#musicbrainz

aerozol[m]: Hello again, I hopped on the forums and found this post. Seems like it is the most similar to what I am asking, but it has not seen activity in what appears to be a few years. https://community.metabrainz.org/t/any-plans-to...

I scrolled down and saw some insights from what appears to be the leader (?) of MetaBrainz. Had no idea he had a relation to RJ (one of the founders of Last.fm). I thought that was cool.

I have not asked for the ability to have private profiles, but I would not mind such a feature either. I think making the public data export psuedoanonymous would be a precursor to that. If a user has a private profile, there would then be no public way to link the activity to the public dataset, which would seemingly make the "pseudo" part of that disappear.

Hey techman! Just (re)visiting that thread - It would be amazing to give people control over what is private and what is public on ListenBrainz. But the LB team is too small to add that complexity, and the to-do list is too long, still. Currently it can be achieved by hosting the database locally, but that’s not very casual-user friendly. Perhaps you know some developers and UX designers with some free time ;)

This post also answered something I was asking myself, which was whether it was worth it to make a complaint like this. Turns out that people getting targeted for their music listening history is rather unfortunate.

I think we have a fundamental disagreement with what is anonymous - for me, making up a username and having that link the listens together is no different to having a random number link your listens together. I see no difference between the two, unless you make your username traceable back to you, or we reveal email addresses or something like that.

outsidecontext sums it up a bit better in the thread: “One problem I see here is the way of anonymizing the data. I see two options:... (full message at <https://matrix.moviebrainz.org/_matrix/media/v3...>)

I could make another alias for just tracking user listens, but that separation becomes pointless if I share a link to that profile to others who know me by my main alias (this one)

What changes about that if you send them a link to a profile that has a random number string instead of a username and say “check out my listens”?

I think the sticking point is that I personally think that linking listens together in any way (whether under a username or not) is not truly anonymizing data. Whether corporate scraping robot thinks of me as user ’1295486298546’ or user ‘coolrunnings29’ doesn’t seem important

Well, that is the thing. I am separating the social aspect of a service like this from the public dataset, where the intent seems to be for research and analytics. At the aggregate level, there is no need to have user IDs and usernames directly linked. A random identifier generated for that dataset would probably be sufficient. Still allows grouping.

I was a little lazy with my use of anonymous earlier. After doing a little more research, pseudoanonymous is more appropriate.

That would allow grouping, but it would be veryu easy to look up the data on the website (the social part), and link your username. I agree with atj that giving that illusion of privacy would probably increase risk to non-savvy users

For context I used this as a basis for what to look into further, which points out stuff from GDPR: https://dam.ukdataservice.ac.uk/media/622919/20...

slide/page 6

With profiles public, pseudoanonymous would be more appropriate. The link that the slide talks about would be your profile.

Following this logic, if profiles were private, there would be no public link. Only people with access to the profile (friends, site admins) would be able to link the data to the public dataset.

Right, well ListenBrainz depends on listens to be public to achieve its goals (generate interesting statistics, link with other listeners, etc)

We already hide the information that links the information to a ‘real’ entity, e.g. the email. If a user sets their username to be identifiable that is their choice

What you could say is that user data is *pseudo*anonymous - the only way to link the public data back to a user is by looking at their profile directly

However, doing such a thing seems like a waste of time if you are already looking at the profile in the first place

All of LB is open source. The data dumps are just a convenient ‘package’ to grab the data

If someone is super privacy conscious imo they have two options: A) submit their listens for personal use (to be able to get recommendations and playlists, stats etc), use a random username, and not share their profile with anyone or a select group of people B) not submit their listens to a public database

So, basically, having a random identifier in the dataset generated for a user would be enough to allow grouping (and allow research) but not directly identify individual users. For aggregate data, I think that would be acceptable.

It still might be worth making a ticket to remove usernames from the data dump - but on the other hand it might be good that it spooked you. Because the reality is that your username could be linked no matter what, same on last.fm, etc. It is not quite the same on Facebook etc because they don’t have a public API, but they are just selling your grouped data to corps that can afford it (but the closed nature means it would be very hard for

those corps to ever link it to your real name)

The grouping is the user

But if you think it would be worthwhile replacing usernames with numbers you should def open a ticket: https://tickets.metabrainz.org/secure/Dashboard...

Uh well I deleted the JSON that I looked at yesterday but assume you replace whatever the user ID and username fields are in the output with a UUID4 or something for that particular user. Yes, you could group listens to that UUID, but the only way for you to then find out the user behind said UUID is to have already looked at their profile to make a comparison. At that point, there would be no

point though.

Look at, or have a program or bot that links the whole dataset and re-adds the usernames. That’s my concern

Someone could certainly go through the effort of such a thing while all user profiles are public...but they would need to have a list of all user accounts to do that. It would be some work (probably front loaded) to do this for every bulk data export if the UUID4s are being regenerated every export.

It would be a lot of effort for me to do… but I bet that half the people in this IRC channel could do it in an afternoon :P

Maybe not. But I guess my thing is even if it’s really hard to link ID’s to usernames, if it’s possible then I wouldn’t want to imply to our users that it’s private. And it’s a lot of effort to go to when they could just use a random username that doesn’t link back to their real person

Generally, I would say that I like the idea of ListenBrainz. Public data for research is great. I do think there should be a way to contribute data without having it easily linked back to me, though. At least for the public dataset. You can have UUID substitutes in place of directly identifying users in the export. With public profiles, it is still possible to be linked in said data, which

would make things pseudoanonymous.

psuedoanonymous seems to be the actual term for what I am asking for at this point

Did you know that I can pull your whole last.fm listening history any time? Without needing a data dump

Username and everything

I do indeed use Last.fm currently but I would not mind switching off of it. I do think it is personally kind of bogus to charge a subscription for generating listening reports (among other things). At least for now, it seems like the alternative to dealing with that and their commercial interests is to just put my data publicly on the internet anyway. That is, unless I want to self-host.

Cool, it’s good to be aware that all your data on last.fm is completely public, except for stuff like email address (same as LB)

Last.fm does have some privacy options, but I am not versed on how that shows up in the API. They do not have a public export (at least via searching) so I cannot say what is in there for sure.

If it really is all public though...wow. My mind would be blown.

Well, you can download your whole last.fm listen history into a listenbrainz account by just supplying your username (…or anyone elses)

Oh, I did see that but it looks like I need to have my recent listens enabled.

Hey, I think you’re right! Then they do have more security settings than listenbrainz :P

aerozol[m]: RE: your thing about making a ticket, do you think it is better for me to make a forum post first? At some point I suppose I will need to transfer all of this context into a (hopefully more conscious) set of paragraphs. Removing usernames (but presumably keeping IDs) would be a nice start but the UUID thing sounds like it would need to be pretty fleshed out before any solution is

worked on.

It depends if you want community feedback or dev feedback (forum or ticket, respectively)

Either way the LB team is very busy - making a ticket is a step towards it happening, but if you were invested in getting this done soon you would have to do it yourself or find someone to do it.

But I find it helpful to make tickets just to get things out of my head and feel productive, sometimes. Community discussion can get a bit more heated (but interesting) ;)

Perhaps having both would make sense. There would already be two tickets, one for removing usernames and the other for the UUID thing. A larger "community" post could be a hub for feedback as well, since it looks like people have asked similar things but not exactly what I have said so far.

о>