Why the growth in Wikipedia article count is starting to tail off: one opinion
As a logistic curve, the growth in article count is starting to tail off. Ward Cunningham predicted this at Wikimania in 2006. I would suggest to WikiDashboard's Ed Chi that a limiting factor in the growth in the article count in the encyclopedia is the length of an individual editor's wikipedia:watchlist.
My personal experience is that as a constructivist, I have had to limit the length of my watchlist to less than 75. Any more than that leads to more misery for this editor. [1] In the jungle that is Wikipedia, a new article faces deletion — any edit to it must contribute to its 'fight to survive' or must transform it enough to 'blend in the scenery'. I would guess that watchlist sizes are of similar magnitude (i.e., not thousands) for other constructivist editors who monitor well-trafficked pages. Currently a watchlist is private to the editor, so that the watchlist cannot be gamed by vandals, who must then assume that the vast majority of articles is under the watchful eye of someone. The net result is that about half of 75, [2] or 35 items will show up in my watchlist at any given time. For this editor, admittedly a sample of one, the number of items in my watchlist report is currently:
| Watch for how long | Watched pages with an edit |
| Last hour | 1 |
| Last 2 hours | 2 |
| Last 6 hours | 6 |
| Last 12 hours | 8 |
| Last 24 hours | 12 |
| Last 72 hours | 21 |
| Last week | 30, about half the length |
| All | 60, about the length of my watchlist |
Assuming that one can use information like Ed Chi's statistics, [3] e.g., 50% of all edits are by core editors, this suggests that the maximum number of manageable articles is less than <number of core editors>*<length of their watchlist>*<%articles tended by core editors>*<number of edits per month>.
Thus a maximum number of manageable articles suggests there is an upper limit on the success of any project which is trying to increase its quality, for example by flagged revisions. [4] Before watchlists assumed their critical role, Recent changes was small enough that the community could monitor it and jump in to work on an article together, as late as 2003. [5]
So (for manageable articles) if we plug in numbers, averages, or estimates we get
| Population | Watchlist size | Successful edits | Edits/month | Contribution/month |
| Admins | 75 | .99 | 100 | 1600*75*.99*100 = 11,880,000 |
| Core | 25 | .85 | 10 | 8500*25*.85*10 = 1,800,000 |
| Active | 25 | .75 | 1 | 140,000*25*.75*1 = 2,625,000 |
| Total | 16,000,000 |
This amounts to 5 edits per manageable article per month, which is just about the historical average for the encyclopedia. These averages mask some heroic efforts; it is unjust to characterize 100 edits per month as only bureaucratic. Witness, for example User:Anthere, at one time the Chair of Wikimedia foundation, who once stayed up 24 hours straight to work on the French edition of the encyclopedia (this was before we knew her as Florence Nibart-Devouard). [6]
What does this model have to do with reality? I would suggest that, in order for an article to survive, it has to be of reasonable quality, so that the deletionist forces in the Wikipedia project can find no reason to delete it. Before the deletionist faction gained the upper hand, the constructivist faction was able to increase the number of articles exponentially, each positive edit feeding the mutual interest of others, leading to more mutual edits. This has not been true since 2007. Andrew Lih, User:Fuzheado has estimated that critical mass is about 20 editors per successful article (my personal experience is that this number can be as small as 5). It is currently more difficult to freely create more articles, possibly because so much time has to be spent maintaining the current number of articles. For this editor, at least, attempting to maintain the items on his watchlist consumes his wiki time. [7]
Finally, to state the obvious, no one would add a item to one's watchlist if one did not care about it. -- Ancheta Wis ( talk) 15:45, 26 September 2009 (UTC)
- ^ It was noticeably unbearable for a watchlist of 90 articles. I no longer remember which 90, or perhaps I have blocked the memories.
- ^ This style of reasoning is called a Guesstimate; it is explicitly taught to physicists and/or operations researchers.
- ^ Wikipedia approaches its limits August 13, 2009
- ^ This proposal does not take into account the effects of the vandalism watch by the Recent Changes patrol or the successful function of anyone's project bots. But even assuming that a bot can multiply the manageable watchlist length of a human editor by a factor of 10 or 100 or even 1000, there also needs to be a way to compress the relevant detail to be surveyed by the human editors.
- ^ Here is a year 2003 example of the usefulness of Recent changes: the initial edit history of homunculus shows that within the space of 20 minutes after this article's creation, 5 different editors jumped in to add their contribution, with even more to come in the hours, days, weeks, and years which followed. That sort of mutual cooperation could have only come about when a Recent change occurred only every 20 minutes or more. Today hundreds of changes occur per minute, too many to follow immediately; a new article is 'starved out', to languish after first creation, prey to the first deletionist to find it. The homunculus article started out as a one sentence article; today, even if a inclusionist editor were first to get to it after initial creation, likely as not it would be tagged as a stub, rather than be added-to, and grown organically.
- ^ Anthere has recounted a similar experience as above; one article also started as a single sentence: "An apple is a fruit."
- ^ So perhaps Recent changes might be sub-categorized so that a community of editors might once more simultaneously monitor articles in a specific sub-category. Or perhaps a wikiproject might encourage its members to simultaneously monitor a set of sub-categories as a community.
