CDA 230 Problems: Do Algorithms Threaten to Undermine Speech Protections?

A few weeks ago, after I published a blog post raising the question of what might happen to CDA 230 when internet intermediaries like Facebook invoke First Amendment protections – which civil liberties lawyers’ were calling on Facebook to do in the wake of the controversy over its trending newsfeed – I was fortunate enough to have a sustained email exchange with UCLA Law Professor Eugene Volokh. There, he pushed back on my claim that there is a tension between the First Amendment and CDA 230, pointing to precedent about intermediaries – including bookstores (Smith v. California; Fort Wayne Books, Inc. v. Indiana), newspapers (New York Times Co. v. Sullivan), and movie theatres (Freedman v. Maryland; Vance v. Universal Amusement Co.) – that show an intermediary can claim both constitutional rights and statutory protections in the same piece of content (e.g. selling a book, running an ad, or screening a film). His reading of the law is correct, and can and indeed has been applied to the online context, including to website operators and bloggers. So within this framework – within the scope of what types of facts have been contemplated by precedent – his skepticism was valid.

Professor Volokh’s incisive questions pushed me to clarify the distinction between the world he described, where intermediaries could invoke both the First Amendment and CDA 230 (or similar statutory immunity) in the same piece of content, with the emerging world that I’m interested in, where precedent is limited when applied to new situations that are substantially different in kind such that analogic reasoning fails. From this perspective, we can see the impending clash between the First Amendment and CDA 230, and, as a result, can ask the normative question of when the First Amendment should – or should not – apply.

This blog post represents the arguments and thoughts developed through my email exchange with Professor Volokh, to whom I am indebted for his time and thoughtful questions. It proceeds in three parts.

First, I explain the functional transformation happening in online intermediaries that changes their relationship to third-party content, which results in an open legal question as to how and if both the constitutional rights and statutory protections apply.

Second, I discuss how this open legal question will result in a tension between the First Amendment and CDA 230, such that the former may cannibalize the latter. This outcome turns on two factors: technological, business, and legal dynamics pushing intermediaries to the boundaries of the functional spectrum so they act more like speakers and editors on the one hand, and courts that are primed to reject any expansion of CDA 230 on the other.

Third, I argue that, in this context where we’ve reached the limits of precedent – where it’s unclear how the First Amendment and CDA 230 relate to one another – ethical considerations and especially a concern with the public interest should drive how we think about how the law ought to apply. This methodology matters not only for tackling intermediary speech questions, but moreover offers an approach for academics thinking through questions of First Amendment coverage, or what Frederick Schauer describes as the right’s “scope of application.”

1.     Changing Functions of Intermediaries

As a preliminary matter, it’s helpful to understand how the function of intermediaries is changing such that it is no longer clear how they can claim both First Amendment rights and CDA 230 immunities in the same content, as they have been able to do. (The blogger who has a First Amendment right to host comments, but a CDA 230 immunity from the content of those comments is a quintessential example.) The key insight is that today there is a spectrum of functional roles for intermediaries, due largely to the affordances of digital media technologies, which enables them to play different roles than they could in the past. As a result, these functional changes distinguish emerging intermediaries from those contemplated by our First Amendment precedent.

The precedents that Professor Volokh cites about bookstores, movie theatres, and newspapers are useful to understand online intermediaries that perform similar functions to those brick-and-mortar intermediaries, like online newspapers and online bookstores. For First Amendment purposes, these intermediaries function like neutral hosts or conduits, making a basic editorial choice to either sell or not sell a book, run or not run an ad, or screen or not screen a film. (Of course, brick-and-mortar intermediaries could have transformed the content that they were selling, but they didn’t; it wasn’t common practice, for instance, for theatre owners to splice new images into a third party’s film.) But they can also operate non-neutrally; as Professor Volokh points out, newspapers are protected if they edit op-eds, excerpt letters to the editor, or impose editorial standards around advertisements. Further, intermediaries can seek First Amendment protection for exercising non-neutral editorial judgments, like choosing to exclude certain viewpoints with which they disagree (Miami Herald v. Tornillo; Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston).  

