16 ms·

Maybe, maybe not. But I have a number of academic colleagues working on marketing/news/misinformation on online platforms, and Facebook has always been a proble

by pmyteh 8y ago

Maybe, maybe not. But I have a number of academic colleagues working on marketing/news/misinformation on online platforms, and Facebook has always been a problem for analysis. It's next to impossible to get raw data unless you're Facebook, which makes investigating these questions very difficult. For comparison, Twitter is comparatively easy to work with (though getting harder as time goes on) and Reddit is very open. This is one reason why there are lots of papers based on Twitter data and very few based on analyses of Facebook.

bad_user 8y ago
Any legitimate reasons for not allowing access to raw data? Could it violate people's privacy for example?
- SiempreViernes 8y ago
  I'm sure it could, but of course you can always anonymise data first. In any case it seems a bit rich for facebook to restrict access due to privacy concerns, in reality the issues is likely that reason is simply that academics can't pay as well as ad agencies.
  - lvh 8y ago
    Data anonymization is an incredibly complex problem and de-anonymization is an entire subfield. I’m by no means a Facebook apologist, but “we don’t trust our ability to deanonymize data so it would be suitable to publish” rings true.
- pmyteh 8y ago
  Yes, there are substantial privacy implications from social media research. And they're particularly problematic on a platform where privacy controls restrict access to certain data/conversations to particular friends (and Facebook employees, of course). Not everything is like this, though. For example, there's been next to no public data about how the news feed algorithm works and no way of tracking how ads are being targeted. Both of these are commercially sensitive (and it's clearly in Facebook's interest to keep its cards to its chest) but a royal pain compared with, say, newspaper research where you can see what is being served to whom along with which adverts on a day-by-day basis. And there's essentially no public API even to get those things to which you do potentially have access - so there's no way of getting the membership of public groups without web scraping or recording by hand, AFAIK. And certainly the first of these is prohibited by TOS, which means many university ethics committees will reject the research even if the researcher does have the capability to get it. It's basically a large black box. And, as a political scientist, black boxes with political salience concern me.
- detaro 8y ago
  That's certainly something to be careful about. Although one would think that in at least some data about advertising could be made available without much risk - outside of specific targeting lists, how much personal data is gonna be in there?
mankepo 8y ago
Anyone know a marketplace for ppl to sell SARs to reseachers - with filters like: public funding, open results, xyz de-personalisation standards or auditors? Come to think of it any proxy depersonalisation services that maybe cache xyz data then show you how you payload would look were you to participate in the study? Edit: no need to cache, download some js, xslt, rdf or similar to allow you to verify what you were sending locally? Public comment section for discussion of implications of studies and releasing data sets.. lots of interesting possibilities in this space of taking ownership of your data but also sharing it.

by pmyteh 8y ago

bad_user 8y ago

Any legitimate reasons for not allowing access to raw data? Could it violate people's privacy for example?

SiempreViernes 8y ago
I'm sure it could, but of course you can always anonymise data first. In any case it seems a bit rich for facebook to restrict access due to privacy concerns, in reality the issues is likely that reason is simply that academics can't pay as well as ad agencies.
- lvh 8y ago
  Data anonymization is an incredibly complex problem and de-anonymization is an entire subfield. I’m by no means a Facebook apologist, but “we don’t trust our ability to deanonymize data so it would be suitable to publish” rings true.
pmyteh 8y ago
Yes, there are substantial privacy implications from social media research. And they're particularly problematic on a platform where privacy controls restrict access to certain data/conversations to particular friends (and Facebook employees, of course). Not everything is like this, though. For example, there's been next to no public data about how the news feed algorithm works and no way of tracking how ads are being targeted. Both of these are commercially sensitive (and it's clearly in Facebook's interest to keep its cards to its chest) but a royal pain compared with, say, newspaper research where you can see what is being served to whom along with which adverts on a day-by-day basis. And there's essentially no public API even to get those things to which you do potentially have access - so there's no way of getting the membership of public groups without web scraping or recording by hand, AFAIK. And certainly the first of these is prohibited by TOS, which means many university ethics committees will reject the research even if the researcher does have the capability to get it. It's basically a large black box. And, as a political scientist, black boxes with political salience concern me.
detaro 8y ago
That's certainly something to be careful about. Although one would think that in at least some data about advertising could be made available without much risk - outside of specific targeting lists, how much personal data is gonna be in there?

mankepo 8y ago

Anyone know a marketplace for ppl to sell SARs to reseachers - with filters like: public funding, open results, xyz de-personalisation standards or auditors? Come to think of it any proxy depersonalisation services that maybe cache xyz data then show you how you payload would look were you to participate in the study? Edit: no need to cache, download some js, xslt, rdf or similar to allow you to verify what you were sending locally? Public comment section for discussion of implications of studies and releasing data sets.. lots of interesting possibilities in this space of taking ownership of your data but also sharing it.