Thank you for your answers. It's worth noting that we are using the Wikia dataset too and that we are trying to identify when we consider a wiki "death" or "inactive". We are looking for a rule that should be as much as independant as possible of the topic, stage or size of the wiki. Indeed, we could say as well that we are interested in when the wiki community dies and isolate from any external relation (like, for instance, if it's a wiki around a TV serie that nowadays it's been nowadays, we don't care about that external fact, but whether the wiki is still active or not)

On Mon, Nov 5, 2018, 8:37 PM Jeremy Foote jdfoote1@gmail.com wrote:

Hi Abel,

We have been working on a paper that looks at wiki survival on Wikia. We ended up using a similar measure to Zhu et al. We are more interested in when the community around a wiki "dies", and so we measure death as a 30-day period in which fewer than two people edit the wiki. I think that the generalization of this idea is to look for the first period of X days in which N or fewer people make an edit, where X and N might change depending on the question your are looking at (or might be informed by the data).

Best, Jeremy

On Mon, Nov 5, 2018 at 11:35 AM Ziko van Dijk zvandijk@gmail.com wrote:

...

Hello,

Interesting mail - at the moment I am busy with thinking about chronological aspects of a wiki. One idea is that a wiki can have a finality, that means, that the founders had only a limited goal in

mind.

...

If accomplished, the wiki is no longer needed. There is a paper about open source software that you might already know? (Schweik 2014)

Kind regards Ziko

Am Mo., 5. Nov. 2018 um 14:44 Uhr schrieb ABEL SERRANO JUSTE < abeserra@ucm.es>:

...

Hello fellow researchers!

We are conducting a research about "mortality in wikis" and we are

looking

...

for a good definition to determine when a wiki is considered "death", "inactive" or "abandoned".

So far, I've only found this definition from Haiyi Zhu, Robert E. Kraut

and

...

Aniket Kittur in their paper: "The impact of membership overlap on the survival of online communities" https://dl.acm.org/citation.cfm?doid=2556288.2557213. We define a community to be dormant (the inverse of active) in a given month if the community did not have any activity (including discussion pages and community pages) in the given month and the preceding two

months.

...

Any other references you could point me out? any better ideas?

Thank you in advance!

Saludos, Abel. _______________________________________________ Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

---

Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

---

Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

While perhaps not relevant to the original enquiry ...

For Wikipedia, do we need to know (or care) about the "death" of:

* a WikiProject * a Portal * a category * an article

While there are articles on historical topics which might, once written, arguably not need further updating, there are many articles, categories, portals, projects which do as they involve topics that are current in the real world. As a simple example, for a town, we report on population and temperatures. For an electorate, there are elections and new people elected. Sports results etc. If readers visit articles with out-of-date information, they may see less value in the article and therefore Wikipedia as a whole. Have we ever thought of having some system of tagging articles of this nature, either as a whole or in sections or in infobox fields that indicates when the information could be considered out of date. E.g. we have the Australian census every 5 years (the last being 2016). It takes them about a year to release the data to the public (so 2017 we had first access to the 2016 data), so we might think it desirable that all Australian places with census data have been updated by 2018 and certainly we would not think it acceptable if there were any with 2006 (or earlier) census data (except as historical information). Yet of course there are many such out-of-date Australian articles, but probably we don't have an easy way to know which ones. (Before anyone rushes to tell me about Wikidata solutions, I would point out that the average Australian Wikipedia editor neither knows nor cares about Wikidata and our attempt to add 2016 census data from Wikidata more-or-less collapsed from lack of community support). I'm thinking here about solutions that Wikipedians might understand, such as templates which have a tracking category that is activated when the article misses an update deadline based on some template in the article.

Of course, on Wikipedia, many articles have the illusion of being actively maintained in the sense of edits occurring, thanks to vandalism and reverts of vandalism, endless re-categorisation, automated changes of a trivial nature (e.g. dash length), the Internet Archive Bot and other bots, copyedits etc. As someone who does her watchlist diligently, I am seeing increasing activity over articles (my daily watchlist seems to be growing faster than the number of entries on my watchlist) which suggests we are more active, but, when I look at the edits, relatively few of them are updates to the information content. So activity should not be equated to information currency. Note, as anyone who deals with visible metrics soon learns, people game them and our edit counts are a classic example. I sometimes wonder what would happen if we suppressed that information. Or better still, counted something that we value more than "number of times clicked Save". What if we only counted the number of citations added (or counted it in addition to plain old edit count)? Would that drive behavioural change from less information-productive activities towards more information-productive activities?

But if we can have some measure of information-activity/inactivity for an article, then I presume we can aggregate this across any natural groupings of articles (e.g. Category trees, Portals, WikiProjects) to discover where we are stagnating and then let humans decide if that topic space is one that can stagnate (because it is historic) or one that must be updated periodically to be considered useful and whether the correct frequency of updates seems to be occurring, either macroscopically or (ideally) microscopically around particular time-sensitive factoids.

Can we measure "information growth"?

Kerry

Again, thank you for your responses.

As I stated previously, our research is more on the Jeremy Foote's line: We want to study the "wiki survival" or "wiki mortality" in highly diverse wiki ecosystem as it is Wikia and we want to focus more in *the death of wikis as a community*, so probably not so related to how much useful or outdated is the knowledge posted.

