Twitter generated a fair bit of conversation this week with the dual events of its founder’s public concession that “we didn’t fully predict or understand the real-world negative consequences” of how people might misuse its platform and the release of a new initiative to engage the broader research community to help it devise new metrics for assessing the “conversational health" of Twitter to help it improve how society uses it. Could this new effort transform Twitter into a Utopian public square or does it represent just another failure of Silicon Valley to understand the world it seeks to reshape in its image?
Like all soaring dreams when they meet the crush of reality, the early dream of the Internet has given way to the darkness of the human condition. Social media, in particular, has helped centralize and expand the reach of some of the web’s darkest horrors, from terrorist recruitment to horrific hate speech and threats of violence. The algorithmic surfacing and short message format of social media, coupled with its largely mobile consumption, favors the kind of brief emotional, entertaining and highly shareable monologues that bring out the schoolyard bully in even the most rational of beings.
In the US we spend at least a decade of our lives in school being refined from the raw clay of the worst elementary school playground behavior, where the biggest, loudest and most aggressive one wins and physicality reigns supreme, to the refined work of art that appeals to our intellectual and rational being. Social media works to undo all of that, offering an algorithmically mediated world in which bullies are able to run free and the ones with the loudest voices and least concern for others can terrorize their fellow netizens, shouting, screaming, streaming profanities and epithets and threatening all around them until the rational voices depart and the only ones left are their fellow bullies.
Meanwhile, the unseen hand of the social media platforms themselves is silently shaping that discourse, from who is allowed to speak to what they are permitted to say. Even routine decisions can have powerful consequences when they are magnified by nearly two billion people. Whether by policy or in error, even simple accept/reject decisions for advertisements can powerfully reinforce harmful, false and outdated narratives. This often creates a strange duality where the same companies touting equality and diversity freely allow their paying customers, the advertisers, to discriminate against minorities, market to bigots, and run ads of half-naked men, but ban a fully clothed women reading in the dark. Of course, the platforms aren’t in any way doing this knowingly and fiercely condemn these horrific misuses of their platforms and work to halt them when uncovered, but at the same time, it raises the question of just how closely they are willing to pay attention to what happens on their platforms if their automated and human filtering miss such blindingly obvious cases and just how much non-obvious activity is occurring every day.
From standing up to Congress for the free speech rights of terrorists to reversing course and dedicating substantial resources to forcing them off their platforms, the central role social media companies play in our modern world means they find themselves in the middle of nearly every societal debate. Even simple policy decisions can be exploited by powerful and skilled adversaries, as repressive governments find ways to turn every policy into a loophole that can be used to quash dissent or turn them into the ultimate surveillance tools. After all, even a seemingly unquestionable policy of banning “terroristic speech” could quickly become a powerful tool of censorship in the hands of a dictator who simply passes a new law that classifies all anti-government criticism and opposition politicians as “terrorists,” leaving the social media companies to unwittingly and unwillingly act as defacto government censors.
Into this dark and foreboding world comes Twitter’s proposed new floodlight of civility, the idea that we can simply instrument our online world with a set of metrics that assess its “health,” create a dashboard of our online “conversational health” and then drive the platform towards the green zone on the dials.
Perhaps most centrally, Twitter’s new effort both presumes that conversational health can be quantitatively assessed and reduced to a series of indicators displayed on a dashboard and, most importantly, that there is a natural path forward to optimize the platform towards the positive side of each of those indicators. Unfortunately for Twitter, this oversimplification misunderstands the immense complexity that is our globalized world and assumes a single unified understanding of “health” that is shared across the world’s immensely different cultures. It is doubtful, for example, that a vibrant discussion of democratic rights would be understood to be “healthy” by the world’s non-democratic nations, while a discussion of the merits of one religion might deeply and profoundly offend members of another and so on. Regardless of the civility and thoughtfulness with which such topics are discussed, they are likely to create great distress among some portion of Twitter’s population.
