In Defence Of Twitter

allen farrington
13 min readMar 16, 2019

Twitter allows anybody to speak, anybody to listen, and everybody to engage.

Everybody loves to hate on Twitter these days. Often on Twitter, but in more reputable spaces too. Within the past month, The Atlantic, The Wall Street Journal, and Slate have had less than glowing takes. My usual haunt, Quillette, published no less than three such. Recode was more positive, but the starting point is still the extent of Twitter’s maladies, from which it might recover. These articles follow in the wake of CEO Jack Dorsey doing a round of podcasts which seemed to achieve little other than turn these shows’ fans against the hosts for not being harsh enough on him. Joe Rogan’s first interview has one of his channel’s worst dislike ratios (possibly the worst), and the top comment on Dorsey’s appearance on Sam Harris’s show wonders whether Dorsey, “is on some sort of Internet tour ruining people’s ratings”. The fallout was so great that Dorsey went back on Rogan last week flanked by Vijaya Gadde, head of trust & safety, legal, and public policy at Twitter, with Tim Pool taking over from Rogan to argue the case on behalf of the aggrieved.

And yet I find that a great deal of criticism of Twitter somewhat misses the point and that these same critics, often for the very same reasons, ought to find room for a little praise. Many seem to see Twitter as a kind of Satanic influence on the culture. Perfect. Allow me to play devil’s advocate and present the defence.

I will be clear upfront on what I am not saying. I am not saying I approve unconditionally of how Twitter is run, or that it has never made mistakes. I am not saying that Twitter does not censor political outsiders, or that its moderation practices are not patently hypocritical. I do not intend this as a rebuttal to any of the articles mentioned above, all of which I enjoyed. I also do not intend to say Tim Pool did not give a great performance; he did, and I enjoyed this even more.

Here is what I will say: firstly, that the criticisms are odd in the first place given that Twitter is clearly a blessing for political outsiders, along with any and all cultural groups not traditionally aligned with legacy media. Secondly, that focusing on the outputs of Twitter that users most easily see, and that are as specific as alleged censorship, is the wrong approach and is highly likely to lead to even greater frustration when presented with Twitter’s or Dorsey’s explanations. And finally, that the blessing to certain political outsiders is only barely political in fact; it follows from Twitter more importantly being a blessing in principle as a uniquely democratising force in the spirit of the Internet itself.

There are surely countless examples of political outsiders aided by Twitter, and equally thousands of allegedly aggrieved groups, but one single example scores very near the top in both categories simultaneously: conservatives in particular would do well to remember that Trump was the first ever social media candidate for president. The Clinton campaign was based around the assumption of leveraging the influence of legacy media, while Trump outsourced the production and distribution of political material to a highly decentralised community that coordinated primarily via Twitter and Reddit. He could A/B test material in real time, and could tap into and amplify whatever rose to the top of his enormous network of meme production. Hence Pepe the Frog, which started as nothing but a joke, ended up being retweeted by Trump himself, and fawned over by legacy media.

Consider the visualisation below of the strength of engagement between and amongst the most influential Twitter accounts following the final presidential debate. The bigger the name the greater the influence and the thicker the line the greater the engagement:

graphic from John Swain’s Medium post

What we see here is that all the biggest names are on the Clinton side, but that there are vastly more connections across a greater range of sizes and more engagement on the Trump side. Dorsey made a related point on the Sam Harris podcast. In the lead up to the 2016 election, “the amount of journalists on the left, who were following folks on the right end of the spectrum, was very small, but the amount of journalists on the right, following left wing people, was extremely high

This all seems to demonstrate that Twitter was instrumental in allowing the Trump campaign, and the right by extension, to simply have a far better idea of what was actually happening during election season, by superior communication and information gathering, and to react to it appropriately. Marc Andreessen, professional baller, silenced the crowd during an onstage interview with Recode by pointing out that the only major media outlet that accurately reported on the election was Breitbart. Which is not to say Breitbart is not propagandistic — of course it is — but it was the only media outlet whose propaganda corresponded reasonably well to the facts, in this case. The Clinton campaign, on the other hand, relied on a paradigm of cultural power and information flow that had been effective for as long as anybody could remember, but which by 2016 had been dramatically disrupted by Twitter. Their total confusion led them to react highly inappropriately, including pricelessly deciding to lecture Americans on the dangers of racist frogs. Of course, there were many other factors that contributed to Trump’s victory, and exploring these has practically become its own cottage industry of social commentary. I do not wish to discuss any others, but simply to say that Twitter was an insufficient but necessary condition.

