Thread by @cfiesler

There has been a very upset/angry reaction to a paper published using tweets about mental health. I'm not RTing because I'd like to talk about this without drawing more attention to the researchers or the community. But it's an important research ethics cautionary tale. [Thread]

The paper is a qualitative analysis of tweets about mental health care. It includes paraphrased quoted tweets that the researchers ensured were not searchable. The study was approved by an ethics review committee in the UK, and the paper cites the AOIR ethics guidelines.

The paper includes an ethics and consent section that includes the above and notes that because tweets are public, consent was not required. The study also included a researcher with mental health lived experience. There do not appear to be any other statements regarding ethics.

The fact that the research was reviewed by an ethics committee, that there was some consideration for researcher positionality, and that they employed ethical fabrication are steps beyond a lot of research like this.

However, the reaction shows that there was still actual harm.

The lead researcher (a student) tweeted a link to the paper, and replies suggest that despite the paraphrasing, multiple people recognized themselves and others in the tweets--including tweets from people who have since died.

I'm going to describe some harms expressed in replies and quote-tweets from people who identify as part of the community studied:
- feeling violated and exploited with respect to their content being used without consent, without knowledge, and for researchers' personal gain

- feels especially invasive with respect to this being a support resource, and that some of the people studied have died
- by studying content without talking to people, it does not properly include the community's voices or provide the opportunity for clarification

- the attempt at anonymization was insufficient since people were able to recognize themselves and others in the data
- the content analyzed is explicitly people talking about some of their worst experiences, and those experiences should not be exploited for research

- this community, and patients in general, are a vulnerable population that should be treated with special care by researchers
- in support communities on twitter people use pseudonyms to protect their privacy and amplifying their content might counteract those measures

So what should the researchers have done? Some pointed out that asking permission to use tweets quoted in the paper would have mitigated some harms. This is a strategy that I've seen before in mental health related research - not consent for analysis, but just for quoted content.

This example highlights something that is really important to remember--not ALL potential research harms are privacy-related. A finding in some of my work about public reactions to research is that some people fundamentally object to being studied without consent. That is a harm.

The fact that there was some ethical consideration from the researchers here (i.e. paraphrasing), suggesting that they had good intentions, also highlights how the norms of academia haven't caught up with the expectations of research subjects.

It's very important to remember there are no checklists here. Not "is it public," not IRB approval, not anonymization. Every research project & population studied has specific context. This was a vulnerable community that required a lot of consideration for potential harms.

For further reading, this is a summary of some of the work from myself and colleagues on this topic, including @moduloone's and my paper about how Twitter users feel about being studied. howwegettonext.com/scientists-lik…

To add to this, I want to point to another paper we published a couple of years ago, for two reasons:
(1) One finding is that some people fundamentally object to being studied without their consent.
(2) We analyzed publicly available content in this study.
cmci.colorado.edu/~cafi5706/Unex…

The context was the Facebook emotional contagion study; we wanted to understand public reactions to (and harms of) a research ethics controversy, while avoiding the obvious selection bias. (People who hate research are unlikely to participate in research.)

We analyzed public comments on news articles. And in doing so we considered possible expectations of privacy (or not), the intent/context of this specific content, and weighed potential risks - and wrote about this in the paper.

But the point here is, some people might find this method to still be problematic, and others will think my explanation overkill. Even I find this tricky and I sometimes doubt myself. I have different practices and standards now for my own research than I did five years ago.