Hi Leila, First of all thank you for the quick response, and pointers to great resources on prior work. I don't think I've read all those papers, so it's a great to catch up with the prior art. I'm sorry my last message didn't explain the full context of my inquiry. I'm collecting data for a project which is directed by Julia Kamin, Nathan Matias and myself (civilservant.io) and Aaron Halfaker (WMF). We are currently finalizing an MOU with Foundation, and it looks like some of the links you provided point to Emily Harburg's work with Natha Matias as well. The idea is to actually conduct randomized trails on the effects of gratitude in *non-english* Wikipedias — this isn't an observational study[1]. We plan to co-design the study with community, so the exact parameters aren't finalized yet. In addition all the data I'm collecting at the moment is only for power analysis. Also, chicovenacio on IRC likewise told me that the revision information tied to each thanks is not public, so that's just too bad. Thanks again for your help. [1] *Testing capacity of expressions of gratitude to enhance experience and motivation of editors* https://meta.wikimedia.org/wiki/Research:Testing_capacity_of_expressions_of… Make a great day, Max Klein ‽ http://notconfusing.com/ On Wed, Jun 13, 2018 at 11:16 AM Leila Zia <leila(a)wikimedia.org> wrote:

If the data on the specific edit is not public, it may nonetheless be able to be guessed. According to the public log, my last (genuine) thanks (User:Kerry Raymond) was to User:Ozesoldier on 11 June. Since we know the date of the thanks, what edits did Ozesoldier do prior to that? Well, that one is easy. That user did 4 edits to the same article Anton Hettrich, having edited nothing else for two years. So yes I thanked in relation to Anton Hettrich. Which of the 4 edits? I have no way of knowing (nothing shows in the history, which is interesting as it does show when it is recent history) and to test if I could find out which of the 4 edits I thanked, I just thanked the user for all 4 of edits to see if any of them said “already thanked”, and it didn’t, but only 3 new thanks appear in the public log, so under the hood, the thanks system knew one of the thanks was a repeat and ignored it). Having said that, I encountered these edits on Anton Hettrich via my watch list and saw the diff of 4 edits rather than diffs of 4 separate edits. Now the thanks system doesn’t allow you to thank a diff of multiple edits (even if made by the same user), so to thank, you have to the extra step of doing to the history and thanking a specific edit, which I do somewhat randomly as I am really thanking for the group of edits. I typically pick the edit of the group that added the most bytes, but it can just be the first one my mouse reaches. So it’s important to understand that a thanks is probably not 100% linked to a specific edit when it occurs as part of a sequence of edits done around the same time. It may mean “I like what you are doing to this article” rather than “I like the way you removed that comma”. So I would argue that you don’t need to do about the specific edit, but that knowing the specific article probably suffices. So can we work that out. Well, if we assume thanks are a response to a watchlist notification (mine almost certainly are of that type but maybe others have different behavioural patterns), then the article in question would be in the intersection of “recently” edited by the receiver of the thanks (which is public information) and on the watchlist of the giver of thanks (not public Information). However, why do you watchlist something? Again, for me, it’s pretty simple. I watch articles I have made a contribution to, by default, and later remove those that generate a lot of watchlist activities but are topics about which I do not deeply care or to which my own contribution was housekeeping rather than intellectual). But I think everything in my watchlist is going to be something I contributed to. And, yes, I had previously edited Anton Hettrich (I started it) and this is public knowledge. So, based on my own user behaviour (which may or may not be typical) I would be tempted to suggest that the article Z that is the cause of the thanks from X to Y must be a recent edit by Y (that occurred before the timestamp of the thank) to an article previously edited by X. And that article (or set of articles) is computable with public knowledge. What do we mean by “recent”? I am honestly not sure, but if it is very recent, it’s faster to compute, so practical computation limitations may determine how recent you are prepared to consider. Maybe you just work backwards through the Y’s contributions until you find an article that the X previously edited as an approximation. Clearly the further you work back the larger the set of candidate articles becomes. If thanks are watchlist triggered, I would think that “recent” would be a one month or less. So while the data is not public, maybe you can make a fair guess at least about the article that is involved and from that the single edit or group of edits that likely to have attracted the thanks. But whether these approximations are adequate for your task depends a lot on your research question. Kerry Sent from my iPad