Since 2016, however, conservatives in particular have been increasingly irritated that Twitter seems to be censoring them, or at the very least moderating them disproportionately harshly and with what seems to them like obvious political bias. A version of my argument above seems to fit this narrative: it is now obvious that Twitter aided the Trump campaign, which was much to the horror of its (presumably?) majority progressive workforce. Therefore, they have since decided not to let the platform be ‘weaponised’ by the right any longer. This seemed to have been all but confirmed when Dorsey said on Harris’s podcast that, “I don’t believe that we can afford to take a neutral stance anymore.

It is certainly possible that at some level of human moderation there are people doing their job badly, without adequate transparency, or with politicised malice. But to read this into Dorsey’s quote, and to attribute such an ideology to the entirety of Twitter, I think is just silly. It assumes an institutional focus on Twitter’s outputs rather than its throughputs that I do not believe can possibly be the case.

Twitter is a platform, and this makes it fundamentally different from just about any other human creation or organisation, software or otherwise. Engineers do not design the outputs of Twitter; they design the architecture by which user inputs are turned into outputs on a massive and ongoing basis, in real time. In the first Rogan podcast, there were many times when Dorsey was alluding to precisely this fact not only as being true and relevant but as being his primary interest as Twitter CEO, and Gadde made the same points again the second time around. Yet if you think their jobs are to craft the outputs as optimally politically balanced, it would have sounded like both were weaselling out of addressing the topic altogether.

For example, Harris, Rogan, and Pool all pressed on what would lead to a user being removed from the platform, and Dorsey replied that Twitter looks at the pattern of conduct, returning to this slightly cryptic explanation several more times, as did Gadde in only slightly more helpful detail. What they mean by this is that Twitter’s content moderation is necessarily almost entirely automated. There is far too much for a collection of employees to even read, never mind effectively judge. In addition, natural language processing is nowhere near advanced enough to be the front line of this automation. In fact, it is debatable whether it can ever be, given the rawness of speech on Twitter, the enormity of assumed context on the part of speaker and audience, and speech’s ambiguity in general. Patterns of behaviour, on the other hand, are potentially very complex, but are capable of being definitively classified based on discrete data points. Twitter’s machine learning algorithms can and are being trained to recognise what is very likely to be harassment, or other breaches of the terms of service. And even this need not be perfect, but if it reduces the enormity of all tweets ever down to a set that human moderators can actually consider, it will be fairly effective. If Twitter can additionally allow users to contribute to the policing, this can provide further independent data points to feed to the algorithms. Of course, this introduces new problems in that this can clearly be abused, which starts the discussion all over on a smaller scale of how to embrace machine learning to crack down on this too, and so on and so forth.

What is probably unique to Twitter’s situation, and is remarkable in its own right, is that this is not even exclusively an exercise in software engineering. There are no quantifiably optimal solutions to these problems because the task itself requires sophisticated philosophical framing. Dorsey attempted to convey this at several points across his appearances, almost never to the deserved reception, since his respective interviewers were too keen to get out of philosophy and back to what happened on Twitter last week. For example, in his first appearance on Rogan, though clearly not a state actor, Dorsey addressed how best Twitter could respect the First Amendment rights of its users in the hypothetical ‘public square’ online. He noted, however, that there is not a perfect correspondence of regulating principles between actual public spaces and the online world because of the differences in capabilities granted to individuals by the technology. As Lawrence Lessig outlined in his brilliant Code 2.0, the transition from the application of laws in an original domain to a novel digital one often sheds light on a ‘latent ambiguity’ in the original law. In effect, Dorsey was advocating for what is called fidelity in translation as an interpretative principle in US Constitutional law — that we ought to tease out the motivating factor behind the law in the original context, such that this motivation can be faithfully reproduced in new circumstances. This is as opposed to, for example, an originalist approach that would rule latent ambiguities as ‘undebatable’ — meaning that the Constitution simply has nothing to say on such matters and the onus is on the legislature to make new law. Lest we worry about getting too far down this fascinating rabbit hole, Dorsey didn’t manage to either; Rogan roped him back into debating the moral limits of harassing feminists.

Dorsey only really comes into his own when freed to speak on his philosophy of technology without the weight of having to tie each and every thought to Twitter. For example, he is one of approximately two major tech CEOs to publicly back blockchain, the buzzword holy grail of empowering decentralisation and philosophy of technology, and to which he seems to crave pivoting since nobody knows quite how that can be reconciled with social media. And unlike Elon Musk (no stranger to courting controversy on Rogan, I might add), Dorsey is in a position to contribute to driving adoption of the technology. Given Dorsey’s responsibility to shareholders I am sure he is not allowed to say any such thing in public, but I am nonetheless absolutely convinced that if blockchain had existed at the founding of Twitter, Dorsey would have built as much of Twitter as possible around integrating this technology both throughout and specifically lower down in the software stack. Given the following train of thought of his on Rogan, this ought not to be surprising:

I was formed through a lot of the ideas of the Internet. I fell in love with what it made possible and never want to run afoul of those ideals and the removal of barriers and boundaries and the connection we have because of it. I reflect often on my role and the centralisation of my role and our company. I want to figure out how we can continue to add massive value and be an amazing business which is us and will always be us, but at the same time be a participatory force in this greater good that the Internet has started. It’s not lead by any individual or company and that’s the beauty of it.”

