[Air-L] Text/Data Mining Software Suggestions: for YouTube, Facebook & Instagram?

Tue Nov 10 05:21:40 PST 2020

Dear Stuart,

Thanks for you reply, but I feel that this is a bit disingenuous. I should certainly have expanded on my "if the topic warrants it", but I think that there are many ways to balance our need for further understanding of platforms and platform life with ethical research practices in ways that include scraping. The AoIR ethics guidelines already provide a nuanced guide to how to weigh different concerns - and respect for user privacy is not the only moral value that drives our decision-making. And not everything is a slippery slope that needs to be dealt with in absolute terms.

I would also like to remind you that the legality of scraping is by no means as clear-cut as you present it here, neither in the US nor in Europe. Even public institutions like the Conseil Supérieur de l’Audiovisuel in France are publishing studies based on scraping. (https://www.csa.fr/Informer/Toutes-les-actualites/Actualites/Pourquoi-et-comment-le-CSA-a-realise-une-etude-sur-l-un-des-algorithmes-de-recommandations-de-YouTube) With respect to the GDPR, I would recommend having a look at the recent opinion statement by the European Data Protection Supervisor (https://edps.europa.eu/sites/edp/files/publication/20-01-06_opinion_research_en.pdf) which highlight the "special regime" research affords within that regulation.

While my comment was certainly too quick and underdeveloped, let's not pretend that this is a clear-cut affair; and how we weigh the different considerations that come into play will necessarily involve compromise and balancing of different value concerns. Given the enormous amount of power platforms wield in our societies, and not simply through their capture of data but already by the sheer fact of mass connectivity, I find it dangerous to exclude certain research methods from the outset.

