Twitter is taking unnecessary risk with its current way of applying "state-affiliated media" labels

Twitter, states & labels

On Thursday August 6, 2020 Twitter’s Head of Integrity @yoyoel announced that Twitter would start labeling “state-affiliated media outlets and government accounts.”

The label is some sort of a symbol, followed by text “[Country] state-affiliated media.” This is what the label looks like:

Which accounts will be labeled with this? According to Twitter,

[a]ccounts belonging to state-affiliated media entities, their editors-in-chief, and/or their senior staff 

So not just Russian or Chinese government media’s accounts, but the editors and senior staff at such media as well. (Twitter does not list any impacted countries although in its example of what the label looks like, the example is an unspecified “Russia state-affiliated media” ).

(Facebook seems to have a similar policy since June 2020 but I don’t know much about Facebook so I won’t comment theirs here; to the extent that they are similar, my analysis will apply to Facebook’s policy as well).

Problems with the “state-affiliated media” policy

There are three main problems with the policy’s “state-affiliated media” part (that is the only part I will talk about here).

Motivation and objectives

First, the motivation and the objectives for the policy are unclear. Why did Twitter do this? What does Twitter want to achieve with this? Will it achieve it? (Hint: no).

Twitter’s news release states that

Twitter provides an unmatched way to connect with, and directly speak to public officials and representatives. This direct line of communication with leaders and officials has helped to democratize political discourse and increase transparency and accountability. We also took steps to protect that discourse because we believe political reach should be earned not bought. In 2019, we banned all state-backed media advertising and political advertising from Twitter. Today we’re expanding the types of political accounts we label. 

So Twitter motivates the new policy by merely stating that it expands its existing policy from 2019 that bans “all state-backed media advertising and political advertising.“ This policy was motivated by Twitter’s belief that

“there is a difference between engaging in conversation with accounts you choose to follow and the content you see from advertisers in your Twitter experience which may be from accounts you're not currently following. “

All this apparently means that Twitter does not want its users to mistake paid content for unpaid content. That’s fine, and this is journalism ethics 101, “news and advertising are strictly separate.” (This 2019 policy puts Twitter close to being a journalistic organization but that’s another story.)

Nonetheless, it is unclear to me how Twitter’s 2019 motivation - users should be able to differentiate unpaid from paid content - extends to the 2020 “state-affiliated media accounts“ labeling policy. Twitter seems to think that it is important that its users can differentiate between state-affiliated media and non-state-affiliated media (“independent media” in Twitter-speak but I will call it “private media”). But why does Twitter think this? Maybe because it would help its users “make informed decisions” about things, as per Twitter’s mission? Fair enough. But how does a user’s ability to differentiate between two types of media - state-affiliated and private - help Twitter achieve that mission? Does Twitter consider state-affiliated media to be less trustworthy than private media? Or the other way around? This is unclear, and both positions, especially outside the U.S., have serious proponents with serious arguments. Is Twitter aware of this?

Poor definitions: what do these terms even mean?

The second problem, related to the first, is that the policy simply cannot work in any meaningful objective and systematic way because the definition of state-affiliated media is poorly developed:

How do we define state-affiliated media accounts for these new labels? 

State-affiliated media is defined as outlets where the state exercises control over editorial content through financial resources, direct or indirect political pressures, and/or control over production and distribution. Unlike independent media, state-affiliated media frequently use their news coverage as a means to advance a political agenda. We believe that people have the right to know when a media account is affiliated directly or indirectly with a state actor. State-financed media organizations with editorial independence, like the BBC in the UK or NPR in the US for example, will not be labeled. As part of the development of this process, we consulted with a number of expert groups, including members of the Digital and Human Rights Advisory group in Twitter’s Trust & Safety Council.

How is Twitter supposed to operationalize this definition of “state-affiliated media?” Without a solid operationalization, it will be impossible to apply and enforce the policy in any meaningful way.