Even assuming that distress is a natural part of a “healthy” platform and that the focus of these metrics would be on the way in which we conduct that discourse, the most important question is just what to do when those indicators trend towards the red portion of the dial. Twitter cites shared attention as one possible metric, yet filter bubbles in the online world have been a long standing topic of conversation with little movement towards their elimination, while agenda setting and its related threads has been a focus of media scholarship for the better part of a century. If pushing the world towards greater conversational overlap becomes a focus, how precisely do we do that? When it comes to news media, we not only don't have an answer, but we embrace elements, such as the inwardly looking focus on "local" as a positive.
One might argue its important to come up with metrics first and worry about how to optimize for them later, but often the most natural metrics are those hardest to optimize for. While countless organizations and governments have worked to create dashboards of “global peace” over time, we are in 2018 living in a world in which conflict still scars our planet. Having a nice visual dashboard gauge of “peace” does little to help you assess how to get there.
The critical missing dimension from Twitter’s call is the question of “Modifiability.” A proposed metric of “Happiness” that measures how “happy” Twitter’s 300+ million users are might suggest itself as a natural metric under the idea that happier users might be less likely to engage in toxic conversation, but it is not immediately clear how to improve that number by making the world’s population happier. In contrast, a measure of “Profanity Prevalence” that assesses the density of profanity on the platform in the idea that it can split users apart is readily optimized through textual and visual OCR filtering of the platform.
Twitter’s initiative appears to emphasize the academic community as the source of answers by asking for lists of “relevant peer-reviewed publications and papers” and expecting as final output “peer-reviewed, publicly available, open-access research articles,” rather than broadening proposal evaluation materials to tools, datasets, blog posts, visualizations and other content related to the call and emphasizing non-academic outputs beyond simple open source tools. Twitter’s profound restrictions over access to its data mean that few academics have extensive experience working with the Twitter firehose or understanding its nuances at scale, whereas explicitly encouraging the commercial research world and the legions of spare-time open data and open research contributors that hail from those worlds, on condition that they make all of their outputs freely accessible, might offer unique perspectives not present in the academic world.
It is also intriguing that Twitter’s call for proposals makes little mention of international and cultural differences across the world in that different cultures across the world may have different understandings of what “healthy” conversation is. Instead, Twitter’s examples propose a set of “universal” metrics that can be applied equally to the entire planet, rather than an understanding of just how richly varied the people that inhabit are world are and the benefits of that diversity and plethora of narratives and views, rather than conforming the planet to a single “Twitter-fied” view of what conversation should be.
Of course, perhaps the most important question is just how far Twitter follows through with this new initiative and how much it changes between its current sweeping vision and the reality of what it ultimately becomes, much as what ultimately became the Twitter Data Grants program was greatly narrowed from the grand vision of some early proposals.
Does this new initiative show that Twitter has finally recognized that it alone may not have all of the answers to how to fix its platform? Or is it merely a stalling tactic to show to governments and populations increasingly concerned about its role to buy more time by saying it is actively working on the problem? Perhaps tellingly, when asked why Twitter was only just now launching this initiative, a spokesperson declined to comment beyond pointing to its founder's previous comments.
Putting this all together, it remains to be seen how serious Twitter is about its new initiative and whether its present grand vision survives the reality of implementation. The focus on “universal” metrics that hold across the whole of humanity ignores the immense diversity of the world’s cultures and push against questions of just how well the richness of human nature can be reduced to spreadsheets of cold quantitative numbers. Perhaps most importantly, however, the success or failure of Twitter’s new initiative hinges on what to do with that ultimate “health” dashboard – as those dials start trending to the red, how readily can they be nudged back into the green? The answers to these questions will determine whether we will ultimately have a powerful tool for fixing Twitter or will we merely have a pretty window through which to watch the decline of online civility and perhaps ultimately the collapse of the world itself?
