Share watchlists/2005 Dec 03 irc log/

partial log from Wikimedia Research Network/Meetings/2005-12-03

[20:46:23] <here> would anyone like to comment on the share watchlists idea
[20:46:24] <here> ?
[20:46:24] <JamesF> cimon> Ah.
[20:46:32] <mav> my IRC connection is crap
[20:46:49] <JamesF> here> Isn't that mostly a thing Cormac is interested in? He sent his apologies...
[20:46:58] <mav> oh god ; shared watchlists? that would kill the servers
[20:47:00] <here> actually i rewrote the whole page
[20:47:04] <here> mostly my deal
[20:47:08] <JamesF> Ah, OK.
[20:47:14] <here> Share watchlists
[20:47:16] <Xirzon> mav: not necessarily
[20:47:30] <JamesF> mav> You want a budget? We could just demand more servers! ;-)
[20:47:36] <here> i think rather than code changes, a script workaround to easily copy users watchlists to a subpage would be best
[20:47:41] <here> (not a priority)
[20:47:47] <cimon> are you going to post th log?
[20:47:58] <JamesF> cimon> Yes, of course.
[20:48:04] <cimon> nice
[20:48:13] <JamesF> cimon> In-promptu minutes are already up at http://meta.wikimedia.org/wiki/Wikimedia_Research_Network/Meetings/2005-12-03
[20:48:26] <JamesF> Lagging behind the discussion by about 10 minutes or so, as I get around to it.
[20:48:55] <JamesF> here> Would shared watchlists be part of a multiple-watchlist system?
[20:49:05] <here> not necessarily.
[20:49:10] <JamesF> here> So you'd have a few private and maybe a few public watchlists.
[20:49:15] <JamesF> "High vandalism targets".
[20:49:21] <JamesF> "Policies"
[20:49:26] <here> actually i'd like to see users actual watchlists shared.
[20:49:30] <JamesF> Might be general public lists.
[20:49:34] <here> not necessarily thematically
[20:50:00] <here> see on en:User:Here/watchlist
[20:50:02] <JamesF> Whereas "stuff I've been editing and/or have randomly watch-listed" would be a private list (and what we all currently have).
[20:50:11] <Angela> a form of shared watchlists exists at http://en.wikipedia.org/wiki/Wikipedia:Watch
[20:50:17] <Angela> I don't know if they're widely used though.
[20:50:26] <Xirzon> here: so why have you been watching gay midget photography exactly?
[20:50:27] <k1v1n> i like the idea of people sharing their watchlist. That would be like del.icio.us bookmark sharing. That could be interesting. Who else is interested in what you're interested in.
[20:50:31] <here> "stuff I've been editing and/or have randomly watch-listed" is exactly what I want to see
[20:50:32] <here> and share
[20:50:34] <nichtich> How about letting each user define the status of his watchlist?
[20:50:40] <JamesF> Xirzon> Certainly, yes, there are privacy issues.
[20:50:46] <here> k1v1n exactly, i think we are missing a huge community builder here
[20:50:49] <k1v1n> who else watches this page...
[20:50:56] <Angela> Sorry, that might be the wrong link. It seems mostly about schools. There was another one by topic.
[20:50:57] * mav_ (n=chatzill@c-71-56-70-11.hsd1.ga.comcast.net) has joined #wikimedia-research
[20:50:58] <here> it is absolutely opt-in first and foremost
[20:51:14] * Solensean is now known as Solaway
[20:51:21] <cimon> xirzon: and why do you care? :P
[20:51:39] <here> xirzon: because i'm interested, or editted it.
[20:51:42] * mav_ is now known as maveric149
[20:51:56] <Xirzon> cimon: people who are nosy about the interests of others rarely have a justification
[20:51:58] <here> i think that the privacy issue is way overblown here, and that enough folks would actually be into this.
[20:52:15] <Xirzon> those who want to share their watchlist can put it on their user page already.
[20:52:19] <Xirzon> and use related changes.
[20:52:19] <here> the results would be a user-user network of watched pages (both friends, enemies, leaders, etc)
[20:52:28] <here> xirzon, no one does this as it takes too much effort.
[20:52:39] <here> all of these concerns are addressed at Share watchlists
[20:52:46] <here> including summary of an extensive discussion from 2003
[20:53:03] <here> what is needed is an easy script workaround and uses for the resulting data
[20:53:06] <JamesF> here> Multiple lists could allow "non-privacy-invading" sub-lists, perhaps.