Why the growth in Wikipedia article count is starting to tail off: one opinion
As a logistic curve, the growth in article count is starting to tail off. Ward Cunningham predicted this at Wikimania in 2006. I would suggest to WikiDashboard's Ed Chi that a limiting factor in the growth in the article count in the encyclopedia is the length of an individual editor's wikipedia:watchlist.
My personal experience is that as a constructivist, I have had to limit the length of my watchlist to less than 75. Any more than that leads to more misery for this editor. [1] In the jungle that is Wikipedia, a new article faces deletion — any edit to it must contribute to its 'fight to survive' or must transform it enough to 'blend in the scenery'. I would guess that watchlist sizes are of similar magnitude (i.e., not thousands) for other constructivist editors who monitor well-trafficked pages. Currently a watchlist is private to the editor, so that the watchlist cannot be gamed by vandals, who must then assume that the vast majority of articles is under the watchful eye of someone. The net result is that about half of 75, [2] or 35 items will show up in my watchlist at any given time. For this editor, admittedly a sample of one, the number of items in my watchlist report is currently:
| Watch for how long | Watched pages with an edit |
| Last hour | 1 |
| Last 2 hours | 2 |
| Last 6 hours | 6 |
| Last 12 hours | 8 |
| Last 24 hours | 12 |
| Last 72 hours | 21 |
| Last week | 30, about half the length |
| All | 60, about the length of my watchlist |
Assuming that one can use information like Ed Chi's statistics, [3] e.g., 50% of all edits are by core editors, this suggests that the maximum number of manageable articles is less than <number of core editors>*<length of their watchlist>*<%articles tended by core editors>*<number of edits per month>.
Thus a maximum number of manageable articles suggests there is an upper limit on the success of any project which is trying to increase its quality, for example by flagged revisions. [4] Before watchlists assumed their critical role, Recent changes was small enough that the community could monitor it and jump in to work on an article together, as late as 2003. [5]
So (for manageable articles) if we plug in numbers, averages, or estimates we get
| Population | Watchlist size | Successful edits | Edits/month | Contribution/month |
| Admins | 75 | .99 | 100 | 1600*75*.99*100 = 11,880,000 |
| Core | 25 | .85 | 10 | 8500*25*.85*10 = 1,800,000 |
| Active | 25 | .75 | 1 | 140,000*25*.75*1 = 2,625,000 |
| Total | 16,000,000 |
This amounts to 5 edits per manageable article per month, which is just about the historical average for the encyclopedia. These averages mask some heroic efforts; it is unjust to characterize 100 edits per month as only bureaucratic. Witness, for example User:Anthere, at one time the Chair of Wikimedia foundation, who once stayed up 24 hours straight to work on the French edition of the encyclopedia (this was before we knew her as Florence Nibart-Devouard). [6]
What does this model have to do with reality? I would suggest that, in order for an article to survive, it has to be of reasonable quality, so that the deletionist forces in the Wikipedia project can find no reason to delete it. Before the deletionist faction gained the upper hand, the constructivist faction was able to increase the number of articles exponentially, each positive edit feeding the mutual interest of others, leading to more mutual edits. This has not been true since 2007. Andrew Lih, User:Fuzheado has estimated that critical mass is about 20 editors per successful article (my personal experience is that this number can be as small as 5). It is currently more difficult to freely create more articles, possibly because so much time has to be spent maintaining the current number of articles. For this editor, at least, attempting to maintain the items on his watchlist consumes his wiki time. [7]
Finally, to state the obvious, no one would add a item to one's watchlist if one did not care about it. -- Ancheta Wis ( talk) 15:45, 26 September 2009 (UTC)
- ^ It was noticeably unbearable for a watchlist of 90 articles. I no longer remember which 90, or perhaps I have blocked the memories.
- ^ This style of reasoning is called a Guesstimate; it is explicitly taught to physicists and/or operations researchers.
- ^ Wikipedia approaches its limits August 13, 2009
- ^ This proposal does not take into account the effects of the vandalism watch by the Recent Changes patrol or the successful function of anyone's project bots. But even assuming that a bot can multiply the manageable watchlist length of a human editor by a factor of 10 or 100 or even 1000, there also needs to be a way to compress the relevant detail to be surveyed by the human editors.
- ^ Here is a year 2003 example of the usefulness of Recent changes: the initial edit history of homunculus shows that within the space of 20 minutes after this article's creation, 5 different editors jumped in to add their contribution, with even more to come in the hours, days, weeks, and years which followed. That sort of mutual cooperation could have only come about when a Recent change occurred only every 20 minutes or more. Today hundreds of changes occur per minute, too many to follow immediately; a new article is 'starved out', to languish after first creation, prey to the first deletionist to find it. The homunculus article started out as a one sentence article; today, even if a inclusionist editor were first to get to it after initial creation, likely as not it would be tagged as a stub, rather than be added-to, and grown organically.
- ^ Anthere has recounted a similar experience as above; one article also started as a single sentence: "An apple is a fruit."
- ^ So perhaps Recent changes might be sub-categorized so that a community of editors might once more simultaneously monitor articles in a specific sub-category. Or perhaps a wikiproject might encourage its members to simultaneously monitor a set of sub-categories as a community.