Sunday, April 20, 2008

Online advertising conference: publisher perspectives

The Law & Business of Online Advertising: Berkeley/Santa Clara

Panel on Publisher/Intermediary Perspectives on Online Advertising

Moderator: Paul Schwartz, BCLT

Oren Bracha, University of Texas School of Law

Bracha discussed his paper with Frank Pasquale, Federal Search Commission? Search engines are advertisers in a broad sense, an important link in conveying information to the public. They’re also more and more entangled with ads in the classic sense—Google is now using snail mail to solicit new AdWords clients. What happens when Google drops your site from #3 to #50, collapsing your traffic? Google may have a legitimate reason, but right now it’s a black box. Does the First Amendment prohibit any regulation?

The internet is not a soap box for everyone, as early plaudits said, because to exist is to be indexed by a search engine and vice versa. Market/technology won’t take care of the problem of hidden bias. With high fixed and licensing costs, entry’s not that easy. And users are unlikely to understand what they’re not getting. Subtle distortions are not easily evaluated; search engines can shape preferences. Eric Goldman argues that personalized search can fix this. Personalized search might solve the structural bias towards commercial sites and majority preferences, but makes it harder to deal with targeted bias towards users/classes of users—personalized search makes each manipulation more valuable.

Peter Swire, Moritz College of Law, The Ohio State University

Two topics: (1) a technical fix for opt-out cookies. A brief summary of his comments to the FTC on technical issues in behavioral advertising. New consumer poll: 63% of consumers rate opt-out’s importance as 10 out of 10, and 84% rate it 8 or higher. The industry also has consensus on consumer choice as an important value. The opt-out cookie is broken, though. Antispyware software deletes all cookies, including opt-out cookies, or browser software deletes it, and tracking is turned on again.


There are simple technical fixes available. There are problems with doing opt-out through cookies, but it’s technically feasible and scalable now. If we don’t do it, the industry and FTC will fail at enabling customer choice.

(2) Privacy and antitrust

Privacy is a nonprice metric against which to measure the competitive effects of mergers. Privacy can be a form of nonprice competition, and all 5 FTC commissioners recognized this point in Doubleclick/Google merger review. Consumers might be harmed if surveillence increased due to less competition. There is competition in privacy now—e.g., AskEraser.

Privacy concerns could also justify putting conditions on a merger. Or it could be part of a second request by the FTC for information on practices in data-rich industries. This points to a new role for Chief Privacy Officers in mergers and other transactions.

Limits: this approach doesn’t treat privacy as a fundamental right or investigate the quality of privacy practices independent of mergers.

Rebecca Tushnet, Georgetown University Law Center

I began with a point Peter Swire made when we were discussing the panel: New, big institutions invite a lot of different regulatory responses. Radio & TV got the FCC; no single regulatory entity has yet emerged in online advertising, but there are lots of issues.

I focused on intermediary liability. Internet intermediaries tend to make 2 claims: (1) we have a First Amendment right to do what we want. For example, comments to the FTC claim a First Amendment right to target users in behavioral advertising, because it’s a publishing decision. (2) it’s not our speech, so don’t hold us liable if it’s unlawful. With search engines, we see this particularly in the trademark keyword sales debates.

However, this has all proceeded with little discussion of First Amendment constraints on intermediary liability. The policy space seems fairly open, and we have different regimes. (1) DMCA for copyright, with notice and takedown in response to allegations of links to infringing materials. An injunction remains possible, but it’s not worth it; the takedown does everything the copyright claimant wants. (2) CDA immunity from being treated as a publisher (or distributor). Eric Goldman has noted that the CDA was enacted at the height of internet exceptionalism. (3) Other IP, including TM, where no special rule applies and courts are still working out how to apply prior precedent on direct and secondary liability.

A couple of points: (1) Intermediaries are not great representatives for speaker interests. The mismatch between intermediary and user interests with respect to the DMCA is well-known. Mark Lemley also makes the point that Google—no slouch with respect to asserting user rights in general—bans the use of competitors’ trademarks in the text of ads, even ads that are explicitly comparative and thus obviously legitimate under US TM law. Google doesn’t want the bother or risk of screening, so advertisers aren’t allowed to do it.

