Thursday, September 10, 2020

Can Free Speech Rights be Extended to Platforms?

Perhaps ironically, political protests by professional athletes might be opening a new legal avenue by those who seek to have fairness doctrines, equal time rules or other public interest requirements on social media platforms. Up tot his point, the U.S. constitution has protected speakers only from censorship by government entities.


But widespread private firm allowance of such political speech might widen the terrain for those who believe some new expansion of speech freedoms must be extended to powerful private platforms as well.


“Speakers in the United States have few or no legal rights when platforms take down their posts,” according to Daphne Keller, director of the Program on Platform Regulation at Stanford's Cyber Policy Center. 


Some use the analogy of must carry rules once imposed on TV broadcasters. To date, lawsuits likening platforms to “public forums” have failed. 


Also, there are different issues related to content: removal of items that violate terms of service, and the way that ranking systems operate. The former deals with removed content; the latter deals with search ranking algorithms. 


The former issue is similar to the ways stories are constructed by news media, for example. Are opposing views treated fairly and with neutral adjectives? Is the amount of space given to opposing views roughly equal? 


The latter is similar to the choice of stories to run, and not the way content is treated once a “publish” decision is made. Which stories are deemed newsworthy, and which are not?


So far, U.S. courts have held that private platforms do not have a legal obligation to carry user speech. Still, some argue that dominant platforms are de facto gatekeepers, and should be regulated as “essential providers” of political speech, or even utilities, with a common carriage obligation. 


But those claims of speaker rights also bump up against the First Amendment rights of the platforms as speakers. Ranking and removal of content is an exercise of editorial judgment, in other words. 


Largely unexamined--so far--are various methods of giving more control to platform users, says Keller. It is not easy, but some advocate more end user content control settings. The problem is that people disagree about what constitutes “hateful speech.”


Some may  want platforms to carry all legal speech. Others might simply prefer more curation, allowing civil dialogue. 


“One possible approach would let platforms act against highly offensive or dangerous content

but require them to tolerate more civil or broadly socially acceptable speech,” argues Keller. 


Again, the problem is disagreement about how to identify such offensive or dangerous content, and not simply because the censoring algorithm or reviewer simply disagrees with the expression of those views. The same sort of problems arise with efforts to apply “fairness doctrines” that essentially preserve the rights of the listener, rather than the speaker. And all such rules limit free speech rights of speakers and platforms. 


Another approach distinguishes between “hosted” content (allowing anyone to speak) and “recommended” content that appears in news feeds, for example. The former is more akin to a town square, the latter more akin to the “curated” news feeds or search results. 


Yet others might prefer some form of unbundling the ranking and sorting algorithms, allowing third parties to create their own curated feeds. None of these would be simple. None would be free of some limitations on free speech. And most could negatively affect the monetization models that make the platform services possible. 


And yet we might be moving in such directions in any case. The recent issue of political protests by professional athletes raises the issue of whether constitutional free speech rights actually have standing even in the case of private firms. 


No comments:

Costs of Creating Machine Learning Models is Up Sharply

With the caveat that we must be careful about making linear extrapolations into the future, training costs of state-of-the-art AI models hav...