Certainly there are a lot of sites with badges that do seem to encourage certain behaviour. On Wikipedia, we have edit count and that seems to generate editcountitis which (when gamed) tends to favour lots of little housekeeping edits over content edits. But one of the things with badges on most sites is that the site assigns the badge. Here on Wikipedia, I can put any badge I want on my User Page (the pre-existing ones are mostly edit-count based but I can roll my own as some users do). Indeed as I discovered, other people can put badges on my user page and presumably take them away. As edit count is our primary KPI, it doesn't address "cultural" attributes. Should we be making more of an effort to promote other KPIs that emphasise positive behaviour like thanks (given and received)? Unfortunately our main interaction mechanism is writing on talk pages and it's hard to tell whether any contribution on a talk page is a "positive" behaviour or a negative one (short of some kind of sentiment analysis). This is an unfortunate consequence of using a wiki for a conversation rather than some more purpose-built tool.
In principle one takes a KPI and then creates a badge to reward a behaviour that improves that KPI. But that's all easier said than done. For content improvements, there are probably some things we can do. For example, I presume looking at the edit deltas, we could tell if an edit to an article added a citation (a pair of ref tag in the new version that weren't there in the old version). Adding citations is a desirable behaviour that we could report on and give badges for (although obviously whether or not that citation in any way supports the claim cannot be determined, so the "gaming" of this is to add random citations to offline sources to lots of articles, which cannot be easily verified). In which case maybe we need to give a better score to an online citation on that grounds it is more likely to be verifiable). But positive "culture" or positive social behaviour is harder to detect and reward. For example, we'd like to close the gendergap but firstly we don't have KPI that measures it on an ongoing basis because we don't actually know which contributors are male/female. And even if we had that KPI, what users or their behaviours would we reward for having positive impact on that KPI? In real-life, we might reward a customer who introduces a new customer. Or we might have a "finders fee" for someone who introduces a "new hire". How could we reward introducing new women to Wikipedia or encouraging them (perhaps through mentoring) to contribute more? Or would we reward contributors who contribute to articles about "women's topics" (which is addressing the content gendergap rather than the contributor gendergap, which aren't the same thing although many believe them to be closely linked). [I won't disgress into the challenge of deciding how "female" an article topic is.] On some sites, you need certain badges to "unlock" certain extra functionalities. Are we happy for RfA to be a question of collecting up enough badges? AFAIK, the only auto-implemented badge we have on Wikipedia is the "auto-confirm" (4 days and 10 edits from memory). I think badges are a good idea but I think the way Wikipedia is implemented makes it challenging to machine-identify desirable behaviours to reward (particularly for social/culture metrics). I think badges have (in the most part) to be machine-calculated and awarded or else it just becomes a popularity content (who's mates with who). I know Aaron (or someone) was toying with the idea of putting a value on each edit (presumably based on some training set of edit data that humans rated). I think it's not impossible to come up with some set of dimensions on which an edit might be valued and, using some human evaluations on a test set, come up with some kind of values for each dimension. It might be rough in the first instance but I guess if it incorporated some ongoing feedback mechanism, it could improve over time. A cheap thing that we could do (and I don't think we do) is have edit count badges for "last week", "last month", "last year". ATM we only have "lifetime" counts, which makes it hard for the new user to get any quick positive acknowledgements for their efforts. Kerry -----Original Message----- From: Wiki-research-l [mailto:[email protected]] On Behalf Of Robert West Sent: Tuesday, 6 October 2015 1:05 PM To: Research into Wikimedia content and communities <[email protected]> Cc: Marti Johnson <[email protected]>; Patrick Earley <[email protected]>; Jacob Orlowitz <[email protected]> Subject: Re: [Wiki-research-l] Reinforcing or incentivizing desired user behavior This paper is on using badges to steer user behavior: https://www.cs.cornell.edu/home/kleinber/www13-badges.pdf On Mon, Oct 5, 2015 at 7:42 PM, Pine W <[email protected]> wrote: > Hi all, > > Some of us plan to have a conversation at the WCONUSA unconference > sessions about ENWP culture. Are there any recommended readings that > you could suggest as preparation, particularly on the subject of how > to reinforce or incentivize desirable user behavior? I think that > Jonathan may have done some research on this topic for the Teahouse, > and Ocassi may have for done research for TWA. I'm interested in > applicable research as preparation both for the unconference > discussion and for my planned video series that intends to inform and inspire > new editors. > > Thanks, > Pine > > > _______________________________________________ > Wiki-research-l mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/wiki-research-l > -- Up for a little language game? -- http://www.unfun.me _______________________________________________ Wiki-research-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wiki-research-l _______________________________________________ Wiki-research-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wiki-research-l