All the best,
Bernhard

> On 10 Nov 2020, at 12:43, Stuart Shulman <stuart.shulman at gmail.com> wrote:
> 
> I feel that "if the social relevance of the topic warrants it" is not an argument university human subjects review panels can use to defend techniques that break laws and violate privacy. Although I try my best to support students and faculty doing socially relevant studies, I draw a red line when it comes to knowingly violating platform rules. I have watched this debate evolve for a decade and sought carefully to engineer compliance into our tools. Meanwhile, concerned academics are writing papers about the API lockdowns as if their own disregard for laws and norms (ex., the right to be forgotten) is not one of the root causes of the lockdowns. Not the sole cause, as there are strategic financial drivers impelling lockdowns, among other legal landmines. However, if you are one of the many thousands of academics sitting on spreadsheets of social data you have never checked for deletions before doing analysis and have no plans to delete all of that data at some point, I would suggest you are failing a basic test for ethical and legal research practices. While some well-funded research groups are working hard, for example, to use the compliance stream with stored Twitter data, the vast majority appear to treat stored social data as if it were the exactly the same as a stamp collection or a scrapbook of newspaper clippings you might store forever and claim complete ownership over. It is not. By pursuing research without the same level of ethical and legal compliance routinely required for interviews and focus groups (ex., we de-identify data, destroy the recordings, and delete the transcripts after the research is complete), the anarchic world of scraping, storing, and mining the personal data of millions of people mimics the very things we like the least about the platforms. Whilst we may deem our own research to be warranted irrespective of any or all laws and norms, anyone from any perspective could use that argument to study anything using any method, no matter how invasive, insensitive, or harmful to the research subjects. After a decade on Twitter as @stuartwshulman, I recently deleted my account. Many of you reading this post may have stored some Tweets I wrote about politics, sports, and growing garlic as well as my family, dogs, and close friends. Please be advised I request that you delete 100% of my Tweets from your databases. If you can comply and actually do so, good work, as you are well ahead of the curve. If you cannot or will not, you should report to the research compliance office at your university (especially in Europe) and explain why you cannot find and delete that data for me, and every other former Twitter user like me who may not have thought of using a listserv post to flag this issue related to my right to be forgotten.
> 
> 
> Dr. Stuart Shulman
> 
> U.S. Soccer Federation C-Licensed Coach
> Northampton High School Boys Varsity Coach
> 
> 
> 
> On Tue, Nov 10, 2020 at 6:44 AM Bernhard Rieder <berno.rieder at gmail.com> wrote:
> Dear colleagues,
> 
> I would like to disagree with Brooke here. Facebook data can still be accessed through non-scraping based API-access, most importantly the awesome Facepager.
> 
> For Instagram, scraping is indeed the go-to technique (instaloader works very well) and I would like to defend the idea that ToS should not hinder researchers if the social relevance of the topic warrants it. Adhering to corporate policy is not the gold standard for what independent research should strive for, in my view. Proposing topics to people at Facebook may be a strategy for certain topics, but for anything that does not fit within the narrow interests of the platform, this will most likely go nowhere.
> 
> For YouTube, you can also check out the YouTube Data Tools that I have been maintaining here: https://tools.digitalmethods.net/netvizz/youtube/
> 
> All the best,
> Bernhard
> 
> 
> > On 10 Nov 2020, at 05:22, Brooke Criswell via Air-L <air-l at listserv.aoir.org> wrote:
> > 
> > Facebook and Instagram are strict and according to terms and conditions
> > they don't allow any data scraping.
> > 
> > Best try is to propose your study to a researcher at Facebook
> > 
> > On Mon, Nov 9, 2020, 2:21 AM Alexandre Leroux <alleroux at ulb.ac.be> wrote:
> > 
> >> Facepager for FB and YT it has a user interface and a decent documentation.
> >> 
> >> There are scrappers for instagram but those don't comply with the
> >> platform terms of use and afaik are terminal only.
> >> 
> >> 
> >> On 6/11/20 14:59, Cristina Migliaccio wrote:
> >>> Dear Colleagues,
> >>> 
> >>> Advance apologies if this question has been addressed (as I am certain it
> >>> has been) in some previous forum/email---does an easy to use text/data
> >>> mining software/platform exist that works across these 3 social media
> >>> platforms: YouTube, Facebook & Instagram?
> >>> 
> >>> I would like to collect data on alphabetic features but also
> >> paralinguistic
> >>> features such as likes, shares, etc.
> >>> 
> >>> Any suggestions whatsoever for a text/data mining beginner would be
> >> greatly
> >>> appreciated (videos, lectures to this end also appreciated!)
> >>> 
> >>> Warm thanks-
> >>> Cristina Migliaccio
> >>> _______________________________________________
> >>> The Air-L at listserv.aoir.org mailing list
> >>> is provided by the Association of Internet Researchers http://aoir.org
> >>> Subscribe, change options or unsubscribe at:
> >> http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> >>> 
> >>> Join the Association of Internet Researchers:
> >>> http://www.aoir.org/
> >>> 
> >> 
> >> --
> >> Alexandre Leroux
> >> Ph.D candidate
> >> Group for research on Ethnic Relations, Migrations and Equality (GERME)
> >> Université Libre de Bruxelles (ULB)
> >> alleroux at ulb.ac.be
> >> _______________________________________________
> >> The Air-L at listserv.aoir.org mailing list
> >> is provided by the Association of Internet Researchers http://aoir.org
> >> Subscribe, change options or unsubscribe at:
> >> http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> >> 
> >> Join the Association of Internet Researchers:
> >> http://www.aoir.org/
> > _______________________________________________
> > The Air-L at listserv.aoir.org mailing list
> > is provided by the Association of Internet Researchers http://aoir.org
> > Subscribe, change options or unsubscribe at: http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> > 
> > Join the Association of Internet Researchers:
> > http://www.aoir.org/
> 
> _______________________________________________
> The Air-L at listserv.aoir.org mailing list
> is provided by the Association of Internet Researchers http://aoir.org
> Subscribe, change options or unsubscribe at: http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> 
> Join the Association of Internet Researchers:
> http://www.aoir.org/