Talk:User Interaction Consultation/Who wrote that?

For a discontinued external tool doing this, see WikiTrust. Showing that information would (IMO) be very valuable, but also a rather complex software engineering project.

Yes. This was talked about around 2 years ago but given wikitext compiles to HTML and doesn't give you a sound way to go the other direction this is technically hard to solve.

With parsoid this should be more possible but is still a large investment of work.

That said I have always thought this would be a useful and interesting feature.

I don't think this involves (or is helped by) HTML -> wikitext translation. It can be either be done by figuring out a way to "blame" pieces of wikitext plus figuring out a way to turn wikitext positions into HTML positions, or by blaming HTML history directly. In any case, WikiTrust did this (and more - they also counted editor and edit reliability ratings based on how often that editor's edits survived subsequent changes) well before Parsoid existed.

The task for this is T120739/T2639.

Possibly related: more intelligent diffs (T121469), content persistence research (IIRC @Halfak (WMF) is working on that currently to get editor productivity estimations).

This tool seems to be more potentially useful for editors than for readers.

I concur and amplify. A blame-tool is extremely useful for editors, but it would be an extremely bad idea trying to surface this for readers. There is zero value telling readers that user DingleDoodle wrote a particular sentence, and steering reader comments to editors would be a disaster.

It would depend on how easily and prominently this information is surfaced.

Displaying "freshness" and "trustworthiness" metrics calculated from the author of the text, like WikiTrust did, could be very valuable information for some readers. Adding some extra provenance data like the exact author of a piece of text seems like the next logical step and is certainly very useful for editors. Probably much less so for most readers, though, as they don't have the skills or interest in figuring out if that particular editor is to be trusted or not. Yet, we don't hide the History page for anonymous readers either. If they really want to figure this out then they can do so.

On the other hand, I agree that making this information available too easily or too prominently is not a good idea. I mostly phrased this proposal in the manner it reads now, to get it to fit into the constrained "user interaction" format. I'm not a big fan of how the mobile site is currently displaying the name of the last editor that has touched an article (there are a lot of issues with this, and no clear benefits). And if there's anything we have learned from the Article Feedback Tool, it is that making the bar for interaction too low is not always a good thing.

Do you imagine this kind of system to feed into revision scoring, potentially?

The other way around, I would think. It would be a means to visualize the quality (scores) of particular revisions to readers. (I'm not intimately familiar with either ORES or WikiTrust, but I also think the scores of WikiTrust were more fine-grained than ORES. ORES working on complete revision, WikiTrust working on individual words and sentences in a revision.)