Again, if all we want is explanations for apparent speech policing, this will sound like vacuous techno-utopianism and will likely be extremely frustrating. But if we are assessing the attitude of the chief engineer of a platform, this is almost exactly what we want to hear.

What Dorsey and Gadde consistently tried to draw attention to was that Twitter wants to leverage as many as possible of the features of controlling a massive digital network to cleverly engineer around the quagmire of having to judge all content on the basis of language and context. Ideally this would also be done as efficiently as possible given the real-world constraints of limited computational resources and the time of human moderators. This is an extremely complex and interesting engineering problem that really has nothing to do with the content itself. The most this argument could be strained to invoke a bias against any particular content is that the engineering of the process itself has been ineffective such as to predispose the combination of algorithmic and human moderation to focus disproportionately on some kind of content. I don’t actually believe this argument, but wouldn’t it fit nearly perfectly with the rest of the conservative culture war line that their political opponents are far more likely to abuse whatever tools Twitter gives its users to help police content? Why isn’t this a conservative meme — and why do they let progressives off the hook and go after Twitter instead, which, under this theory, would merely be accidentally enabling this behaviour? It could well be that moderators don’t think they are being harsh in their assessments, and perhaps they objectively aren’t, but the fact of the disproportionality they see to begin with could be biasing the outcome regardless. All that said, even if the moderators are at that point acting maliciously for political reasons this would still a small part of a complex problem, the solution to which is better philosophy and better engineering, not better politics.

I have faith that this engineering will happen in time, without the need for further politicised fallout, precisely because I have faith in the philosophy that drives the engineering to begin with. I’m sure many readers do not share this faith, but I find it in two places. I find it firstly in interpreting my discussion of politics in 2016 above through a lens that is not itself political. While ‘the right’ won a superficially political victory, it was really a better riding of a monumental technological and cultural wave of change. Next time they might lose. If Cruz had run against Sanders they probably would have, and in fact Sanders and now Alexandria Ocasio-Cortez and Andrew Yang have proven especially adept at using Twitter. This wave was one towards decentralisation, distributing power and influence away from stale, centralised institutions. This is exactly what has allowed the likes of Sanders, Ocasio-Cortez, and Yang to emerge as forces of legitimate challenge to the mainstream of the Democratic Party, as Trump was to the Republican Party, and as none would have come anywhere close to being even five years prior. Twitter allows anybody to speak, anybody to listen, and everybody to engage. It short-circuits the centuries old legacy media infrastructure that mediated access to practically all impactful debate. It is in theory the essence of the freeing spirit of the Internet. The extent to which it has failed to live up to this in actuality seems to be so miniscule on balance that giving up on it would surely be a gross overreaction.

I also find faith in Dorsey himself. I feel sorry for him that his attempts to explain the philosophical quandaries implicit in maintaining the largest public forum in history largely continually fall on deaf and angry ears. I feel sorry for him that this is likely a microcosm of the task of running Twitter itself in that its success as an engineering project has given it a power that is probably corrupting to some. But I still think the burden of proof should be on the critics that such corruption, even though it may exist, cannot be overcome. Political leanings surely corrupt even the best of us, and although I’m sure Dorsey is not immune in this respect, I have faith in him regardless because I don’t think he considers them to be anywhere near as important as his philosophy of technology and his unique position to engineer it into reality. The most explicitly political I have heard him at any point was the following during his first appearance on Rogan:

I think the Internet allows for a very healthy scepticism of nearly everything. I’m from Missouri, it’s a ‘show-me’ state. We are all sceptics. My mom was a Democrat, my dad was a Republican. My dad listened to Rush Limbaugh and Hannity all the time. I found myself somewhere in the middle but one of the things I appreciated was that we had a ton of fights and arguments and yelling matches around the kitchen table but I appreciated the fact that we could have them. I felt safe to do so. I felt like — obviously they were my parents — but they weren’t judging me because of what I said.

That does not sound to me like the thoughts of a wilful censor. Nor somebody desirous to direct a public conversation, nor to be easily politically corrupted towards any end. I could well be wrong — perhaps the devil’s advocate has been tricked by his own client. But, until further evidence comes before the court of public opinion, Twitter is something I am happy to defend.

follow me on Twitter @allenf32



allen farrington

I’m an investor. I think about things. I write some of it down.