The “financial resources” part

The first part - “the state exercises control over editorial content through financial resources” - sounds possible to operationalize: determine whether a government gives money to a media outlet in such a way that the money impacts the editorial content of the media (i.e. what the media covers and how it covers it.) There are two parts here, (1) is money given by government or not, and (2) if money is given, does it impact editorial content. Twitter might be able to investigate whether money is given, although I am not sure how they will investigate this in cases where government gives money to private media by buying ads or giving tax breaks or radio waves or something else of monetary value. I think such investigations are, in theory, possible and if Twitter comes up with some thresholds as to how much monetary value given to a private media by a government is sufficient for the“financial resources” criterion to be met, then let’s assume Twitter can determine whether government gives money to a media.

But how is Twitter supposed to determine that money, when given, impacts what a media covers and how it covers it (this is the “editorial control” criterion of Twitter’s definition)? If the U.S. government gives money to San Diego Union-Tribune by buying an ad for the U.S. Navy (a made-up example, but a realistic one), how is Twitter supposed to determine whether and how this money given impacted how the paper chose to cover a Navy related incident? Establishing such links has been a big topic in media studies for decades, and as far as I know it is an unresolved one. Does Twitter think they can resolve it? I think there are advances to be made in this line of research with big data analysis so Twitter might be able to do something there, but I’m not sure this is what they are intending to do, and even with data however big, it will still be difficult to establish such links with any reasonable certainty. Which brings me to my next point, which is what is the threshold for such links to be considered established? What if a Navy ad in a paper tipped an editor’s scales once and they chose not to run a story on Navy but ran an otherwise equally good story of similar length that one time in 2007, does that mean the the “the state exercises control over editorial content through financial resources“ criterion is now met and the paper and its editors and senior staff now get the “state-affiliated media” label? Most of us, myself included, would perhaps say that ‘nah’ that’s not enough - what what would be, and why? And so on.

In sum, the “financial resources” part of the definition needs to be thought out much more carefully and in such a way that it can be meaningfully operationalized, and I don’t think Twitter can do it. Again, measuring the specific impact of a media organization’s financial resources in editorial decisions is a big question in media studies that to the best of my knowledge remains unresolved. Or who knows, maybe Twitter can do it, and I’ll be happy to be proven wrong.

The “political pressures” part

The second part - “direct or indirect political pressures” - faces the same operationalization issue than the first one, but on a whole new level. How is Twitter supposed to establish that “political pressures” - direct or indirect, whatever this distinction means - exist in a media organization? Again, as implied by Twitter’s definition, if “political pressures” do exist, then a media is “state controlled” and shall be labeled as such. Don’t political pressures exist in any organization, everywhere, all the time, including in families? Well, you could say that Twitter doesn’t mean that kind of political pressures. OK. What do they mean, then? This should defined by Twitter. Before such defining can be done, however, Twitter should have a good theory of “politics” (currently it does not, and it seems it does not even realize that it should have one, and this lack of understanding - or maybe it is meta-lack of understanding? - generates most if not all of Twitter’s problems in its policies that deal with politics).

Once they have a good theory of politics - assuming they can arrive at one, and I think they can if they hire some political theorists and place at least one of them in the top management team - then and only then will they be in the position where they can then start operationalizing the concept of “political pressure.” They should be able to do this, but then they run into similar operationalization issues as they did with “financial resources” - establishing that “political pressures” exist and that they impact editorial decisions over some subjective but justified impact threshold. ()

The “production and distribution” part

The third part - “control over production and distribution” - suffers from similar lack-of-definition issues as the previous two parts: these concept are not defined. What does “control” mean here? What does it not mean? Same with “production” and “distribution.” Because these concepts are not defined, they are not operationable in any meaningful way (is there currently a team at Twitter trying to do these operationalizations-for-labeling without any conceptual definitions? As far as I know, labels are being applied as of yesterday, in consultation with Twitter’s Digital and Human Rights advisory group). Defining and operationalizing these concepts in general is certainly doable, but it does not mean that it is doable in any meaningful way for the particular purpose of Twitter here. Which reminds me that the particular purpose of this policy of Twitter’s here is unclear (so start there, @jack). As things currently stand, including the lack of relevant expertise in the top management, I don’t think Twitter can do any of this definitional work in any meaningful sense.

So Twitter is unable to meaningfully define “state-affiliated media“ (the sentence in the statement saying that ‘[u]nlike independent media, state-affiliated media frequently use their news coverage as a means to advance a political agenda‘ does not do Twitter anything useful for the purposes of defining anything).

