, maybe not "anyone" -- vandals and assholes won't be able to edit, but they can't now anyway because they're quickly reverted.

Wake up, Wikipedia! Stop pushing a flawed philosophical approach, and just do the right thing and turn off anonymous editing! Dr. Cash (talk) 23:04, 5 February 2009 (UTC)

It'll never happen. And no, IP users make a LOT of helpful edits. I would argue they tend to be minor as a rule (a spelling error here, a wikilink there, a date correction over there), but that's a very good thing. They are the exact types of edits 'drive by' editors will make and are needed -- the reason WP works so well is because of the collaboration, multiple eyes seeing things. I can't see how requiring people to create accounts is somehow better than still allowing anyone to edit any time, with a bit of a safety net toward what's displayed. There's other issues with your post, but I don't feel like repeating myself for the eighth time. ♫ Melodia Chaconne ♫ (talk) 23:23, 5 February 2009 (UTC)

vandals and assholes won't be able to edit - Unless of course they make an account, which takes all of 15 seconds. Mr.Z-man 23:39, 5 February 2009 (UTC)

I remmber reading someplace that 73% of good edits come from ips, and only 11% of vandalism.--Patton^t/c 12:43, 6 February 2009 (UTC)

If you remember reading it "someplace" that it shouldn't be to hard to find a ,... Dr. Cash (talk) 19:23, 6 February 2009 (UTC)

I don't know if this is what Patton123 referred to. It mentions the number in opposite context, but further down the article are some interesting findings. 212.200.240.232 (talk) 19:44, 7 February 2009 (UTC)

User:Dragons flight/Log analysis? Hut 8.5 19:26, 6 February 2009 (UTC)

Can we add "Flagged Revisions" to WP:Perennial proposals? Please? What is going on with FR now, anyway? All Hallow's Wraith (talk) 23:23, 6 February 2009 (UTC)

I am also wondering "what is going on" since everything seems to have stalled and there seems to be no established course of action. I am only aware of this page and the pages linked from the "Related discussions" section. Is there discussion going on somewhere else? Barrylb (talk) 05:55, 7 February 2009 (UTC)

Standard response: the essential difference between flagged revisions and no-anonymous-editing is that flagged revisions do not raise the bar for anyone to contribute, it only delays those contributions from becoming visible. It affects the sense of instant gratification, but doesn't add hoops for inexperienced contributors to jump through. Dcoetzee 06:07, 7 February 2009 (UTC)

• This thread's original poster assumes that disallowing anons to create articles would get rid of all newly-introduced inaccuracies in articles, which is simply not true. There is a significant amount of vandalism and well-meaning errors from registered users that need to be addressed in order to increase the accuracy of WP. - Mgm|^(talk) 11:24, 13 February 2009 (UTC)

Flag protection, patrolled revisions and deferred revisions

I have made three proposals related to Flaggedrevs, for which we could run trials. The proposals are a variant of Flag protection, Patrolled revisions and Deferred revisions. Comments would be appreciated to see if there is support for some of them and what to do next. Cenarium (talk) 15:23, 22 February 2009 (UTC)

Wikipedia:Flagged protection and patrolled revisions

It is proposed to run a trial of Flagged Revisions at Wikipedia:Flagged protection and patrolled revisions. The proposal is divided in two parts:

• Flagged protection: an article can be 'protected' by an administrator so that the version viewed by readers by default is the latest flagged version. This is a modified version of the original flagged protection proposal.
• Patrolled revisions: a 'passive' flag used to monitor articles, especially blps, for vandalism, blp violations, pov pushing, etc, that can be used for all articles, but has no effect on the version viewed by readers.

The proposals are independent but supplement each other. They involve the creation of a 'reviewer' usergroup. This implementation can support secondary trials. The main trial should run for two months, then a community discussion should decide the future of the implementation. Cenarium (talk) 22:42, 15 March 2009 (UTC)

A poll has started at Wikipedia talk:Flagged protection and patrolled revisions/Poll. Cenarium (talk) 18:33, 17 March 2009 (UTC)