My Senate Testimony About Platform Transparency

This week I participated in an unusually collegial and productive Senate hearing about approaches to platform transparency, presided over by Senator Coons. My detailed written testimony, including appendices listing other resources and attempting to identify platforms potentially covered by proposed laws, is here. It captures a lot of detailed questions and concerns that I have been thinking about for a while, but not had time to write about anywhere else, including about surveillance issues. It also includes concerns addressed in more detail in two earlier blog posts (on privacy and on costs/benefits) and in this list of practical hitches in even the most basic transparency reporting.

Testimony from the other witnesses -- Brandon Silverman, Nate Persily, Jim Harper, and Jonathan Haidt -- is here. 

My oral testimony is below. I ad libbed some, so this isn’t verbatim.

***

Thank you for the opportunity to testify. I work on platform regulation. Transparency is incredibly important to the future of platform regulation. I’m particularly honored to speak beside Nate Persily and Brandon Silverman, who have built real-world, functioning, effective transparency systems. I’m excited to talk about this at a time when the EU, in its major new Digital Services Act legislation, has really paved the way for a new era of platform transparency.  

I think we will hear today about a real diversity of transparency tools, and a whole range of tools is what we need. Not every approach is useful for every research topic, or for the way every platform functions. As someone who has worked on multiple approaches to platform transparency for well over a decade, including in-house at Google, I would like to find a way for the law to enable a whole range of approaches.

• That includes fixing the laws that constrain so-called “scraping” of data from public websites. This is low-hanging fruit, and would have the side effect of lowering some barriers to interoperability and competition.
• It includes building APIs, which are just channels for computers to talk to each other, so people researching things like bias in algorithms can submit bulk queries and look at bulk results.
• And it includes mandatory disclosures like those contemplated in laws like the discussion draft of PATA. [Like *some* of those in PATA. I maybe should have emphasized that more, as my written testimony did.]

​All of this can vastly improve public information both about online harms, and about what regulatory responses will actually make things better. 

To make these laws work, though, there are some pitfalls we need to navigate. Some are practical, some are constitutional, some are political. My written testimony on this was… very long. Sorry! It raised some questions that should go to agencies for resolution. Getting transparency rules right will be picky, time-consuming, in-the-weeds, and iterative. Things will keep changing, and the law should not fall behind. But there are also some real policy decisions here. I think making those is your job.

One is about privacy and surveillance. For researchers to examine what is being said and claimed and propagated online, and how platforms influence that, they will need to look at information about people. There are unavoidable tradeoffs in deciding how much they get to do that. Sometimes I think the need for public information should win out, other times user privacy should. Congress should provide guidance on that – ideally through federal privacy legislation, but if not, then right here in these bills.

When it comes to government surveillance, I think there is a brighter line. Nothing about these transparency laws should change Americans’ protections under the Fourth Amendment, or under laws like the Stored Communications Act. I don’t think that’s anyone’s intention here. But clear drafting is essential to make sure the government can’t effectively bypass the Fourth Amendment by harnessing the unprecedented surveillance powers of private platforms. 

A second big issue is about competition, and the practical costs and benefits of using different transparency tools for different companies. Laws designed for a Google or a Facebook are a bad fit for companies that are both far smaller in measures like revenue and employee and user count, and also far less relevant for the problems that laws like this are trying to solve. I don’t think we need a live dashboard showing us which hotels are most popular on TripAdvisor. I think the problems of designing rules for giant incumbents and then applying them to a whole competitive ecosystem of smaller companies speak for themselves.[1]

A third issue is about CDA 230. As I explained in my testimony, I don’t think tying transparency to liability for content serves anyone’s goals. Instead of responding to any particular concern or critique of CDA 230, it would effectively create a lottery, where inevitable errors in meeting a complex new set of requirements will unpredictably open the door to litigation in some cases but not others.

The last thing is the First Amendment. I want transparency mandates to be constitutional. But there are serious challenges. Y’all need to put some really good lawyers on that. [I cut this for lack of time, but wanted to say: Also… Maybe if you just apply these requirements to the giants, who need to do a lot of this in EU anyway… they might not be willing to bring those First Amendment claims. It would make them look pretty bad.]

So those are my concerns. In platform-speak, they are the things to solve before launch, so that transparency laws do what we want them to do and do not lead to unintended consequences. The other thing I wrote about, and am happy to answer questions about, is the substantial new transparency regime in the DSA.

Transparency laws are important. We should do this. We should do it right.