[20:53:17] <maveric149> shared whitelists would be a better idea
[20:53:18] <nichtich> It should be able to 1) opt-in to publish your watchlist 2) add/remove single articles to you public watchlist
[20:53:19] <Angela> I'm not sure that sort of network would be useful. If you look at social networking sites, a lot of people have the sole aim of getting 1000 people on their friends list. With shared watchlists, the aim might turn from watching for vandalism to trying to show who your "friends" are in the hope they watch your page too.
[20:53:31] <here> hmm.. perhaps i really am overestimating individuals willingness to share everything on their list
[20:53:41] <JamesF> Angela> Yeah, there is that.
[20:53:53] <here> nichtich: that would be nice, but not necessary
[20:54:09] <here> angela: i think that noise would be minimal and/or interesting
[20:54:32] <maveric149> my watchlist is already too big ; why would I want to add somebody else's watchlist to that?
[20:54:35] <nichtich> Maybe you want to publish your watchlist but not gay midget photography
[20:54:42] <brian0918> It seems like shared watchlists could end up being used to push agendas in the same was as AIW or ADW
[20:54:55] <nichtich> what is AIW / ADW?
[20:54:58] <here> there is also the ability to see what others are watching, ideally through an xm l output, and even see their related changes.
[20:55:09] <Angela> It could also lead to accusations of stalking if you're watching a lot of stuff that another user edits a lot.
[20:55:09] <brian0918> Association of Inclusionist/Deletionist Wikipedians
[20:55:09] <k1v1n> the reason people opt in on del.icio.us is not to connect with other individuals, it is to find what other people with similar interests also find interesting. It's a mechanism to "discover" new and interesting things.
[20:55:31] <brian0918> They point like-minded people's attention to votes on AFD in the hopes to corral votes in their favor
[20:55:31] * Eskimo (n=Eskimo@wikipedia/Eskimo) Quit (Read error: 104 (Connection reset by peer)?)
[20:55:40] <brian0918> Shared watchlists would probably do the same thing.
[20:55:40] <here> k1v1n: exactly.
[20:55:44] <Angela> we have WikiProjects to discover those things and for people working on similar topics to "meet"
[20:56:10] <Xirzon> Angela: the thing about WikiProjects is that they exist in the wiki namespace and are reviewed to be in line with our policy goals
[20:56:13] <here> would anyone present be interested in sharing their entire watchlist (other than myself?)
[20:56:30] <Xirzon> I'm not sure that would be the case with the way any shared watchlist system is used that is layered on top of everything else
[20:56:39] <k1v1n> sure.
[20:56:44] <nichtich> Do we speak about sharing or just making it accesible like user contributions?
[20:56:52] <Xirzon> if it's just a software feature rather than integrated into the site properly, there is the serious risk of it being used to coordinate sock puppetry, GNAA and other nonsense
[20:56:56] <here> accesible, in that meaning.
[20:57:06] <here> xirzon: how?
[20:57:26] <Xirzon> here: groups forming and using their shared watchlists to coordinate votes, reverts, etc.
[20:57:38] <nichtich> I'm waiting for a method to get my watchlist as RSS for years - so you can combine watchlists from several wikis
[20:57:49] <Xirzon> here: don't get me wrong - I think *grouping* is essential, and better tools for grouping are a good thing to work towards.
[20:57:49] <here> xir: seems like that would be equally possible on the positive side?
[20:57:58] * k1v1n (n=k1v1n@cpe-066-057-113-059.nc.res.rr.com) Quit (Remote closed the connection?)
[20:58:00] <here> nicht: this is highly related to watchlist RSS feeds.
[20:58:17] <Xirzon> here: I'm not necessarily opposed to shared watchlists on principle. I just think what goes on should be transparent - the groups that form should be public, just as the watchlists themselves.
[20:58:37] * k1v1n (n=k1v1n@cpe-066-057-113-059.nc.res.rr.com) has joined #wikimedia-research
[20:58:41] <Xirzon> and then I'm not sure if what we have with wikiprojects is not already very close to those goals.
[20:59:08] * soufron5 (n=soufron@vol75-8-82-233-239-57.fbx.proxad.net) has joined #wikimedia-research
[20:59:31] <here> I'm just extremely curious what watchlist networks actually look like
[20:59:35] <nichtich> Xirzon: do you think of groups as watchlists that are owned by multiple people?