Twitter and the state

Also, Twitter says that

We believe that people have the right to know when a media account is affiliated directly or indirectly with a state actor.

But why? Why do people need to know this? What are the they supposed to do with that information? Twitter doesn’t say, but the unstated assumption here seems to be that users should view information by state-affiliated media differently from information by private media, perhaps because one of them is less trustworthy than the other. But which one? Twitter doesn’t say, but the new policy seems to imply that it considers state-affiliated media less reliable than private media. But this is a U.S. centric view. Did Twitter realize that about half of Europe would have a beef with this assumption: every single European country has a state-affiliated media (that is why you have the Eurovision song contest, which apparently is coming to America next year), and people tend to trust these media. How did Twitter miss this fact? It seems that they may not have fully missed it, as the policy excludes some state-affiliated media:

State-financed media organizations with editorial independence, like the BBC in the UK or NPR in the US for example, will not be labeled.

So state financing is actually OK, as long as the editorial control is independent of the state “like the BBC in the UK or NPR in the US.” (Note: control of production and distribution with editorial independence is not discussed here for no apparently good reason). The underlying theory - it’s editorial independence that matters in determining the trustworthiness of a media - makes a lot of sense, but how is Twitter supposed to know that editorial control in a given state-funded media like BBC is independent of the state? Maybe they just read the BBC’s Charter, which states that “[t]he BBC should provide duly accurate and impartial news, current affairs and factual programming.” And after reading it they just believed it, which would be an easy thing to do, because, you know, everyone knows that that’s what BBC does. And maybe all the NGOs in the Digital and Human Rights advisory group then nodded their heads and said ‘Yep’ and off they all went.

But then again, RT, a media with a newly acquired “Russian state-affiliated media” label, says on its About us page that “RT is an autonomous, non-profit organization that is publicly financed from the budget of the Russian Federation.“ Did Twitter just look at this and said ‘Nah, I don’t think so’ and apply the label? Or did they even look at the page? Not that it would have made a difference, I think, and looking at the charters or “About us” pages is not the right way to do this. But how did they do it? How does Twitter determine the “editorial independence” of a state-financed media? This is a key question that Twitter leaves unaddressed. And if the policy is to be meaningful for any other purpose than @jack avoiding interrogation from the U.S. Congress, this is a question that Twitter must address seriously.

So what?

But so what? Who cares if Twitter can’t define the concepts? After all, they are a private company and many of their product decisions (this policy is a product decision) are at their discretion and rightly so. With their “Digital and Human Rights advisory group” under the Trust and Safety Council, why can’t Twitter just start applying the labels to Russian and Chinese state media, with no regard to the idea of proper definitions?

Well, that seems to be exactly what they are doing:

Again, whether applying the labels without solid definitions and theories behind the definitions makes sense depends on the purpose of the policy; if its real purpose is to demonstrate to the U.S. Congress that Twitter is doing something to counter the online fake news, the policy might succeed. I would not be surprised if the Congress were OK with Twitter labeling Russian and Chinese media with the label and leaving U.S. and European media out. After all, everybody knows who the bad guys are.

The problem is that Twitter can become viewed as a hypocrite. Predictably, Russians are at it already:

In addition to the Russian state-affiliated media accounts not liking their new designation, some individual Europeans, presumably representing the portion of the European population that views state-affiliated media as less trustworthy than private media, a view presumably in line with that of Twitter’s, seem to be on Twitter demanding that their country’s state-affiliated media be labeled as such.

Not just Europeans:

Why it matters

Is Twitter now going to go and investigate all of the state-affiliated media and determine with some certainty which of these need the new label? I don’t think so, not least because, as I have argued above, such an investigation is not doable based on the current definition given by Twitter. What, then? Just use the “BBC argument” and say that these are excluded because, well, we just think there is editorial independence because, well, we say so? They can certainly do that, and maybe all of the critique (including mine) is just Twitter noise by a small number of activists and some states that Twitter does not need to care about. Maybe all of Twitter’s inconsistency will lead to nothing substantial. Maybe Twitter users will keep using Twitter like before, no dent to the brand or user experience, and the advertisers will keep advertising like before.

