My latest review for Jotwell is live. In it, I salute Felix Wu’s Collateral Censorship and the Limits of Intermediary Immunity, which is arguably the best paper ever written on Section 230, the provision that shields the Facebooks of the world from liability when users defame each other. Wu gives a clever new interpretation, one that explains not just what Section 230 is intended to do, but what it isn’t. Here are some excerpts:
Wu’s move, so elegant that it is obvious in hindsight, is to recognize that there are really two questions about Section 230, not one. The first is how strong its protection should be: this was the issue in Zeran and it is the one on which scholars have mostly divided. The second is when that protection should apply at all: this part has received less attention. If we have two sliders to play with, perhaps we should set them differently. Section 230 could be broad and shallow: shielding intermediaries in a variety of factual settings but offering only a thin immunity that can be overcome with a sufficient showing of malice or unconcern on the intermediary’s part. Or it could be narrow and deep: protecting intermediaries only from defamation and closely related torts, but offering absolute protection when it does.
Having distinguished these sliders, Wu offers guidance on how to set them. He does so by reconstructing a theory of what Section 230 is supposed to do: prevent “collateral censorship.” It’s a commonplace that an online intermediary can’t be counted to stick up for its users when its own ass is on the line. (Exhibit A: PayPal and Amazon disgracefully dropped WikiLeaks based on little more than Joe Lieberman’s disgraceful jawboning.) Faced with even the vague and distant threat of liability for user speech, the rational intermediary will yank the challenged content. It has nothing to gain and everything to lose by doing anything else. This gives opponents of speech an easy-to-use heckler’s veto: just threaten the intermediary. A robust, deep immunity recognizes that the intermediary has much weaker incentives than the original poster does.
As Wu demonstrates, however, this rationale only works some of the time. …