An American served >1 year in prison for conduct that is 100% legal in Europe. But it’s not drugs. It’s copyright. Here’s why it matters.

The Administration is receiving some unusual advice from the content industries as it undertakes to renegotiate the North American Free Trade Agreement (NAFTA). It’s not surprising that content providers would weigh in, given the multiple obligations Article 17 of NAFTA imposes on Canada, Mexico, and the United States concerning intellectual property protections. If it’s your ox being gored, you’re going to have some strong opinions; it would be weird if you didn’t.

What makes the content industries’ position unusual is that they are apparently attempting to take some issues off the table entirely as subjects for negotiation. According to this recent TechDirt article (emphasis mine):

the entertainment industries are arguing that exceptions and limitations are outdated and unnecessary in trade agreements. They say that copyright holders should be protected from piracy and unlawful use of their works, claiming that any exceptions and limitations are a barrier to the protection of American artists.

That is, the content industries want the final agreement to mandate strong copyright protections without mandating (or, depending on how you read the word “any” in the preceding quotation, permitting) signatory countries to recognize exceptions or limitations on those rights. Copyright rights, on this view, are a one-way ratchet: they can only be strengthened everywhere, never reduced.

There are a lot of reasons to quarrel with this view, assuming it has been accurately reported. It’s profoundly ahistorical; every international agreement that defines copyright rights also fixes outer boundaries of protection. Article 10(1) of the Berne Convention requires every country to allow quotations from published works, and other parts of Berne allow copyright exceptions for teaching, news reporting, and other socially valuable purposes. The WTO TRIPS Agreement, another landmark multilateral treaty, incorporates long-recognized copyright limitations (for example, Article 9(2)’s idea/expression dichotomy) into international law. An argument that copyright limitations and exceptions have no place in trade agreements isn’t an argument against NAFTA; it’s an argument against every copyright treaty ever concluded.

An argument for international copyright law to ignore copyright limitations and exceptions is also bad policy, which is why civil society and public-interest groups have been loudly insisting that NAFTA renegotiations protect fair use and maintain balance between the rights of content creators and users. It’s why the American Library Association, the Center for Democracy and Technology, and over 80 other individuals and organizations from several nations issued the Washington Principles on Copyright Balance in Trade Agreements last fall as NAFTA renegotiation got underway in earnest.

But most of all, it’s just shortsighted even from the perspective of people who want to maximize the scope of copyright rights to take limitations and exceptions off the table as a subject of trade negotiations. To understand why, let’s look at two recent cases involving an American citizen and a European company who did the same thing: they distributed a tool that allowed owners of Nintendo Wii entertainment systems to play games that had not been authorized by Nintendo. These systems, colloquially known as “mod chips,” bypassed the internal authentication system that the Wii used to verify that the games users were seeking to play on their consoles had been approved by Nintendo. Nintendo complained to the authorities in each jurisdiction that helping Wii owners play unauthorized games was against the law because it circumvented a technological protection that Nintendo had put in place to ensure that only authorized games were playable.

Both the United States and Europe have laws against that sort of thing. They have laws of that type because they are required, under Article 11 of the WIPO Copyright Treaty, to

provide adequate legal protection and effective legal remedies against the circumvention of effective technological measures … that restrict acts, in respect of their works, which are not authorized by the authors concerned or permitted by law.

The EU Court of Justice, however, refused to condemn the making of a device to circumvent the Wii’s technological protections, because those protections appeared to the court to be broader than necessary to protect Nintendo’s copyright rights. It required, in assessing the legality of the circumvention tool, a consideration of “how often [the] devices are in fact used in order to allow unauthorised copies … and how often that equipment is used for purposes which do not infringe copyright[.]”

In contrast, the American who circumvented the Wii’s technological protection went to prison for twelve months and one day; the opinion of the Court of Appeals did not even address the question whether the “mod chips” he sold could have been put to noninfringing uses.

These two cases involved the same underlying offense and essentially identical substantive laws, written to implement a single treaty obligation, yet they reached opposite outcomes for reasons neither opinion said a word about.

If you believe in meaningful limitations to copyright holders’ exclusive rights (or, for that matter, in the rule of lenity which usually governs complex criminal issues of first impression), you probably prefer the European outcome to the American one: perhaps tools should be evaluated based on whether they are mostly put to infringing or noninfringing use. Maybe the limitation on copyright holders’ exclusive rights that the European court recognized would be a good rule for the United States, too.

On the other hand, if you believe that copyright holders should have the power to control how their works (such as Nintendo’s game console) are used, then the European approach probably infuriates you and the American one probably seems correct. Maybe you wish that European law didn’t recognize a limitation on copyright holder’s exclusive rights in this instance, and that the American view prevailed instead.

The point is, whichever outcome you favor, the only way to get to that outcome is for international treaty negotiators to discuss limitations and exceptions. Taking limitations and exceptions off the table makes it impossible to move the ball in any direction, not just the one the content industries claim to oppose. That’s a foolish way to conduct negotiations, and one all the players in the NAFTA debate should reject.

Is Copyright’s Muddled “Conceptual Separability” Doctrine Coming to Patent Law?

In Samsung Electronics Co. v. Apple Inc., 137 S. Ct. 429 (2016), the Supreme Court held that damages for infringement of a design patent could properly be awarded even even where the patent covered only an individual component of the product as sold to to consumers rather than the design of the product as a whole. Most products aren’t sold to consumers in separate pieces, however, but only as a completed whole. When only some of the components of a finished product are covered by a patent, how should a court assess the damages for infringement of that patent? The Court’s decision naturally raised, but did not answer, the question how to apportion damages where a particular component of a product as sold to end-users infringes a patent but the remaining components (and the product as a whole) do not.