[20:59:43] <Xirzon> nichtich: yes
[21:00:03] <Xirzon> that's one example of what could constitute a group
[21:00:05] <here> i think it could provide insight into community, how people use watchlists, and sharing difficult to find pages -- without much negative
[21:00:11] <Xirzon> but wikiprojects already do this to some extent using topic lists
[21:00:18] <Xirzon> which are then checked using Related changes.
[21:00:36] <Xirzon> you can do this for categories, too.
[21:00:45] <nichtich> Categories are also a way to group articles and make them watchable
[21:00:47] <JamesF> Yes.
[21:00:54] * sannse (i=sannse@wikipedia/Sannse) Quit (Read error: 104 (Connection reset by peer)?)
[21:01:24] <Xirzon> group-forming will be very relevant when it comes to peer review, because then you have to notify people that an article in an area they're an expert in is under review.
[21:01:25] <nichtich> but it's much more easier to add/remove an article to a watchlist than to add/remove a category
[21:01:37] <here> nichtich: precisely.
[21:01:52] <Xirzon> nichtich: maybe it should be made easier to add categories then ;-)
[21:02:02] <nichtich> Xirzon: I agree to yoir sentence before "because"
[21:02:02] <here> again "those random pages on your watchlist" is exactly why this is significantly different than groups and/or related changes vaults.
[21:02:17] <JamesF> Hmm.
[21:02:26] <here> and why manually copying this info is not enough, b/c no one will do it.
[21:02:44] <nichtich> Xirzon: that also
[21:03:09] <Xirzon> here: I'm sure plenty of the researchers here will be interested to look at the patterns in watchlist data, yes.
[21:03:30] <Xirzon> that's more of an academic project without an immediate practical benefit we can predict, I think.
[21:03:50] <here> xirzon: yes, i would agree
[21:04:07] <JamesF> That's another thing, BTW.
[21:04:11] <JamesF> On the agenda.
[21:04:20] <nichtich> Xirzon: but we don't want vandals to know wich pages are not watched
[21:04:33] <JamesF> How to reconcile analysis of watchlist data with our privacy policy.
[21:04:41] <JamesF> But we can skip that for now.
[21:04:43] <Xirzon> nichtich: not sure - if we are faster than the vandals (and we are more) we can make sure that they are.
[21:04:45] <JamesF> It's quite late.
[21:04:49] <Xirzon> but a thing which could be done fairly quickly
[21:04:52] <JamesF> This meeting is starting to drag a little.
[21:04:59] * mav (n=chatzill@wikipedia/maveric194) Quit (Read error: 110 (Connection timed out)?)
[21:05:06] <Xirzon> is to basically implement an opt-in for sharing watch lists, and then export all those which are shared, and let them be analyzed by researchers.
[21:05:07] <JamesF> ... as Mav crashes.
[21:05:27] <here> ok, well thanks for the feedback.... i'll try to garner more support and work on a script workaround.
[21:05:35] <here> any tips or resources for such a script would be appreciated.
[21:05:43] <JamesF> Xirzon> Previous meetings have seem to come the the consensus that opt-in watchlists are by their very nature inappropriately skewed.
[21:05:55] <here> jamesF: i don't see any other way of doing it.
[21:05:56] <nichtich> Do we agree that some method to opt-in to publish your watchlist is wished (and maybe more functionality)?
[21:06:10] <here> necessary evil of self selection, same as many systematic bias issues here
[21:06:29] <k1v1n> it probably should be an opt-in for each item on your watchlist that you would be willing to share.
[21:06:44] <JamesF> here> Just take the data and have the data analysed by a trusted party (one of the developers, perhaps) running provided scripts and returning only the results?
[21:06:45] <Xirzon> JamesF: Yes, it's probably uselessly skewed if you wanted to do a study on, say, relations between articles on sexual topics
[21:06:52] <here> k1v1n, perhaps -- but honestly i'd reather encourage full disclosure.
[21:06:54] <Xirzon> JamesF: but that doesn't mean any analysis of it would be worthless.
[21:06:57] <JamesF> Xirzon> And thus high-view topics.
[21:07:03] <nichtich> JamesF: you can better anonymize the data
[21:07:08] <here> if that had been the case upon arrival, it would be the norm, and the network would be heavily used!
[21:07:09] <k1v1n> you might want to share some (say a research watch page, but not a political watch page), but not all.
[21:07:11] <JamesF> Xirzon> 'Cos our sex-related topics are well-read.
