@rra and I have a new, #OpenAccess article out: "Shifting your research from X to Mastodon? Here's what you need to know"
https://www.cell.com/action/showPdf?pii=S2666-3899%2823%2900323-9
This article is an opinion piece in which we argue that social scientists cannot simply port their work from X to #Mastodon or the rest of the #fediverse. There are key differences in culture, expectations of privacy, and of course topology to consider.
Thanks for this very interesting work. It's certainly only one of the first steps in a lot of research that needs to go into this. For examples, I've found out that when I added some information about the wider study, responses to my polls increased a lot.
But I am writing to ask you whether you've had any thoughts on one big issue: demographics. Any social research would need to frame its validity within the sample of respondents. Since on social networks participants are self-selected, it is useful if we have some data about demographics and representation. The polls provided by the platform are a great tool, because they are much more immediate to access, as opposed to dedicated survey links. However, the way they are currently implemented in Mastodon, individual posts with polls are not bundled to the demographics of their respondents.
One approach I've had in mind was a posting a series of polls in a short period of time, consisting of (necessarily brief) demographics and few questions.
Another, possibly better, alternative would be to have a platform that strongly encourages users to provide demographics data and that packages this with poll responses. It would be a hard sell in terms of privacy, so I wouldn't say realistic.
Did you have similar questions? Any thoughts?
Consent. That's key. The big difference between here and corporate social media is that people can share what they want. And that they aren't positioned via data analysis (e.g., Facebook will make inferences about people's demographics through analyzing their relationships and behaviors).
So demographic details in fediverse-based research need to be opt-in, which might be suboptimal for the social scientist, but... maybe the era of easily gathering data on people should end?
@rwg @rra
Sure. In the EU we have GDPR which is quite explicit about this.
A nice constraint (from the perspective of privacy) about polls on Mastodon is that they don't give you a mechanism to know who responded (in theory one might try to trace back user exposure to posts and time of increments in votes to trace these back, but this is fiction for now). We don't even have a way to confirm that when we provide two polls (one demographics and one on the research topic), the same person would respond to both. So in terms of privacy, we have a conservative setup, which - as you say - is fair game.
However, my claim is that if we inform users about the way we want to interpret data and have similar response rates to the two, we can safely assume that the answers to the demographics question closely relate to the demographics of the respondents to the research question.
Robert W. Gehl
Martin Ruskov