The first step to approach this is to define "Death". Indeed, It can correspond somehow with the activity of the wiki, as the activity represents the aliveness of a community. But definitely bots, changes made globally to all wikis by Wikia and, maybe, small format corrections should be kept apart. Thank you all for pointing those out!

There is a major question that maybe is the key to unlock the definition of these concepts:

*Is inactivity equal to death community or is inactivity just an indicator of possible death?*

So, in the first definition I brought out, they used the term "dormant", leaving some room for the hope that the wiki can be retaken or awaken at some later point. So I believe that, what we are looking actually forward is for a definition closer to what Leila said about dropping-out users editors (Thank you very much for the reference!), but applied to wikis, so something like "dropped-out wikis" could be what we are looking for in order to have a proxy for wiki death or wiki abandonment. @Leila: Do you think that we could do some equivalent study to evaluate when a wiki shifts from "inactive" to "abandon" ? Is this what you meant when saying "inactive may or may not be equal to dead in your definition."? Or what did you mean by saying so?

I want to add, and I think it's worth noting, that apart from activity data, we also have data regarding the visits to the wiki. This give us an idea on how much attention or usefulness the wiki is still providing to visitors. But I find this data more related to the interest or, even, "activity" of the wiki rather than the aliveness of the community behind.

El mar., 6 nov. 2018 a las 14:03, Leila Zia (leila@wikimedia.org) escribió:

Hi Abel,

Some years back, Erik Zachte and I came up with a proposal related to your question. (inactive may or may not be equal to dead in your definition.)

A project is considered active if it satisfies the following conditions:

• The wiki should be hosted by the Wikimedia Foundation
• Open (not closed/locked or private)
• The wiki should have less than a very permissive threshold of

activity: At least three active editors with 5+ edits/month/editor needed. This threshold is perhaps too permissive but it already cuts away a large number of almost inactive projects.

We haven't put the above anywhere public (we totally should).

And since you may run into this other question soon, too: If you want to see how many months of inactivity for a username means the user has dropped out of the project, please check section 1B in https://meta.wikimedia.org/wiki/Research:Understanding_thanks#Discussion.

Best, Leila

On Mon, Nov 5, 2018 at 2:05 PM Kerry Raymond kerry.raymond@gmail.com wrote:

...

While perhaps not relevant to the original enquiry ...

For Wikipedia, do we need to know (or care) about the "death" of:

• a WikiProject
• a Portal
• a category
• an article

While there are articles on historical topics which might, once written,

arguably not need further updating, there are many articles, categories, portals, projects which do as they involve topics that are current in the real world. As a simple example, for a town, we report on population and temperatures. For an electorate, there are elections and new people elected. Sports results etc. If readers visit articles with out-of-date information, they may see less value in the article and therefore Wikipedia as a whole. Have we ever thought of having some system of tagging articles of this nature, either as a whole or in sections or in infobox fields that indicates when the information could be considered out of date. E.g. we have the Australian census every 5 years (the last being 2016). It takes them about a year to release the data to the public (so 2017 we had first access to the 2016 data), so we might think it desirable that all Australian places with census data have been updated by 2018 and certainly we would not think it acceptable if there were any with 2006 (or earlier) census data (except as historical information). Yet of course there are many such out-of-date Australian articles, but probably we don't have an easy way to know which ones. (Before anyone rushes to tell me about Wikidata solutions, I would point out that the average Australian Wikipedia editor neither knows nor cares about Wikidata and our attempt to add 2016 census data from Wikidata more-or-less collapsed from lack of community support). I'm thinking here about solutions that Wikipedians might understand, such as templates which have a tracking category that is activated when the article misses an update deadline based on some template in the article.

...

Of course, on Wikipedia, many articles have the illusion of being

actively maintained in the sense of edits occurring, thanks to vandalism and reverts of vandalism, endless re-categorisation, automated changes of a trivial nature (e.g. dash length), the Internet Archive Bot and other bots, copyedits etc. As someone who does her watchlist diligently, I am seeing increasing activity over articles (my daily watchlist seems to be growing faster than the number of entries on my watchlist) which suggests we are more active, but, when I look at the edits, relatively few of them are updates to the information content. So activity should not be equated to information currency. Note, as anyone who deals with visible metrics soon learns, people game them and our edit counts are a classic example. I sometimes wonder what would happen if we suppressed that information. Or better still, counted something that we value more than "number of times clicked Save". What if we only counted the number of citations added (or counted it in addition to plain old edit count)? Would that drive behavioural change from less information-productive activities towards more information-productive activities?

...

But if we can have some measure of information-activity/inactivity for

an article, then I presume we can aggregate this across any natural groupings of articles (e.g. Category trees, Portals, WikiProjects) to discover where we are stagnating and then let humans decide if that topic space is one that can stagnate (because it is historic) or one that must be updated periodically to be considered useful and whether the correct frequency of updates seems to be occurring, either macroscopically or (ideally) microscopically around particular time-sensitive factoids.

...

Can we measure "information growth"?

Kerry

---

Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

---

Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l