In at least two post-Samsung cases, courts have adopted a four-factor framework to help analyze the apportionment question. Drawn from a test originally proposed to the Supreme Court by the Solicitor General, these courts have described the relevant considerations as:

  1. “[T]he scope of the design claimed in the plaintiff’s patent, including the drawing and written description”;
  2. “[T]he relative prominence of the design within the product as a whole”;
  3. “[W]hether the design is conceptually distinct from the product as a whole”; and
  4. “[T]he physical relationship between the patented design and the rest of the product,” including whether “the design pertains to a component that a user or seller can physically separate from the product as a whole,” and whether “the design is embodied in a component that is manufactured separately from the rest of the product, or if the component can be sold separately.”

See, e.g., Apple Inc. v. Samsung Elecs. Co. Ltd. (N.D. Cal. Oct. 22, 2017); Nordock, Inc. v. Systems, Inc. (E.D. Wis. Nov. 21, 2017).

The third, and to some extent the fourth, factors in the courts’ framework look a very similar to a question that comes up frequently in copyright cases concerning the design of manufactured goods: whether the appearance of a useful article is “conceptually separable” from the utilitarian aspects of the good. “Conceptual separability” is shorthand for language in the Copyright Act that declares that “the design of a useful article … shall be considered a pictorial, graphic, or sculptural work only if, and only to the extent that, such design incorporates pictorial, graphic, or sculptural features that can be identified separately from, and are capable of existing independently of, the utilitarian aspects of the article.” 17 U.S.C. § 101 (emphasis added).

But “conceptual separability” doctrine in copyright is a bit of a mess right now, and it’s not clear whether it contains any useful points of analogy for courts in patent cases. In its decision in Varsity Brands, Inc. v. Star Athletica LLC, 799 F.3d 468, 484-85 (6th Cir. 2015), the Court of Appeals identified nine different tests the courts had developed to ascertain whether the design of a useful article contained any “conceptually separable” copyrightable elements. Rather than trim down the number of competing approaches, the Court of Appeals proceeded to broaden it by developing its own novel five-step approach to assessing conceptual separability, id. at 487-89. And last year the Supreme Court, affirming, enunciated an unhelpful two-part test that it claimed was drawn directly from the above-quoted statutory language—the same language that generations of prior case law had found inadequate to explain which features of a useful article’s design could be protected under copyright.

It’s not clear to me that the courts who are now insisting on an assessment of “whether the design [of a patented article] is conceptually distinct from the product as a whole” recognize that they are treading adjacent to a doctrinal minefield. Given the courts’ great difficulties answering copyright’s “conceptual separability” question, what reason is there to expect clearer or more satisfying answers in patent cases?

Fake News and the First Amendment

Mark Verstraete and I are at a great symposium on Fake News and the First Amendment, run by UNC-Chapel Hill’s First Amendment Law Review. We will have a draft of our article available soon – for now, here’s a link to our white paper on fake news. Many thanks to Mary-Rose Papandrea, David Ardia, and Lindsie Trego for having us!

Come See The Slants at Arizona Law!

The Slants – fresh off their win at the Supreme Court in Matal v. Tam – are coming to Arizona Law! You’re invited both to Simon Tam’s talk about his experience litigating over whether the band’s name could be registered as a mark, and also to their concert. Details below.

When: Tuesday, 12 September 2017

12PM: Talk by Simon Tam, College of Law, Room 168

5PM: Presentation to the BA in Law Program, Fred Fox School of Music, Room 146

7PM: Concert, College of Law (1201 E. Speedway – or, corner of Mountain Avenue & Helen Street)

Who: The Slants

For more information, see this article on Arizona Law’s Web site.

The Untestable Marketplace of Ideas

This post will summarize my essay, The Empirical First Amendment (forthcoming in the Ohio State Law Journal, as part of a symposium on “the Expanding First Amendment”). The essay argues that the First Amendment’s protections are surprisingly shallow when it comes to scientific discovery.

I say “surprisingly” because free speech law aspires to be scientific. The Founders were influenced by the ideas coming out of the scientific revolution. In the modern era, the Supreme Court has consistently placed its bets on the “marketplace of ideas” theory that truth will emerge from competition between factual claims, much like scientific progress itself.

Fortunately, the scientific approach to the creation of knowledge is extraordinarily good, at least compared to its alternatives. As Jonathan Rauch neatly explains, it requires adherence to just two simple rules: First, no claim, no matter how absurd, unconventional, or unpopular, can be removed from consideration. That is, there is no final say, by the government or any other institution, about what is and is not true. I will call this “Claim Liberty” for shorthand. This is the familiar freedom to speak. Second, every person should be able to test the claim for himself to ensure that he gets the same results. That is, nobody should have to rely on the personal authority of the speaker making a claim if they would prefer to verify its accuracy themselves. I refer to this as “Empirical Liberty,” but it can also be understood as a freedom to test. Empirical liberty often (though not always) requires people to have the means to collect information, observe evidence, or experience things directly from the world.

First Amendment law gives full and unabashed support to Claim Liberty. Even verifiable lies and hateful claims are protected from official suppression in order to avoid the problems that arise if the state has final authority over what is and is not true. The First Amendment “presupposes that there are no orthodoxies—religious, political, economic, or scientific—which are immune from debate and dispute.”

But the First Amendment does not and cannot give that type of support to Empirical Liberty. To see why it does not, consider the claim that the iPhone is water-proof. Nearly every part of an empirical test, from the purchase of the phone, its transport, and its submersion into water involves non-communicative conduct that can be regulated on the basis of its non-communicative effects. To see why the First Amendment cannot give full support to Empirical Liberty, one needs only a brief tour through the broad range of descriptive claims that are theoretically testable but not practically so. Claims that a particular food or exercise reduces the risks of heart failure, or that a particular tax structure creates more jobs, are empirical claims, but a strong right to test the claims with the best possible experiment would enable the tester to impose serious interventions on other peoples’ lives.

Thus, the First Amendment’s support for the scientific process of truth-seeking is incomplete. It permits every claim to be submitted to the “marketplace of ideas,” (Claim Liberty), allowing a cacophony of proposed hypotheses, many of them wrong, many provably wrong. But First Amendment case law and theory has done much less to give listeners the means to rationally choose between those competing theories.

This is not entirely surprising because Empirical Liberty is a modern fixture in the philosophy of science. During the time of the Founding Fathers, a commitment to Claim Liberty was revolutionary enough by taking control over the claims that can and cannot be put forward out of the hands of religious or government decision-makers. It left the decision about which claim is acceptable, and why, up to the individual listeners.

The scientific community did not fully flesh out and appreciate the importance of empirical methods until the early twentieth century, when Karl Popper formalized the requirement that a claim must be testable and falsifiable in order to be valid. This requirement advances human knowledge by assuring all participants that they can see for themselves whether a claim gives an accurate account of the world, and therefore helps kill off the theories that do not match the evidence.

The value of Popper’s contribution was also illustrated in real time by Albert Einstein’s Theory of Special Relativity. The theory caused an uproar. The scientific community mobilized to publish a collection of essays titled One Hundred Authors Against EinsteinThough it may be apocryphal, I would like to believe that Einstein really did have a doozy of a retort: “If I were wrong, one would have been enough.”  If his theory were wrong, any one person could go collect astronomical data that would disprove all or part of the theory.

Popper helped elucidate the importance of Empirical Liberty to scientists, but jurists and free speech theorists have not caught up. Without Empirical Liberty, people do not have the means to gather data and test competing propositions for themselves. A First Amendment that promotes Claim Liberty without protecting Empirical Liberty may not facilitate knowledge very well if the laws leave few opportunities for testing. Instead, it trades one monolithic, governmental or religious authority over claims for millions of individual king-popes who must resort to unscientific hunches or trusting faith in others since they cannot empirically test any of the competing claims.

I am not the first to observe that the First Amendment is an imperfect vehicle for generating factual truth, particularly since its design permits false claims to fester. Robert Post, Vincent Blasi, and Fred Schauer have made similar critiques. But the scholars who have tackled the epistemic problems in First Amendment theory have advanced corrections that abandon Empirical Liberty (and even Claim Liberty in some contexts) and encourage reliance on experts in order to avoid the pernicious effects of false claims.

These are serious concessions to the ideal of building a First Amendment in the image of the scientific method. But they are also so practical that they seem to cause no major loss. After all, the best scientists rely on expert opinions in fields outside their own. Even if we had access to very good evidence or had unlimited opportunities to run experiments, no single person could test and verify every plausible claim (let alone every possible claim) that interests him or has a material effect on his life. It will, therefore, always be necessary to rely on other people’s expert opinions in order to rule out inaccurate claims. For the foreseeable future, the law will play some role in that process in some contexts where health or safety are at stake.

However, without sufficient attention on the protection of Empirical Liberties, the legal rules that optimize collective knowledge for the short term, under assumptions of static or slow-moving access to empirical evidence, will not be optimal in the long term. Over time, they may require a person to remain reliant on the purported expertise of others even if there is no impediment to empirical inquiry. If the state directly restricts a person’s ability to test and verify competing claims, or if the state indirectly and unnecessarily restricts his ability to do so, then any claim that the law must treat listeners as dependents on (state-selected) experts rings hollow. More importantly, it raises legitimate questions about what value strong Claim Liberty can have if our Empirical Liberty is anemic by design.

Can the First Amendment protect the people’s right to test proposed claims for themselves, either through experimentation or through access to relevant information? The courts have not explored the logical implications or necessary limitations of Empirical Liberty. But if we are committed to a scientific approach for First Amendment theory, and I hope we are, then Empirical Liberty must receive more consideration and protection.

Protection of Empirical Liberty does not need to threaten the distinction between speech and conduct. There are at least two areas of low-hanging fruit where Empirical Liberty can be protected without risking serious detriment to the government’s ability to regulate more generally.

First, the First Amendment should scrutinize legal restrictions on research. Philip Hamburger has written about the constitutional flaws in human subjects research (as have I.) Federal research restrictions apply any time a researcher has the intent to produce generalizable knowledge, even if the acts in which he engages are otherwise perfectly legal. Since these restrictions target research qua research, any level of commitment to Empirical Liberty should trigger some form of constitutional scrutiny of these laws.

The second area ripe for reform is regulation that directly and purposefully restricts access to information. The First Amendment has long protected the right of “insiders” to voluntarily share information with “outsiders” by applying scrutiny when the law burdens or restricts information dissemination. But until recently, courts have been blind to the First Amendment’s application to another category of rules: those that restrict outsiders from trying to access information without the consent or cooperation of insiders. These laws include civil and criminal trade secrets laws, privacy laws (including the Wiretap Act, the Fair Credit Reporting Act, and the intrusion upon seclusion tort), and anti-hacking laws (including the Computer Fraud and Abuse Act).

These laws differ from conduct regulations of general applicability because they are designed to target and obstruct information-gathering activity. However, they also serve concrete and important state interests. Trade secrets can help foster innovation. Anti-hacking laws prevent espionage, denial-of-service attacks, and downstream identity theft. Wiretapping laws and legal duties of confidentiality provide a healthy and necessary respite from the pressure of social judgment. Many of the laws designed to manage information flows have a sufficiently compelling explanation at their center that they should withstand facial challenges and avoid wholesale razing.

Nevertheless, First Amendment interests in Empirical Liberty should give litigants an opportunity to challenge the contours of information-gathering prohibitions through as-applied challenges when the laws are used to intentionally interfere with harmless empirical inquiry. For example, as Orin Kerr has written about at length (both in his scholarship and in his blog posts), the CFAA has been interpreted in some jurisdictions to apply serious civil and criminal penalties to anybody who accesses a website for a purpose that violates the website’s terms of service. Both the criminal and civil remedies in the CFAA expose violators to much more liability than an ordinary contract claim would. In theory, a website’s terms of service could restrict access to users who prefer dogs to cats, and unsuspecting cat-lovers would become criminally liable by continuing to engage the website. Of course, websites do not have these sorts of wacky terms, but they do have terms that forbid scraping publicly available information or the use of fake identities to make online accounts.

The CFAA is a big threat to researchers who test online services for evidence of racial bias or who scrape publicly displayed information in order to put it in a more usable form for sociologists. The ACLU is currently representing a group of researchers to challenge the CFAA on First Amendment grounds using a theory quite consistent with Empirical Liberty, and I wish them luck.

Of course, courts cannot give the robust constitutional protection to Empirical Liberty that they currently apply to Claim Liberty. Unbridled experimentation and information-gathering would run roughshod over the rights and interests of other people. But courts can and should recognize a constitutional interest when the law cuts off empirical inquiry that poses little or no social harm.

Assessing the Value of Mandated Informational Disclosures

My last post listed mandated informational disclosures as a means to promote consumer education without using censorship rules. Mandated disclosures are an appealing option to help consumers make better-educated decisions. In theory, compelled informational disclosures can help reduce some of the foreseeable failures in the marketplace of ideas by adding good information rather than subtracting the bad. Federal and state regulators have used informational disclosures to try to nudge consumers into thinking about the “full price” rather than the sticker price of goods, and to nudge retailers and manufacturers into avoiding harmful practices. But the results have been mixed.

Some disclosure regimes changed the market for the better without heavy-handed restrictions. For example, the introduction of restaurant grades in Los Angeles appears to have improved compliance with public health codes.  But many mandated disclosures, possibly even most, do not achieve their goals of consumer education.

Omri Ben-Shahar’s and Carl Schneider’s book, More Than You Wanted to Know: The Failure of Mandated Disclosurewhich expands an earlier article, provides an explanation and examples for why mandated disclosures fail to change behavior. One key theme is that disclosure overload allows important information to fall into the background din. Moreover, disclosures can do worse than nothing; they can actually make markets worse. Daniel Ho’s article, Fudging the Nudge: Information Disclosure and Restaurant Gradingshows that contrary to the results from Los Angeles, restaurant letter grading systems did not improve hygiene in most cities because the disclosure systems were quickly distorted by grade inflation and inconsistency. Mandated disclosure can also lead to pernicious gaming, as seems to be the case with hospital grading systems. To get better grades, hospitals do all they can to avoid the sickest patients.

Also, highly visible and salient disclosures can over-correct for the problem they are attempting to address, as seems to be the case with black box warnings on antidepressants. Or they can be so discouraging that the consumer makes worse, more self-defeating decisions than he otherwise would. When Mexico tried an innovative disclosure system that alerted debtors about the magnitude of their credit problems, the disclosures caused default rates to go up.

For a long while, First Amendment law had little to say about mandated disclosures. In Zauderer v. Office of Disciplinary Counsel, the Supreme Court found that compelled informational disclosures, as opposed to compelled ideological disclosures, were permissible regulations on commercial actors. But Zauderer is now on shaky ground, in part because of the proliferation and problems with mandated informational disclosures, and in part because the line between “informational” and “ideological” disclosures is blurry. Consider, for example, requirements in Texas that abortion providers perform a sonogram and show the patient her fetus before terminating the pregnancy. (The Fifth Circuit upheld the law, but the opinion is controversial.)

Regular Volokh conspirator Jonathan Adler has taken up the issue of compelled informational speech in his work, which has been influential in my thinking. Adler argues that mandated disclosures concerning product attributes other than basic product contents and functions should trigger the heightened scrutiny that normally applies in compelled speech cases.

In a recent article, A Bad Education (published by the University of Illinois Law Review), my coauthors and I propose a way to assess mandated informational disclosures as a matter of good policy. In a related piece, I adapted the methodology for use in compelled speech cases. We used vignettes in an experimental setting to see whether mandated disclosures concern a material latent attribute, and whether they cause an over-reaction. Here’s the theory behind the work:

Imagine we knew the highest price a person would be willing to pay for a particular product under three conditions. In the “unwarned” condition, the consumer uses only their preexisting knowledge and assumptions about the product attribute that is of interest. In the “warned” condition, the consumer is given a simple disclosure in the style of a just-in-time mandated disclosure. In the “perfectly informed” condition, the consumer has as much knowledge about the product attribute as experts.

If we were to plot out the distribution of prices that all consumers were willing to pay for the product under all three of these conditions, we would see curves that look something like this:

Δ1 measures whether mandated disclosures are material for the majority of consumers. If Δ1> 0, then consumers care about the latent attribute on average, and their consumption decisions may be benefitted from a disclosure regime.

But mandated disclosures cannot make consumers perfectly informed. The best they can realistically do is provide a quick alert and rough education about the latent attribute. It’s possible that this realistic warning may not actually shift consumer behavior, in which case the warning is useless. It’s also possible that the warning can cause an overreaction, which is illustrated here as Δ2. If the overreaction is significant, if it is greater than Δ1, for example, then the mandated disclosure may be harmful to the market even if the latent attribute is a material one. Arguably, these concepts could be useful for First Amendment law, too. Compelled disclosures that are either useless (about non-material attributes) or misleading (that cause over-correction) are arguably unconstitutional because they needlessly or harmfully interfere with communications.

In our experiment, we randomized subjects into three conditions—unwarned, warned, and “well-educated” (the closest we could get to “perfectly informed” by supplying additional information about the costs and benefits about the latent attribute) to see how their willingness to pay was affected by the information they received. Here’s an example of what a disclosure about an ingredient used in shaving cream did to willingness to pay.

The subjects saw either the first line of information, the first and second, or all the information, depending on what group they were assigned to:


Their willingness to pay was affected by the disclosure levels:

Subjects who saw the warning without having additional information about what the chemical “Laureth-4” does were naturally wary of it. But with more information about its benefits and lack of risk, the curve returned to where it was without a warning. This example illustrates a non-material attribute (Δ1 = 0) and an over-reaction (Δ2>0). It would obviously be unwise for the government to mandate disclosures about “laureth-4.”

We used the shaving cream example to make sure that our methodology was sound—that it would pick up the effects of warnings and of education. If we didn’t get the results we see above, we would have worried that the experimental design was flawed. But with the reassurance of these results, we tested a range of other product attributes, some of which came from actual First Amendment compelled disclosure controversies.

Warnings about drug side effects and conflict minerals produced results that suggest the disclosures were material and did not cause over-reaction. (The finding on conflict minerals is interesting in light of a recent case before the D.C. Circuit Court of Appeals that struck down SEC laws mandating disclosure of conflict diamonds.) The labeling of a GMO food, on the other hand, appeared to be non-material and to cause an over-reaction—findings equivalent to the “Laureth-4” example. Disclosures related to privacy (data collection by an email provider, data sales, and the use of de-identified research data) seem to be non-material, although as we explain, the results for data selling are peculiar and interesting.

Our study results are tentative, meant to provide just a proof of concept for the methods. Our sample sizes were good but not great, and some will quibble with the way we characterized risks and benefits for the “well-educated” groups of research subjects. But the theory and methods, if not the specific results, can add value and precision to the debates about mandated disclosure. Lawmakers and courts ought to be asking whether prospective listeners in a mandated disclosure regime actually care about the information that is being provided, or if instead the disclosure is being used by the government to create a concern that does not actually exist.

Promoting Good Science Without Censorship

My last two posts summarized the extent to which the First Amendment should constrain the government when it seeks to censor weak scientific claims. To the extent that the government uses advertising restrictions in order to promote good science, free speech doctrine has already begun to get in the way.

This could nudge federal agencies and other lawmakers to use other means to promote research and consumer knowledge. That’s a good thing. The government could do much more than it currently does using its own powerful voice and by using other non-censorship regulations. Here are a few examples:

1. Funding and supporting more independent research.

One of the greatest drawbacks to our current system of food and drug testing is that we rely on self-interested companies to produce the bulk of research that regulators and consumers must use to make important decisions.

This is so in part because randomized controlled trials—the “gold standard” for research—are very costly. The companies that are likely to profit from positive research findings are the most eager to fund it. But for the company, the research is only worth its costs if it comes out the right way.

Randomized controlled trials can be gamed, and often are. Although controlled trials are supposed to avoid the selection effects that can mar studies of observational data, they can be distorted if researchers put conditions on study participants that ensure all the research subjects are unusually healthy, and therefore unusually likely to respond well to medicines and to avoid side effects. Independent researchers would have more interest in selecting the pool of research subjects in a way that better represents the future users of the product.

I do not think it’s realistic to expect NIH and other federal research organizations to have large enough budgets to run all clinical trials themselves. But more can be done with post-market observational data. While observational studies will always raise questions about whether there are sufficient controls to ensure that an effect is truly causal, they do have the virtue of coming from the real world, and reflecting what has happened in the field. Randomized controlled trials, by contrast, reflect the population of patients who happened to come down the pipeline and are willing to consent to an experiment. Controlled trials also tend to be quite small because of their costs, and may lack the power to detect rare but important therapies or side effects.

Compared to controlled trials, analysis of observational data is cheap. The federal government should encourage researchers to access medical data and link them to other consumer databases in order to uncover latent health benefits and risks. Instead, federal privacy and research law tends to do the opposite, discouraging data collection and erecting barriers to access.

2. Publicizing or certifying the findings of independent or well-conducted research.

The government can also use its own voice to promote the research that meets its favored scientific standards. Rather than using censorship rules, the government could certify health claims that meet its evidentiary standards, and allow manufacturers to proudly mark that certification in its advertising. The government can also contribute to consumer education more actively through its websites or through public service campaigns in the mainstream media.

3. Using ex-post liability and safe harbors

Regulators could also rely more heavily on ex post enforcement than the modern regulatory state typically does. Strict liability and ex post fines for products that prove to cause harm can do a lot to motivate research and caution by manufacturers. The government could also create strong incentives to engage in pre-market research using safe harbors from the fines and liability that would otherwise apply if a product turns out to be dangerous for its anticipated use.

4. Mandating disclosures about the evidence base (or lack thereof.)

The government could also make better use of rules requiring companies to provide disclaimers or other information. The compelled speech doctrine seems to be more forgiving than the free speech rules that apply to censorship, at least when compelled disclosures are used to provide relevant, purely informational content. Thus, the state could require companies that choose to make scientific or health claims in their advertising to provide the evidence supporting and contradicting each of their claims on a website or through a central clearinghouse.

Mandated disclosures have a tendency to proliferate quickly, though, and they don’t always have salutary effects. In the next post, I will address some serious limits to the ability for mandated disclosure to effectively educate consumers.

Free Speech and Anti-Knowledge

My last post described a risk-based approach to First Amendment protection of scientific claims, and showed why this shift would constrain some of the legal rules that currently apply to commercial speech. Although a risk-based approach would give the government less latitude to regulate scientific claims made by commercial speakers, it should give the government more latitude to regulate scientific claims made in public discourse. In circumstances where junk science is not only very likely to be false, but also likely to be influential and harmful, fines or liability rules should be able to survive First Amendment scrutiny.

The most tentative section of my article, Snake Oil, makes the case that some forms of “Anti-Knowledge” could be controlled through common law doctrines like negligence without doing mischief to core free speech values.

Let me start by defining some terms. Anti-Knowledge is a subset of factual statements. Free speech law already has to make a distinction between factual and non-factual speech. Factual claims are falsifiable and can, at least theoretically, be compared to objective reality. Non-factual speech (like opinions, moral or political philosophies, and artistic expression) do not assert claims about the objective world and therefore are not falsifiable. Sometimes the division between fact and opinion is hard to make, but courts and lawmakers can and frequently must do this (for example, in defamation law, which applies only to false statements of fact and not to opinions, no matter how hurtful they may be.)

It’s helpful to further subdivide the vast and diverse universe of factual claims into three categories: Accepted Knowledge, Anti-Knowledge, and Contested Knowledge.


Accepted knowledge is comprised of factual claims that that pass an appropriate epistemological standard for “truth.” We shouldn’t pretend that accepted knowledge is the same as truth (for reasons that Thomas Kuhn elaborated), but based on the best instruments and standards for evidence at a given time, a subset of factual claims can be specially recognized as reliable and verifiable.

Anti-Knowledge gets its definition from the Accepted Knowledge category. It contains any statement that is in direct conflict with statements contained in the “accepted knowledge” category. This leaves a large residual category that I have been calling “Contested Knowledge.”

My last post focused on advertising restrictions on Contested Knowledge. Within Contested Knowledge, First Amendment law has under-protected commercial speakers. I do not think that current law or First Amendment doctrine under-protects commercial speakers with respect to Anti-Knowledge. In fact, when it comes to Anti-Knowledge, the First Amendment has over-protected speakers—particularly speakers who engage in public discourse.

There is a popular misconception that Fox News successfully won a legal “right to lie” in a 1990s case involving reporting about dairy farms. Although this folk legal analysis is not correct, it is undeniable that Anti-Knowledge is strongly protected from government interference when a person speaks to a broad, indiscriminate audience about a matter of public concern. The 2012 Supreme Court opinion in United States v. Alvarez (which begins “Lying was his habit”) said in no uncertain terms that false statements are protected by the First Amendment as a general rule. But it also left open the possibility that lies may be regulated if they cause “legally cognizable” harm.

One would think that physical harm is the quintessential “legally cognizable harm,” yet tort cases against publishers and public speakers tend to lose on First Amendment grounds, even in instances where a speaker should know that his falsehood, if believed, are likely to cause grave physical harm or death. In the most famous case along these lines—Winter v. G.P. Putnam’s Sons—the First Amendment protected the publisher of an encyclopedia of mushrooms that assured readers a particular type of mushroom was safe to eat when in fact it was toxic. And other cases I discuss in the full draft suggest the First Amendment supplies broad immunity that protects charlatans, celebrities, and other authors who make pseudoscientific statements in the press or public gatherings.

The strong First Amendment protection for factual claims in public discourse stems from an admirable commitment to epistemic modesty. As the Court said in Alvarez, “Our constitutional tradition stands against the idea that we need Oceania’s Ministry of Truth.” But protection goes too far if it can’t be outweighed by the public interest in deterring reckless anti-knowledge that causes foreseeable physical harm.

What I’m suggesting here runs against the trend to expand and strengthen First Amendment protections (a trend that I’ve supported in most of my work). So consider this hypothetical to test whether liability for pseudo-science can withstand strong free speech scrutiny: should courts or legislators be able to create civil liability for doctors or documentary-makers who publicly provoke fear about trace amounts of mercury used in vaccinations? If the family of a child who has a severe case of measles can provide credible evidence that a book or documentary persuaded his family, or the family of a child who transmitted the disease to him, to avoid the measles vaccine out of unfounded concern that the shot could increase the risk of autism, should the First Amendment come to the speakers’ defense?

It seems to me that the social benefits from this sort of legal intervention are at least as compelling as the benefits from well-working defamation law. It is true that courts will have to determine on a case-by-case basis whether a factual claim was false (that is, whether it is anti-knowledge), whether the authors had a negligent or reckless disregard for the truth, and whether the speech actually influenced listeners. But courts must make these same sorts of factual determinations in defamation law, too.

Although defamation is often listed as one of the special categories of unprotected speech, the principled explanation for those historically unprotected categories  is that they mark distinct areas where the non-speech interests outweigh the interests in free speech. Other regulations that target harmful misinformation should be able to prove that they strike an appropriate balance, too, albeit within strict scrutiny instead of outside of it.

Free Speech in the Fog of Scientific Uncertainty

Donald Trump’s assault on journalistic integrity and shared verifiable facts has ignited a reaction among public intellectuals to demand fealty to scientific truth. Unfortunately the reaction, like so many produced in the haste of political controversy, has oversimplified and overcorrected for the problem. One common assumption within the resistance is that existing systems for regulating scientific claims are self-evidently wise.

My new article, Snake Oil (forthcoming in the Washington Law Review), is therefore coming out at a rather inconvenient time. I loathe the Trump administration’s disregard and delegitimization of scientific institutions including expert federal agencies. Nevertheless, the methods that some expert agencies use to constrain the information flow to consumers are highly flawed. Where evidence is strong and abundant, fealty to an expert opinion by non-experts and lawmakers is sensible and efficient. But deference to an expert authority is not as wise where the science is weak and still-developing. Too often, scholars and policymakers justify broad government powers to regulate speech about weak and contested science by bootstrapping their arguments to examples where the scientific evidence is strong (like global warming), and where deference to experts is therefore highly sensible. But most of the hard choices that humans have to make, whether they are gravely important or relatively minor, are hard precisely because they concern factual questions with only a weak evidence base. Most decisions must be made where the science is still a fog.

Consider, for example, the Food & Drug Admistration’s and Federal Trade Commission’s regulatory powers over health claims. Both agencies have broad powers to censor advertising that is made without sufficient scientific support. And they use high standards (usually two clinical trials, though one court has suggested that one should be sufficient for foods), effectively silencing commercial speech where an evidence base is weak or still developing. Scholars largely support the existing rules requiring agency pre-clearance, and typically defend their constitutionality on the basis that unsubstantiated claims fall outside the coverage of protected commercial speech.

I disagree with the scholarly consensus. The strict standards applied to advertising are not consistent with free speech law, and they do a poor job fostering debate and science literacy. They often have the effect of screening out the best available evidence because it falls short of the agencies’ demanding requirements. At the same time, the standards tend to spur over-confidence in the claims that do meet the applicable standards, or that skate around the rules by making non-specific, comforting claims about health or wellness.

The greatest flaw in the FDA and FTC approach to advertising regulations is that the rules are designed to detect falsity rather than risk. Risk is, of course, related to the chance that a scientific claim turns out to be false, but falsity is not a good proxy on its own. The consequences from being influenced by a claim if it turns out to be false and the consequences of screening out a claim if it turns out to be true are often asymmetric.

The FDA’s restrictions on advertising for e-Cigarettes illustrate the problems with the current approach to commercial speech regulations. E-Cigarette manufacturers cannot make promotional claims that vaping is healthier than cigarettes even though the best available evidence suggests it is. Moreover, the risks from screening out this information if true are much greater than the risks from allowing the claims to be broadcast even if they turn out to be false since the worst case scenario for e-cigarettes is something comparable to the health effects of smoking. Yet the advertising ban has slowed the progress that had been made in persuading current smokers to switch to vaping. The concern that vaping will induce more teenagers to start smoking cigarettes also appears to be unfounded; states that allow the sale of e-cigarettes to minors actually have lower rates of teenage cigarette smoking than states that ban the sales, even with controls for preexisting propensity to smoke.)

A similar lesson can be learned from the FTC’s clumsy history with the egg industry, which I use as the key example in the introduction to Snake Oil. The FTC prohibited the egg industry from stating that there is no evidence that the consumption of eggs increases blood cholesterol or heart attack risks—a statement that was correct at the time and has stronger support today. Of course the egg industry’s ad campaign was completely self-serving, but the campaign would have served the public health as well since the fear of egg cholesterol drove consumers to eat even less healthy proteins like pork and red meat. There are other examples where FTC or FDA rules served public health interests, but these victories can be preserved by shaping a commercial speech rule around risk or harm rather than around “truth”.

To a large extent, this problem I am describing is imbedded in First Amendment doctrine. Central Hudson, the foundational commercial speech case, denied First Amendment protection to promotional speech that is false or misleading, so federal agencies and other law makers were encouraged to focus on the accuracy of advertising claims rather than the risks. But courts should begin the hard work now of readjusting free speech doctrine so that First Amendment protections are overcome only by concrete risks and harm rather than by inevitable doubts about the “truth” of a factual statement that is very difficult to test. The adjustment will put some constraints on agencies like the FDA and FTC (as well as courts interpreting state and federal false advertising law) to ensure that enforcement targets specific claims that are not only likely to be wrong, but to cause dangerous influences in consumer behavior.

By the way, this adjustment does not necessarily require the dismantling of current regulations. Many of the practices at the FDA and FTC could be justified using a risk-centered approach. For example, First Amendment law should allow the FDA to defend a position, based on historical evidence, that drugs with no history of human consumption are presumptively harmful or ineffective until proven otherwise. When it comes to already approved drugs, or to foods that have long been consumed by humans without acute negative effects, a categorical pre-clearance rule will be harder to defend on the basis of risk, but not impossible. Advertising rules that take both the weakness of an evidence base and the net effects of a false claim into account should be able to withstand First Amendment scrutiny. Free speech challenges can slowly steer false advertising law in this direction, and make commercial speech doctrine more compatible with the general trend in First Amendment law to permit regulation only when the government has credible evidence of concrete harm.

Commercial speech is not the only area of First Amendment jurisprudence that suffers from too much focus on truth and not enough on risk; when it comes to speech of public concern, First Amendment law has similar flaws but with the opposite valence. It overprotects statements that contradict accepted and well-supported scientific knowledge that are made in public debate, even if the speech is very likely to cause foreseeable reliance and harm. In my next post, I will discuss how a harm-based approach can give the government a freer hand to regulate public discourse in instances where junk science is likely to be false, influential, and harmful.

Product-Defining Speech

The state of Florida requires all milk sold in the state to contain Vitamin A. Thus, skim milk products, which have most of the naturally-occurring Vitamin A skimmed off with the cream, has to be fortified with added Vitamin A in order to meet Florida’s standards. This was a problem for the Ocheesee Creamery, which prided itself in creating dairy products without additives. The state provided a route for Ocheesee to sell its skim milk under its regulations for imitation milk, but Ocheesee would have been prohibited from labeling its skim milk “skim milk,” and would instead have had to use the term “imitation milk product.” Yum.

The creamery challenged the regulation under the First Amendment, and the 11th Circuit Court of Appeals sided with the creamery. The result of the case is not surprising given how absurd the facts are. Although the state has admirable reasons for wanting milk to be fortified with Vitamin A, it was torturing the English language by prohibiting sellers from using the words “skim milk” to describe what is, in fact, skim milk in its purest possible form. The facts would have been less absurd for the state if the popular understanding of the word “milk” was not the stuff produced by cows (and other mammals) but a product that always necessarily contains Vitamin A. This, though, does not fit common usage.

Even though I think the court got this case right, I also believe that what the state of Florida was doing in this case is not at all unusual. The holding in this case may have broader implications than the litigants or 11th Circuit panel realize. In a nutshell, what Florida wanted to do was force all milk producers to ensure that their products contain Vitamin A for public health reasons. Not every product has to have Vitamin A, of course, just milk. And how do we know what “milk” is? Rather than defining the nature of the substance, regulators can look to see how the sellers describe their own products. If they are selling something that they themselves refer to as “milk,” then it’s milk, and must comply with the Vitamin A rules.

This approach to product definition is very common. Regulators frequently define the scope of their product and service regulations based on the promotional speech of vendors. Let me give two examples to demonstrate that it is not unusual:

First, credit reports. The Fair Credit Reporting Act creates a set of rights and obligations for credit reporting agencies and the creditors or employers that access consumer reports. But what is a “consumer report?”

Here is the statutory definition:

15 U.S. Code § 1681a

(d)Consumer Report.—

(1)In general.—The term “consumer report” means any written, oral, or other communication of any information by a consumer reporting agency bearing on a consumer’s credit worthiness, credit standing, credit capacity, character, general reputation, personal characteristics, or mode of living which is used or expected to be used or collected in whole or in part for the purpose of serving as a factor in establishing the consumer’s eligibility for—

(A) credit or insurance to be used primarily for personal, family, or household purposes;

(B) employment purposes; or

(C) [some other uses associated with credit, insurance, or business transactions]

And a consumer reporting agency, in a nutshell, is any entity that assembles information “for the purpose” of furnishing consumer reports to third parties. (15 USC §1681a (f).)

The problem, though, is that employers (and creditors, too) frequently use information outside standard credit reports to make decisions about lending and hiring. For example, employers sometimes google prospective employees or look at their public Facebook accounts before making interviewing or hiring decisions. Google has knowledge that its search services are used to make employment decisions and stands to benefit (at least marginally) from searches by employers. Does that mean Google is a credit reporting agency?

To avoid the dire consequences of applying the FCRA in a way that locks in a vast range of information service providers, the FTC and courts have limited the scope of “credit reporting agency” to instances where a purpose to create consumer reports is demonstrated through speech—that is, when a company has promoted the use of its information services for employment or credit decisions. This is why Spokeo was presumed to be subject to the FCRA in the recent Supreme Court case, Spokeo v. Robbins.

Spokeo opted into the regulatory strictures of the FCRA through commercial speech encouraging employers to use its profiles for employment decisions. Since Google hasn’t promoted its search services for use by employers (so far as I know), it has avoided the application of the FCRA.  Thus, the FCRA is a product-defining speech regulation.  There are two layers of speech here because the product itself is also speech– employability and credit-worthiness information. But the requirements of FCRA apply not based on how credit-worthiness information is used, but rather based on whether it is advertised for that use.

Another very important example comes from the Food, Drug, and Cosmetic Act. How do we know whether a chemical available on the open market is a “drug,” requiring significant testing and pre-approval by the FDA? The meaning of a “drug” (as well as a “food” or a “supplement”) is surprisingly obtuse. Just listen to this FDA official avoid defining “drug” in this Freakonomics podcast, and refer the show’s host to FDA attorneys.

Another very important example comes from the Food, Drug, and Cosmetic Act. How do we know whether a chemical available on the open market is a “drug,” requiring significant testing and pre-approval by the FDA? The meaning of a “drug” (as well as a “food” or a “supplement”) is surprisingly obtuse. Just listen to Margaret Hamburg, a former FDA Commissioner, avoid defining “food” and “drug” in this Freakonomics podcast, and then refer the show’s host to FDA attorneys (starting at minute 7:50.)

The boundaries of the term “drug” are made by product-defining speech. If a company says that its product can be used for a therapeutic purpose, then it is a drug and must be approved and labeled appropriately for that particular medical use.

I should make clear that this is not a universally accepted view. My colleague Chris Robertson argues that the FDCA is not a regulation of speech because speech is used only as evidence of the seller’s intent to introduce a drug into the marketplace. I disagree. As I explain in my new draft article, Snake Oil (forthcoming in the Washington Law Review), the FDA has carefully avoided using forms of evidence that would easily meet the standards for a knowledge or purpose mental state. For example, a drug manufacturer that knows that a drug is used frequently (even more often than not) for an off-label purpose and happily supplies it for those uses is not treated by the FDA as introducing a “mislabeled” drug. But if the company acknowledges through speech that the drug has been successfully used for an off-label treatment of some sort, then the manufacturer faces heavy fines and criminal liability. The sine qua non nature of speech to prove “intent” is evident not only from the history of FDA enforcement, but from the FDA’s own memo on the First Amendment’s application to public health law, which anticipates using commercial speech as evidence of intent and provides no examples of what other, non-speech conduct could be used to prove intent.

So, should free speech jurisprudence treat regulations of product-defining speech with the same skepticism and probing scrutiny that it applies to other regulations of speech? Or is product-defining speech a sensible way to limit the scope of product regulations without having to define the essential nature of things like “milk,” “credit reports,” and “drugs?” After all, regulations that hinge on commercial speech give sellers some flexibility. Sellers can avoid a set of regulations by avoiding certain forms of speech (like off-label promotion). There is also little room for doubt about when a regulation will apply, since manufacturers are necessarily on notice when they label or describe their products.

But these benefits (if they even are benefits) do not outweigh the free speech problems. First, when regulators avoid defining the scope of a product regulation using objective terms, it means that they are leaving significant loopholes for unregulated products. If there is a serious public health risk from all or a subset of off-label prescription drug use (or from milk or creditworthiness information, for that matter), current drug regulations leave the public exposed to much of that risk by allowing doctors to prescribe the drugs, and pharmacies and drug companies to supply them. Conversely, to the extent off-label drug use is appropriate and even good for patients, suppression of commercial speech comes with significant public health consequences, too, by removing a particularly powerful form of information-sharing.

Promotional speech, in other words, is often used as a rough proxy for a product’s risks when regulators could and should identify the scope of a product in objective terms and directly manage its risks. The best defense of these types of uses of speech regulations, which again comes from my colleague Chris Robertson, is that these regulations can act as a sort of brake on the market that incentivizes sellers to produce research and gather evidence to evaluate whether a drug is safe and effective. But FDA pre-screening is a costly and circuitous way to accomplish these goals, particularly for foods and drugs that have a history and known track record for human consumption.

Others have justified regulations of drug manufacturers’ speech on the basis that such promotion, without appropriate pre-vetting, are presumptively false or misleading. As my next post will explain, determining whether words are false or misleading is not nearly as simple as the free speech cases and scholarly literature suggests.