But the functional role that intermediaries can play online is qualitatively different. Intermediaries can be more than neutral hosts or ideological content curators. They can be editors and speakers with respect to third-party content, and can engage in transforming that content in qualitative ways on a scope and scale unparalleled in the analog world. Consider, for example, if an online content aggregator deployed an algorithm that edited headlines of third-party content to generate more clicks, so that sometimes it changed nothing, sometimes it changed a word or punctuation mark, and other times it changed the headline completely.

This hypo, which illustrates the functional spectrum available for online intermediaries, complicates the application of precedent that finds a constitutional right and statutory immunity for neutral hosts and non-neutral content curators in the same piece of third-party content to situations where intermediaries are more robust editors or speakers with respect to that content. In other words, when the intermediary takes on a more robust function – editorial and speaker functions that are more available today due to the affordances of digital media technologies – it is unclear if both First Amendment rights and CDA 230 protections attach. Specifically, the expansion of the functional spectrum available to online intermediaries with respect to how they interact with third-party content simultaneously bolsters their First Amendment claim and weakens their CDA 230 immunity argument. As a result, how the First Amendment, CDA 230, or both apply to online intermediaries playing new functional roles is an open legal question.

2.     Resulting Tension Between the First Amendment & CDA 230

This open legal question will only become more salient due to technological affordances, business imperatives, and legal ambiguity around the scope of CDA 230. And this change will take place in a context where courts have expressed their disdain for CDA 230, suggesting that the statutory claim will become vulnerable if courts are given an out via a constitutional argument. The result might be that the integrity of CDA 230 is undercut by First Amendment claims (a concern that prompted my earlier blog post).

Technological, business, and legal forces push intermediaries in new functional directions.

First, social media and technology companies likely will blur and push the boundaries of the ways that they transform, interact with, and curate third-party speech. Even if the Facebook curation example does not rise to this level – one could make an argument either way as to whether or not it’s functioning as an editor – one can imagine a variety of more aggressive actions that intermediaries could take (recall our example of an algorithm that substantively changes headlines of third-party content). The combination of algorithms, machine learning, and big data strongly suggest that companies will have the opportunity and the business incentive to transform third-party content in new ways (for instance, to transform headlines to get more clicks).

Second, these changes will take place in an ambiguous legal context. As EFF correctly explains in its legal guide for bloggers, “courts have not clarified the line between acceptable editing and the point at which you become the ‘information content provider.’” In other words, current law imagines either (1) an intermediary that doesn’t change content and so gets both CDA 230 and First Amendment protections (the blogger who hosts comments is the quintessential example) or (2) an intermediary that changes content so much that, while it retains First Amendment rights, it loses CDA 230 immunity because it acts like an information content provider.

What is interesting and new – and poses the risk to CDA 230 that motivated my earlier blog post – is a qualitatively different type of online property that lives on the boundaries of CDA 230 and is able to play a variety of functional roles. Because these online properties will be poised to claim First Amendment protections and CDA 230 immunity, the statutory claim will wade into the uncharted legal waters between “acceptable editing” and total transformation of third-party content. Regardless of whether or not the current Facebook trending news example rises to the level of an intermediary playing a new functional role as editor or speaker, what matters is that it’s a harbinger of things to come; social media and technology companies will innovate and push the boundaries of how First Amendment precedent has imagined engagement with third-party content, either as a neutral host or ideological curator. (Again, think of the permutations of the algorithmically edited headline hypo.)

Courts are primed to limit CDA 230’s application to intermediaries’ new functional roles.

These changes put courts in a position to decide between constitutional and statutory claims. Specifically, they will have to decide whether or not the new activity in which these online properties are engaged amounts to that of an information content provider, a question at the boundaries of CDA 230. And because of courts’ explicit disdain for the statute, litigants presenting functionally contradictory arguments – a constitutional interest as editor or speaker of a potentially new kind, and a statutory interest as a host or conduit – give courts the perfect opportunity to limit the statute: reject an expansion of CDA 230 and rely on the First Amendment claim.

Consider the opening lines of a First Circuit case, Doe v., decided earlier this year against the plaintiffs because of CDA 230: “This is a hard case—hard not in the sense that the legal issues defy resolution, but hard in the sense that the law requires that we, like the court below, deny relief to plaintiffs whose circumstances evoke outrage.” If intermediaries increasingly portray themselves as “creat[ing] or develop[ing] [] information provided through the Internet or any other interactive computer service” (i.e. as an “information content provider” per CDA 230(f)(3)) to claim First Amendment rights but simultaneously claim CDA 230 protections – in other words, if intermediaries perform contradictory functions for constitutional and statutory purposes – then courts might use the First Amendment as an out with which to reject CDA 230. The point is that the threat of First Amendment claims cannibalizing CDA 230 is not hypothetical; it’s quite real given where courts are likely going.