[21:07:14] <here> (full watchlist sharing)
[21:07:27] <JamesF> nichtich> Not really. Almost users will watchlist their own page.
[21:07:35] <Xirzon> JamesF: and well-illustrated.
[21:07:36] <k1v1n> there are sex pages on Wikipedia? I had no idea.
[21:07:41] <nichtich> damn
[21:07:53] <here> heck, even anonymous sharing would be fine!
[21:07:53] <cimon> just have a statistical breakdown; page watched by x byroucrats, Y admins, and z logged in users.
[21:07:58] <JamesF> nichtich> Certainly, there will be a very strong correlation between the edits they make (public information, don't forget), and the pages they watch.
[21:07:59] * soufron (n=soufron@vol75-8-82-233-239-57.fbx.proxad.net) Quit (No route to host?)
[21:08:14] <nichtich> The number of watchers of each page is already very interesting for reasearch and us
[21:08:19] <JamesF> nichtich> I think that it is /very/ unlikely that you couldn't guess who my watch list is of, for instance.
[21:08:28] <JamesF> Yes, exactly.
[21:08:36] <nichtich> JamesF: you're right
[21:08:38] <JamesF> But we seemingly can't release the watchlist data publically.
[21:08:45] <k1v1n> true, anonymous would work although a person could probably put two-and-two together by looking at the edits.
[21:08:54] <JamesF> So we'd have to have a developer (or me, or someone) do that part of it.
[21:08:57] <here> k1v1n: true.
[21:09:04] <JamesF> k1v1n> Exactly.
[21:09:09] <nichtich> But opt-in data with gaps is better than no data
[21:09:11] <Xirzon> oh, one big skewing factor would be people who have the [ ] Watch all my edits option turned on.
[21:09:17] <JamesF> nichtich> Yes.
[21:09:20] <Xirzon> You'd probably want to filter based on that.
[21:09:23] <JamesF> Xirzon> Yes, like me.
[21:09:35] <here> Xirzon: that data is important.
[21:09:36] <JamesF> Xirzon> Except I switch it off as appropriate.
[21:09:49] <Xirzon> here: yeah, but you want to be able to analyze it separately
[21:09:50] <here> again, how people use them, and "all that random junk"
[21:09:51] <cimon> JamesF: it would be useful to have even the watchlist data of a sampling of pages ,, not all of them...
[21:09:55] <k1v1n> good research there on what people are willing and not willing to share.
[21:09:57] <JamesF> Xirzon> Pages I edit but /don't/ watch is probably quite interesting, actually.
[21:10:07] <JamesF> cimon> Interesting idea.
[21:10:36] <here> Bug 727: Number of watching users per page to be shown on recent changes
[21:10:42] <here> (fixed using Enotif)
[21:11:14] <cimon> maybe not identifying even the article in question but for the category it belongs to, and/or the length of it and/or the number of incoming and exiting links...
[21:12:00] <cimon> if you obscure the title of the article, and the identity of the persons watching it, there is no privacy issue
[21:12:22] <here> what data is left?
[21:12:29] <cimon> quite a bit.
[21:12:45] <here> #watchers alongside other individual article metrics?
[21:13:02] <cimon> right, sort of..
[21:13:17] <JamesF> cimon> And it also becomes almost useless.
[21:13:30] <JamesF> Well.
[21:13:36] <cimon> I don't necessarily agree. :)
[21:13:44] <JamesF> Unless you also record number of bytes and number of categories, or whatever.
[21:13:52] <JamesF> At which point the original problem strikes again.
[21:14:05] <JamesF> Because you can work out backwards from that point which article you've got data for.
[21:14:51] <JamesF> And with a dozen articles it wouldn't be a problem, but with a random 10% of the database, say, you could see what the anonymous "editor #34576" watches, and guess who they were.
[21:14:56] <JamesF> Again breaking privacy.
[21:15:00] <JamesF> It's a bugger of a problem.
[21:15:41] <JamesF> OK, well.
[21:15:45] <JamesF> We seem to have stopped.
[21:15:49] <JamesF> Yes?
[21:15:52] <nichtich> yes
[21:15:52] <here> thanks for sticking around for that ;)
[21:15:56] * Solaway is now known as Solensean
[21:16:10] <JamesF> OK, if there's nothing else, I will declare this meeting closed.
[21:16:17] <JamesF> Yes?
[21:16:22] * here nods
[21:16:24] <Xirzon> aye
[21:16:24] <k1v1n> soundg good.
[21:16:27] <JamesF> OK, closed.