Sunday 20 November 2022

Follow the yellow brick road, part 2: How should decisions about new features be made in the fediverse (and based on what data?)

woke up to this exchange on Twitter’s use of user data (and user demands) to develop new features, and it pushed so many of my buttons that it immediately made me cough up this giant hairball. With apologies from the chef.


“Hello I must be going”


“I cannot stay
I came to say
I must be going
I'm glad I came
But just the same
I must be going, la-la!”


The original thread was from a ex-Twitter designer who, on his way out to start a Mastodon sabbatical, gave sage advice to Mastodon co-builders about “listening to its users” when deciding to adopt new features (or not) going forward. Most importantly, he seemed to suggest that a rethink regarding three specific features that currently make Mastodon different from Twitter might become necessary - namely its reverse-chron timeline, the absence of quote tweets, and the absence of algorithmic curation of feeds, i.e suggesting and hiding content from users where the algorithm has decided, based on their previous behaviour, that they are particularly/not interested in it.


Now, I have seen discussions on all of these features in my short time since I joined Mastodon and overall, I love that they are being discussed. I love it because the different perspectives that people bring to the table on each of them allow me to look “under the hood” of each of those features and understand why some people or communities might see them essential while others are vehemently opposed. What I appreciate the most is that those discussions are taking place at all, out in the open, where even non-techies like me can start to understand the basics and where many different communities can and do provide input from their own lived experience.


I’ve been on Twitter for 13 years and I’ve never seen anything like this.


I expect to change my mind several times yet on the desirability of many of the features under discussions, including QTs, CWs etc. I am less certain that my views will change on the algorithmic curation of my feed. I was literally one of the people mentioned in the original post that hated this feature on sight and constantly complained about it and who made it responsible for much of what Twitter has become. I would be desperately sad to see this introduced on Mastodon.


But my view on individual Mastodon features is not why I spent my morning writing this blog post. I’m writing it to pick up on two related things that came out of the original thread and the QT/reply.


Give users what they want


The original poster (an ex-Twitter developer, remember) justified Twitter’s decisions to introduce all of the above features by saying that it was “what 97% of users wanted and thanked us for”.


Taken apart, this statement has three effects, of course. 


1. It justifies design features that were, in practice, introduced by fiat of a centralised authority without much public consultation on the basis that they were really introduced as a response to some sort of majority-based (if unverifiable) verdict. It thus frees Twitter from responsibility for the subsequent negative impact of those features, which arguably contributed to the easier siloing of people into information ghettos, which in turn made them easier targets for disinformation and manipulation of all kinds. With some consequence for democratic discourse.


2. It pitches people against each other by painting the (alleged) 3% percent of Twitter users, who (allegedly) had concerns about these features, as both elitists, who think they are somehow better than the “unenlightened public”, and as dinosaurs, who will ultimately have to bow to the wishes of the majority, if they want their social media space to be successful. Now where have I seen this strategy before?


3. It obscures Twitter’s own motivations for introducing those features, namely fostering an “attention economy” and increasing user engagement, which in turn renders a bigger treasure trove of behavioural data for monetisation and commercial exploitation. Coincidence, surely. Oh look, a squirrel!



Now, I said this before and I am fully aware of how many people will think me elitist for saying so, but what users want is not always what is in users’ best interest.


As a “user” in one sense of the word, many people desperately want to be dependent on A-class drugs, cigarettes, alcohol or sugary snacks. They want to vote for Brexit to “have control” in the face of likely economic carnage, they want to control immigration to protect themselves even if that means asylum seekers in dinghies drowning in the Channel,  and they want the right to refuse to take reasonable precautions in the middle of a pandemic with predictable effects, particularly for the clinically vulnerable, because “free’em”.

Even assuming that giving users what they want was Twitter’s main motivation for introducing these features (narrator: It wasn’t), using “what users want” as your sole or even predominant yardstick for new development is a terrible strategy that opens the door to a majority dictatorship. 

We know all this because we’ve had these discussion before, in the context of our discussions on philosophy, democracy, ethics and human rights. And because we’ve had those discussions, as a society, we look at the consequences that the things “users want” will likely have – for the users themselves and for others – and then we put in place some rules around this (or at least, we used to).

When “what users want” is either detrimental to their interests or the interests of individual “others” (children, partners, etc.) or the community, we impose restrictions. We do this through the various constraints on behaviour at our disposal, law being one of them, and we are mostly very aware that what individual or groups of users want should not be the sole factor we, as a society, take into account in our decision-making. 


So to argue that certain design features on a Social Media site will quite naturally manifest unless elitist designers override “what users want” seems to me quite close to the thinking that was behind the “people are sick of experts” remark so beloved of the UK rightwing. And we all know where that got us,


Personally, as I already said above, I feel that there is a lot of good to be had from the open discussions of the various features and their effects on different communities, because it opens all of us up to different perspectives. What users want should certainly play a part in future development. But so should their motivation for wanting it and the (possibly unintended) negative consequences for heeding them. And I hope that when the decisions are made to adopt or not adopt some of those features, they are made as a result of these – messy, I know – discussions. Not because some central authority decrees it, yes, but also not simply because a majority of users, who may not yet be aware of the longterm implications, “want it”.


If I only had a brain…


Which beings me to the second point the kicked me off this morning, and that is the attitude implicit in the reply (or QT) to the original post.


The poster of that reply expresses a desire to learn from the original thread because the poster was a “Twitter designer” who shared his advice based on a "soak in Twitter's data." There is such longing (for that data) in that remark, it is hard to dismiss it, if you are, like me, a researcher yourself.


And surely, the second poster then goes on to bemoan the fact that outside researchers almost never got access to Twitter’s treasure trove of data because Twitter saw it as proprietary and actively obstructed, for example, independent research on harassment. The poster calls this “a waste”.


And I sympathise, I really do. I have friends in academia, who engage in empirical research and who would often sell their first-born for a chance of doing some work with that kind of data. And not just the raw data Twitter collected, but also – as the second poster highlighted – the results of Twitter’s own  “many thousands of A/B tests”. Releasing those results publicly would, in his view, “ be a valuable project”.

And this “desire for data” is a desire that has already been voiced on my feed with regard to the Mastodon data soon to be generated. How do we secure access to that data for researchers?”, I hear people asking. And I understand the excitement, truly, I do, but I have a question. 

AITA when I say that constantly having my behaviour analysed as part of A/B testing by Twitter without my consent “for research” made me feel like a lab animal and is one of the reasons why I started self-censoring a lot on that platform? Or am I right for taking both an ethics and a lawyers’ approach to this and argue that we must:


·       Challenge (and challenge again after discussions with marginalised communities) our own understanding of what the scope of, and the conditions for, access to public data for researchers should be, and

·       Resist the temptation to use in our own research “fruit from the poisoned tree” (i.e. the results of Twitter’s A/B tests).


I’m very aware that these are questions that many people with more expertise in both empirical research and ethics have discussed for a long and that I’m making my own observations as something of a dilettante savoir. And I get that empirical researchers get excited about the many opportunities for enquiry that these data treasure troves could provide. 

But I can’t help thinking that just asking for the right to do in your project what Twitter itself did behind closed doors is not the “white knight” argument people think it is.  Personally, that is certainly one of the reasons why I have always had such an aversion to projects that were uncritically pushed by “open data” and “data science” prophets, at least where those projects included the use of the personal data of real people.

Don’t pay any attention to the man behind the curtain

What I cherish most about Mastodon so far, is that we as users are intimately involved in the discussion of which features Mastodon should employ. And yes, that is probably down to the structure and the fact that these discussions are happening in smaller communities. This may not scale as more and more people join, and I don’t have a solution to that if I don’t want to be one of those late-arrivals, who wants to pull up the drawbridge behind herself.


But rather than putting my energy into figuring out how we can make Mastodon more like Twitter or pushing to re-create its features because they were what “people wanted”, I’d like to spend that energy on (first) interrogating those features, ALL the motivations for their adoption and also, and most importantly, the research community’s own approach to finding out “what people want” or what is good for them in the first place.


Because “people” is all of us on there, so a more empathic, iterative, discursive, ethnographic, “call and response” approach that involves us not just as objects to be studied but as co-builders and decision-makers may be what we need to develop. 


I don’t know where those on the good ship Mastodon will have sailed it in a few weeks’, months’ or even years’ time. But wherever we may go, I hope we go there intentionally lest we replicate not just all of Twitter’s mistakes but also its method for making them.

[Caption: “Too much of a good thing can be wonderful”]


No comments:

Post a Comment