So for emerging legal claims to which the Facebook example points, courts may not find CDA 230 immunity exists at all. The result? At minimum, a hardening of the current boundary of the statute. And by not expanding CDA 230 to these new types of claims – which I think will be much more prevalent going forward than those by the blogger hosting comments or the online bookstore – then the overall scope of CDA 230’s application to online properties will be diminished. The point is that when these new types of online properties claim both First Amendment and CDA 230 protections for new types of activities, they may offer courts the opportunity to limit and effectively diminish CDA 230 by preferring the First Amendment claim. 

Is this inevitable? No. What’s important is that we think through these very real possibilities – given the changing nature of technology, legal ambiguity, corporate interests, and courts’ motivations – when we argue that potentially new types of activities should be covered by the First Amendment, as lawyers are doing in the Facebook trending newsfeed controversy.  

3.     Ethics Should Inform How to Act at the Boundaries of Precedent

When we reach the limits of precedent – here, of the dual application of First Amendment rights and CDA 230 immunity – academics and the public have an opportunity to make normative arguments about how the law ought to apply. To get a handle on the ethical questions at stake, let’s consider the consequences of the argument that intermediaries who push existing functional boundaries and perform contradictory roles for constitutional and statutory purposes – seeking to be speakers and not to be speakers, to be editors and not to be editors with respect to the same content – should be able to invoke both the First Amendment and CDA 230. This creates an ethical dilemma for the public. How can the public know what type of speaker they are dealing with if an individual or organization simultaneously claims two contradictory identities? More importantly, how can the public hold an individual or organization accountable if it’s unclear what role that individual or organization is playing in our democracy?

This question is most salient in the context of the press. Let’s consider the Facebook trending newsfeed example. If Facebook invokes editorial control in its trending newsfeed, then it understands its actions as similar to those of the press. While that is a perfectly coherent First Amendment position, it’s hotly contested. BuzzFeed’s Nabiha Syed and Ben Smith suggest that platforms are not speakers or journalists, explaining that while “[t]hey power social connection, free expression, and the distribution of news and entertainment on an unprecedented scale,” they “largely don’t create speech themselves,”  and The Association for Education in Journalism and Mass Communication explicitly argues that Facebook “should be judged as a social media platform, not as a news media company.” Even if Facebook positions itself as the press, that’s a role to which a host of responsibilities attach, a point that has been variously made during the recent Facebook brouhaha by danah boyd (calling for public accountability) and New York Times media columnist Jim Rutenberg (calling for more transparency).

By positioning itself as part of the press, and claiming the rights that attach, the public will have an expectation that Facebook, or at least the trending newsfeed part of it, will also take up the attendant responsibilities of the press: that it adheres to certain ethical standards and can be held accountable as such. (Of course, this is an ideal to which the press strives and often fails to achieve, but strives nonetheless.) But if Facebook can in its next breath argue that it’s not a speaker or editor in the very same content for which it just claimed First Amendment protections, all bets are off: that move totally upends the public’s ability to know what role Facebook is playing and how to hold it accountable, and breaks the rights-responsibility nexus that the public expects in its Fourth Estate.

Of course, the question is much bigger than Facebook. What’s at stake is the legal and social identity of new types of actors – hybrid content curators and content creators – the public’s ability to understanding those actors’ identity, and the public’s ability to then hold them accountable. This normative inquiry should animate not only questions about online intermediaries and how constitutional rights and statutory protections apply to them, but also should inform the broader academic question of First Amendment coverage. When we evaluate the application of speech and press rights to new actors and new situations, we should not be guided merely by analogic thinking from precedent, but moreover by ethical considerations animated by the public interest.       


Thanks also to CIS Director of Intermediary Liability Daphne Keller and rising Harvard Law School 3L and Chair of the Harvard Law Review Forum Alicia Solow-Niederman for their insightful feedback and helpful comments as I thought through these questions. 

Add new comment