But because of the new policy Twitter is now fighting the Russian government, which will do everything it can to let the world know that Twitter is a hypocrite and a mouthpiece of the U.S. government. Maybe the Chinese will join the fight. Maybe the Saudis, too, and who knows who else. And the other state-backed internet trolls.

So Twittter, a Californian company, picks a fight against the governments of countries with a combined population of 1.5 billion or more: is this a risk worth taking for Twitter? If so, what is the upside? I just don’t see it. In addition, and not necessarily unrelated to the state fight risk, Twitter risks alienating its users, which can happen if the users begin, with or without help from state actors, to view it as a hypocrite, or, yikes, evil.

As a background to its new policy, Twitter says that

Our mission is to serve the public conversation and an important part of that work is providing people with context so they can make informed decisions about what they see and how they engage on Twitter.

There is nothing wrong with this. But nothing in the new policy is likely to advance this mission.

Organizational and managerial implications

Twitter may be updating the policy at some point, maybe soon. But to really remedy the situation, Twitter should go back to the drawing board and develop actual carefully motivated and defined objectives for the policy: what exactly is it that they want to achieve and why? Twitter’s mission - “to give everyone the power to create and share ideas and information instantly without barriers“ - is good, why not start there?

Once the objectives are in place, Twitter should define the concepts in the policy so that they can be meaningfully applied to the empirical analysis of media accounts and the organizations behind them - a key requirement for any meaningful application and enforcement of the policy.

One possible way to properly define the key concept of “state-affiliated media” would be to use a deliberative body for the job. A body that would use a set of rules as a starting point but be free to exercise judgment based on available facts. It looks like Twitter did use such an approach in 2019 with regards to its state-affiliated media advertising policy:

How will we define state-controlled media in this context?
This policy will apply to news media entities that are either financially or editorially controlled by the state. It has been informed by established academic and civil society leaders in this space. Sources include Reporters Without Borders Press Freedom Index, Freedom House, the Economist Intelligence Unit Democracy Index, the European Journalism Centre’s Media Landscapes Report, the Committee to Protect Journalists, and UNESCO's framework to assess media development and independence.

We will be making policy determinations on the basis of critical issues of media freedom and independence, such as control of editorial content, financial ownership, influence or interference over broadcasters, editors, and journalists, direct and indirect exertion of political pressure, and/or control over the production and distribution process.

What happened with this approach? Twitter doesn’t say. In any case, going this route of this may require establishing a permanent “editorial body” within Twitter that has the relevant expertise to organize deliberative decision-making. The National Science Foundation’s proposal review process with permanent program officers and outside temporary reviewers could serve as a model for this. There are other models, but establishing such a body would probably necessitate some additional re-organizing within the company. It seems that Twitter did use some version of outside temporary reviews model the “state-affiliated media” policy:

As part of the development of this process, we consulted with a number of expert groups, including members of the Digital and Human Rights Advisory group in Twitter’s Trust & Safety Council.

Based on the observable output (non-implementable state-affiliated media policy) this does not seem to be working. More generally, it does not seem to be a useful way to organize Twitter’s policy-making regarding any policy that deals with media or politics. Neither the Trust and Safety Council nor the Digital and Human Rights Advisory group seem to have the required expertise, and Twitter may not even understand that this is the case. (Twitter’s top management team does not have a single person responsible for policy development and policy development only). Currently policy development seems to be under Twitter’s top lawyer Vijaya Gadde, who is a head of “legal, public policy, and trust and safety.“ It would make sense to split the position into three: (1) legal, (2) public policy, and (3) trust and safety. Each section should be headed by a person with relevant expertise, which means a political scientist for policy (or possibly a person with public policy experience, although it is important that this is not a lobbyist position), and possibly a cybersecurity person for trust and safety (depending on what the term means for Twitter).

Conclusions

Twitter’s new “state-affiliated media” label policy is a failure: it is inapplicable and unenforceable in any coherent, meaningful, objective and systematic way. This failure exposes Twitter to unnecessary geopolitical and business risk that Twitter seems to be unaware of. To remedy the situations, organizational and managerial changes are needed.