(2) There is substantial pressure as the internet matures to move away from CDA immunity, especially for business torts. Roommates.com is an example, where the violation is inducing illegal acts by suggesting terms; so is the AdultFriendfinder case, involving false advertising/endorsement; and we may see a similar result in the Subway/Quiznos case, where Quiznos encouraged people to create comparative ads and some of the ads were allegedly negative enough to constitute actionable product disparagement.

We haven’t spent much time asking what the constitutional boundaries of intermediary liability are, because the CDA preempted so much of that. Here’s one story: NYT v. Sullivan, the foundational case for the modern First Amendment, is a case about advertising and intermediaries. The NYT ran an ad about segregationist misbehavior containing minor factual errors. Alabama commissioner Sullivan sued and got a huge libel judgment against the New York paper. The Supreme Court held that liability could not constitutionally be imposed for statements about a public figure without actual malice—knowledge or reason to know of the falsity.

The Court’s opinion was explicitly designed to enable business models that involve transmitting others’ speech. An intermediary is less able to verify the truth of speech given to it than it is to verify its own speech; less committed to the initial speaker’s cause then the initial speaker, and thus less likely to underestimate its chance of being sued or overestimate the extent to which the rest of the world will agree with the justness of its cause; and more likely to have litigation-attracting deep pockets. For all these reasons, it’s easier to chill, and Sullivan is designed to protect the intermediary business model.

This view of Sullivan might give us a way to think about how to regulate intermediaries and possibly revisit the CDA. The NYT can prescreen for content in the print edition; internet intermediaries have more difficulty doing so. Sullivan may be too restrictive a standard to apply online. Even notice of claimed unlawfulness might not be enough to justify a takedown requirement—much less damages.

Q for Bracha: How does the elasticity of demand between organic and paid search results affect your analysis? Both are sources of traffic, and organic is free in theory but you might be paying a SEO for it.

Bracha: He’s much less worried about the paid results. Usually paid ads just trigger TM complaints.

I said that the big lie of the keyword debate is that it’s just about paid search. When you sue websites for TM use in keywords, you affect, or desire to affect, the organic results. The TM use debate is about controlling the whole page.

Swire: The good news is that the organic/paid line is brightly delineated. In other markets, “advertorials” and other practices sneak in ads deceptively.

Varian: Nonprice competition is relevant to antitrust, but another element is product quality/relevance. If there’s a tradeoff, that should also be a concern; why stop at privacy?

Swire: True. Mergers are touted as efficient for consumers. To the extent relevance is part of a merger it should be netted against the disadvantages, including privacy.

Varian: If Google changes its algorithm, there are still just as many results in the top 10 as there were before. Helping some people hurts others.

Bracha: There’s no metaphysically neutral search result. Bias has to be judged normatively. Right now Google’s a black box no matter what. In many situations, Google should be able to choose, but perhaps not all.

Q: By merging, even companies with lousy privacy policies may decrease potential future privacy competition.

Swire: Then we take that into account. But if merger won’t change privacy policies, it’s hard to say that the merger causes privacy harm. Showing a causal role in harm is a high burden for regulators.

My question: Why wouldn’t bad privacy practices necessarily get worse if they joined forces? That is, if Amazon’s privacy policies are bad and suddenly it knows what (bad) eBay knows, isn’t there a synergistic harm?

Swire: The extra harm is not an antitrust problem—it’s just increased harm to consumers, just like increased pollution from a merger.

(Ok, but then I don’t get why we count improved relevance, as Varian said, as a benefit of a merger.)

Varian: Don’t do things like environmental regulation through competition policy: regulate environmental impact directly.

Swire: The counterargument: if you have a harm sitting in front of you, one might ask, why not weigh it? The FCC standard is the public interest and convenience, though that’s not the FTC standard.

Comment: Privacy as a feature is the province of antitrust. Privacy as a good is the province of consumer law.

No comments: