All About That Base Rate

Nov 15, 2023 David FagundesAdd a Comment

David Fagundes

Some years ago I attended a presentation by a musicologist who specialized in giving testimony in copyright litigation. Here’s how he tried to grab the audience: First, he would play a clip from a well-known track by a popular musician or band. Then he would play a selection from an earlier, lesser-known track by an obscure musician or band that sounded similar to the first clip, all while giving the audience a wide-eyed stare. The impression this created was intentional and unmistakable. Clearly the well-known artist had copied from the lesser-known one!

The audience, mostly laypeople, certainly bought it, based on the gasps that accompanied the presenter’s schtick. I did not, and left frustrated that the musicologist-turned-expert-witness had tricked the audience into thinking that he had exposed several instances of egregious copyright infringement. I knew something was wrong but had difficulty putting my finger on just what was the problem with the presenter’s move.

Thanks to Christopher Buccafusco and Rebecca Tushnet’s sparkling essay, Of Bass Notes and Base Rates: Avoiding Mistaken Inferences About Copying, I finally have a clear picture of the error that afflicted that presentation and so much copyright litigation. As the authors explain, the application of copyright’s substantial similarity problem suffers from base rate neglect, which causes courts and litigants to significantly overstate the likelihood that a defendant copied from a plaintiff.

Let’s unpack this a bit. Plaintiffs in copyright infringement litigation must make a threshold showing that the defendant copied their work. This usually entails proof that the defendant had access to the plaintiff’s work and that the works share similarities that are probative of copying. Answering the latter question in the plaintiff’s favor depends on their amassing enough evidence to support an inference of copying. At this stage, plaintiffs often retain expert witnesses—usually musicologists,¹ like the speaker I mentioned above—who testify that the degree of similarity between the two works is so great and distinctive that it could not be explained by, for example, independent creation or copying from a public domain source.

Here, the authors argue, is where base rate error creeps in. When a musicologist testifies that the quantum of similarity between two works enables an inference of copying, that ignores the crucial issue of how likely such similarity would be absent copying (the base rate). After all, there are only so many appealing chord progressions out there, and the plaintiff presumably created the work herself, so unless we know how likely it is that the similarities at issue would occur in the normal course, it is impossible to say whether the similarities in a given dispute are truly probative of copying.

The authors state that their argument is “modest.” I dissent. Not only did their central insight help me understand what was wrong with the musicologist’s presentation I resented so much (thanks!) but it has the potential to change the way we think and conduct copyright litigation. The past couple of decades have seen an uptick in infringement lawsuits by lesser-known artists alleging that hit tracks by popular artists (e.g., Katy Perry, Led Zeppelin, Marvin Gaye, Ed Sheeran, Taylor Swift, Lana Del Rey, Dua Lipa, and many many more) infringe their earlier, lesser-known works. A central feature of these plaintiffs’ cases is the testimony of musicologists that the degree of similarity between the works in suit is so great that it is explicable only by the defendant’s copying.

But Buccafusco and Tushnet have shown the expert testimony in these cases shares a common, fatal flaw: base rate neglect. Not one of the experts in any of these cases actually knew how likely it was that the given similarity would occur absent copying. This means that experts’ central evidentiary contribution to the litigation (quantum of similarity implies copying) is not a reliable, informed opinion but just highly articulate hand-waving. Many critics decry the purported irrelevance of legal scholarship to law, but this essay represents a conspicuous counterpoint.

The authors limit their discussion of the implications of their insight to the admissibility of expert testimony. Yet it may have other important payoffs. For example, some courts have (controversially) held that where the degree of similarity between two works is extremely high, that enables an inference of copying, regardless of evidence of access. Judges applying this “striking similarity” doctrine often commit base rate errors, assuming that a high degree of similarity itself warrants a conclusion that the defendant copied the plaintiff’s work absent any sense of how likely the similarity would otherwise be (i.e., the base rate).

Another tantalizing question provoked by this essay is whether and how the base rate problem varies within and across genres. Music is especially ripe for this fallacy because the number of notes in the scale is low and the number of appealing combinations of them is lower still, all of which suggests a higher base rate of similarity. But some genres of music are notoriously self-similar, such as ska or reggae, which are defined by a syncopated guitar rhythm. Musical works in these categories are especially likely to have a high base rate of similarity. The same may be true in some literary (formulaic romance or fantasy novels) and artistic (traditional portraiture) contexts that are defined by certain core features. By contrast, the alphabet yields many more available combinations of appealing words and phrases than the musical scale does appealing progressions of notes, so the base rate of similarity in most literary works is likely lower. None of these assertions about relative base rates can be reduced to specific numbers (the authors point out that we cannot actually know the base rate of similarity in any genre) but they each illustrate how the idea of base rates serves as a useful heuristic when thinking about copyright infringement generally.

The authors conclude that because we cannot deduce the base rate of similarity in any instance of claimed infringement, at present the wisest move is to simply bar experts from rendering conclusions about whether similarity supports an inference of copying. And while this does reflect the current state of play, it raises an intriguing possibility that may be realized sooner than one might think. If a large language model can learn from many billions of data points to produce astonishingly plausible simulacra of human responses to prompts, why couldn’t a similar artificial intelligence learn from the millions of available tracks to find the likelihood that certain similarities between musical works occur naturally? Big data might hold the solution to copyright’s base rate neglect problem.

By importing a known but underappreciated idea from quantitative analysis, Christopher Buccafusco and Rebecca Tushnet have generated a simple but significant insight that has the potential to change the way lawyers, judges, and academics think about both the doctrine of copyright litigation and how it unfolds in litigation. And they do it all in under 9000 words. This article is based and I rate it very highly.

1. Judges’ reliance on musicologists to compare musical works in suit to preexisting compositions has grown so prevalent that some authors have compared it to the patent law practice of assessing inventions in light of prior art. See Joseph Fishman & Kristelia Garcia, Authoring Prior Art, 75 Vand. L. Rev. 1159 (2022).

Copyright Fiduciaries: Problems and Solutions

Oct 18, 2023 Jessica SilbeyAdd a Comment

Jessica Silbey

In this forthcoming article, Andrew Gilden and Eva Subotnik begin an important conversation about an underexplored area of copyright law. Their focus is copyright law’s inconsistent treatment of mental capacity. Under copyright law, copyright authors can produce valuable copyrighted work but those same authors may lack the legal capacity to make decisions about if, when, or how to exploit that work. For example, children and people with mental illness or disability can be copyright authors, but they cannot license that work (or refuse to license it) without a legally competent surrogate. The authors explain that this inconsistency leads to injustices for which they offer reforms.

The article starts with the engaging example of the Britney Spears’ 13-year conservatorship, controlled by her father, which from the age of 26 prevented her from making decisions about her life and career. All the while, Spears wrote and performed her songs, building a multimillion dollar portfolio over which she had no control. She was the author of her music, but she had no control over it because she lacked the legal capacity to form binding contracts, or so said a court. She resisted the conservatorship without success for over a decade. The article is full of many other such examples, including of teenage authors, elderly creators, and authors with mental illnesses.

The article argues that this “capacity gap” between authorship and control is a problem from within copyright law as a matter of doctrinal consistency and from a fairness perspective of avoiding exploitation. It further argues there is something unique about copyright law’s capacity gap because, unlike a usual trust situation when assets are transferred to a competent person upon incapacity, in a copyright situation the incapacitated person can produce new wealth while being subject to the control of the trustee or conservator. As the authors say “this dynamic creates unique opportunities and incentives for abuse” and undermines copyright law’s solicitude for authors’ wellbeing.

Both Gilden and Subotnik teach trust and estates, and so they are a great pair to explore the intersection of fiduciary law with intellectual property. Both have focused recent writing on problems that arise for copyright authors after death, in, for example, what Gilden calls the “social media afterlife” and what Subotnik has recently called “dead-hand guidance” in a “preferable testamentary approach for artists.” This new article is about authors still creating, some with cognitive disabilities and others who are simply young. Fiduciaries encourage valuable artistic productivity, for the fiduciaries’ benefit and audiences’, no doubt. But it is at best debatable, according to all the examples the article provides, whether the rate and nature of artistic production is in the best interest of the authors.

The article explains that this conflict between author and fiduciary undercuts the utilitarian justification for U.S. copyright law, which focuses on increasing productivity and the financial incentive of copyright’s exclusive rights. As explained, copyright is available to those lacking legal capacity to be incentivized, which challenges the law’s carrot-stick mechanism for creative production; moreover, the copyright incentive is possibly working on what the article calls the “wrong” people who do not have the authors’ or the audiences’ best interest in mind.  The capacity gap also quite clearly undercuts the personhood theory of copyright law, in which the author’s dignitary interests are paramount and protected through authorial control over the work. By situating the copyright’s capacity gap within several major theoretical justifications for copyright law, the article adds to the rich scholarship on copyright’s evolution from its origins to its contemporary manifestations.

This article has many virtues. It teaches those of us in the intellectual property field a lot about critical features of trust and estate law, which frequently intersect with copyright law, in particular, and with which I’d surmise most of us are fairly unfamiliar. It weaves the two fields in an elegant and clarifying manner and is rich with contemporary examples in which the capacity gap problem arises. These examples would be wonderful classroom discussion topics for those of us teaching in either legal field. The article also engages recent copyright law scholarship concerning alternative IP theories and IP’s failure to protect marginalized authors, advancing these burgeoning fields within IP law. And it provides concrete guidance and reform suggestions to limit the risks of exploitation and abuse that can emerge from copyright’s capacity gap.

The last part of the paper titled “Minding the Gap” will be most interesting to practicing lawyers, administrators, and authors. It harnesses some existing mechanisms within the Copyright Act (such as termination of transfers) and of the Copyright Office (registration and recordation) offering some clear-eyed and reasonable first steps toward addressing problems arising from copyright law’s capacity gap. I would not be surprised to see this article cited and used for reason of its last section alone. Overall, the article is insightful and its topic important. It is a commendable contribution to copyright literature.

Name-Dropping Government Agencies in Advertising

Sep 20, 2023 Alexandra RobertsAdd a Comment

Alexandra Roberts

When shoppers see “Now FDA approved!” on a bottle of Excedrin, does it make them more likely to select that option over a competing product? Will consumers choose a brand of dietary supplement marketed as a “patented blend” over one that doesn’t make patent claims?

In Conjuring the Flag: The Problem of Implied Government Endorsements, Michael Mattioli argues that advertisers use claims about intellectual property and regulatory approvals to mislead consumers about their products’ quality, safety, efficacy, or legitimacy. By reference to or use of the US Patent and Trademark Office or the Food & Drug Administration, advertisers borrow those agencies’ halos to imply that branded products from bugspray to hairspray to nasal spray are superior to competitors’ versions. Mattioli cites survey evidence finding most consumers interpret government stamps of approval like “FDA-approved” and “patented” as endorsements of quality. In Conjuring the Flag, he amasses data on the different ways advertisers reference agency or IP approval to appeal to consumers, highlights why that strategy is misleading, and proposes ways FTC could curb it.

What does it mean when an advertiser claims a product like a can of bugspray is “patented”? Patent lawyers know it only means some aspect of the product or packaging was deemed useful, novel, and non-obvious or some element of its design qualified as original and ornamental. For the bugspray, that might mean an inventor or designer acquired a utility patent that covers the spray mechanism on the dispenser or a design patent for the pattern adorning the can, not that the spray itself is particularly safe, effective, innovative, or non-toxic. Mattioli’s study analyzed hundreds of ads that reference patents, but none of the ads in his data set provided patent numbers or identified what the patent actually covered. Patent-related advertising claims appear most often in ads for products intended to be ingested or applied to the body, such as supplements, toothpastes, and skincare products; consumers may crave reassurance that these types of products have been tested and found safe, so “a significant number of companies are using their patents to cultivate an aura of legitimacy and safety in industries that lack rigorous regulatory oversight.”

Mattioli also reviewed eight hundred ads that tout FDA approval, clearance, or registration—these claims appear most frequently in connection with hair care, cancer treatments, and food and drink containers. But FDA approval means only that the FDA has determined a drug or device’s benefits outweigh its known risks for the intended use, not that the product is safe or effective for other uses or that it lacks side effects or complications. Here, too, advertisers conjure the flag in misleading ways to signal safety or superiority to consumers and sell more products, capitalizing on the likelihood that consumers either don’t know or aren’t thinking deeply about what the claim conveys.

When it comes to trademarks, the misleading advertising mechanism is different. Advertisers sometimes use trademark registration to conjure the flag, as with marks like U.S. HEALTH CLUB for vitamins. But more often advertisers obtain trademark registrations that make their patent- and FDA-related claims look even more official: they register marks like JEANS WITH PATENTED FIT for jeans; PATENTED INGREDIENTS for cosmetics; FDA APPROVED MEDICAL SUPPLIES for devices; and FDA REGISTERED for dietary supplements. More aggressive application of failure to function and deceptiveness doctrines could and probably should be used to screen out marks like these at the USPTO, and the agency could update its manual of examining procedure to include that guidance.

While advertising claims and trademark registrations that reference patents and FDA approvals are sometimes literally false, they are more often merely misleading: “Half-truths and implications are, of course, what conjuring the flag is all about.” To address the problem and better protect consumers, Mattioli proposes FTC launch an initiative focused on monitoring and investigating such ads along with a comprehensive public awareness campaign to teach consumers how to identify and report ad claims that imply government endorsement. Because consumer surveys or expert testimony are typically needed to establish a claim is misleading, he further proposes modifying the legal standard to create a rebuttable presumption that any ad referring to a patent, FDA approval, or other government action in a way that implies approval of a product is potentially misleading, shifting the burden of proof to the advertiser to demonstrate the ad is not misleading.

While materiality is a fact-specific inquiry, it is well within FTC’s power to conduct a broader survey or study to gauge whether ads that proclaim products or services “FDA approved” or “patented” mislead consumers in ways that impact their purchasing decisions. In fact, FTC’s Advertising Division has performed surveys and other research on advertising claims like “organic,” “recycled content,” “results not typical,” and “made in the USA” to undergird its guidance and policy statements on those categories of representations. This intervention is crucial because while it’s possible these representations are misleading and constitute unfair competition, it’s also possible consumers skate right past them or regard them as mere puffery. Mattioli’s intuition that these types of claims are misleading is compelling and likely correct, but further study would give FTC a solid basis to pursue claims that conjure the flag more aggressively and systematically.

Gender Equity in the Market for Collegiate Name, Image, and Likeness Rights

Aug 8, 2023 Michael W. Carroll1 comment

Michael W. Carroll

Within the field of intellectual property law, there are not too many legal or economic developments that would qualify for an event study. But, on July 1, 2021, such an event occurred when a new rule issued by the National Collegiate Athletic Association (NCAA) took effect. Prior to that date, intercollegiate athletes were prohibited from exercising their right of publicity or any other rights in their name, image, or likeness (NIL) to endorse products, services, or businesses in a commercial manner. Under the rule change, these athletes, numbering nearly 500,000 at the time, suddenly became free to license or otherwise use their NIL rights commercially, and a new market was suddenly born.

In The NIL Glass Ceiling, Professor Boston explains how the market for intercollegiate NIL rights has quickly evolved in a way that provides these athletes with long-denied revenue but with disparate outcomes for athletes who identify as men or women.² She argues that these disparities are problematic both because female athletes should be entitled to a greater share of the revenue in this market and because these disparities send an unwelcome message to female athletes about the state of gender equity in intercollegiate athletics and in the workplace. She argues that more gender-equal outcomes could arise if schools were subject to scrutiny under Title IX, applicable Department of Education regulations under Title IX, and NCAA rules that govern certain third-party support for intercollegiate athletic programs in the case of disparities in NIL revenues paid by certain third parties directly to athletes.

This article makes three main contributions to the literature. First, it explains that the NIL market is comprised of three main forms of agreement or licensing. Athletes are paid either from deals they self-arrange, from deals they make with their schools, or from those made with third-party Collectives that pool assets to finance NIL licensing. The second contribution is to focus on the outsize role of the Collectives in this market and to argue that the relationships between these and the schools whose athletes they support are sufficiently close to subject payments made by these Collectives to scrutiny under the school’s legal duties, under applicable law, to not discriminate on the basis of gender. The third contribution is to argue that the important role that this new NIL market is playing in intercollegiate athletics requires legal reform to account for this development. Professor Boston discusses a range of options for how such reform might be achieved.

One feature of Professor Boston’s description of this market I found interesting is that even though the right of publicity is not recognized in every state, as she notes, and only some athletes have made uses in commerce of their name, image, or likeness in a way that would qualify for trademark protection, the market for NIL licenses appears to disregard these differences and treat all athletes as being able to supply consideration by granting NIL permissions in exchange for payment. I have no doubt that a court would treat such permission as sufficient consideration under contract law, but it would not be surprising if this new market were to lead to judicial or legislative recognition of a right of publicity where none exists today.

Professor Boston explains that in this market, some athletes represent themselves, or work with an agent, to negotiate NIL endorsement deals. Some sponsors, however, seek NIL arrangements with an entire team. In these situations, the sponsor may seek an alternative to individual negotiations. One alternative is for schools to facilitate NIL transactions on behalf of their athletes. Some states permit this practice, while others prohibit it. Even when a school is permitted to provide this facilitation, the NCAA’s rules prohibit such facilitation for prospective student-athletes.

Most schools do not provide such facilitation, which leaves space for third parties to play a role, which they quickly have moved to do. Professor Boston divides these into third parties that provide technology platforms for NIL licensing, talent agents, and other advisors who provide support for NIL transactions, and traditional Boosters, supplemented by a new entity, the Collective.

There are more than 100 of these Collectives, which use different mechanisms, such as crowdfunding or membership tiers, to create an asset pool that they use to pay athletes for their NIL endorsements. To date, these Collectives limit their support to individual schools. These Collectives focus their resources on schools’ football and men’s basketball teams, which helps fuel the gender disparities in NIL revenues.

Professor Boston explains that advocates for athletes’ ability to monetize their NIL rights thought that such opportunities would play a potentially equalizing role between men’s and women’s sports. That is not how it has turned out so far. Overall, male athletes receive three times as much as female athletes in all three of NCAA’s divisions. She shows that because Collectives provide a significant share of NIL revenues, the gender disparities in their practices substantially contribute to overall disparities in NIL revenues. That raises the question this article addresses: should schools be held legally responsible under Title IX for the gender disparities attributable to Collectives’ NIL practices?

Professor Boston provides a very nice, succinct explanation of applicable law. Under applicable regulations interpreting Title IX, schools’ compliance obligations fall within three categories: (1) athletic scholarships, (2) benefits and services, and (3) effective accommodation of students’ interests and abilities. Within the “benefits and services” category, equal treatment obligations extend to a school’s recruiting and publicity activities.

Professor Boston argues persuasively that even though schools are prohibited from explicitly using potential NIL revenue as a recruiting tool, they do, and they do so in a gender-disparate manner. Similarly, publicity enhances athletes’ NIL licensing opportunities, and women’s sports on average receive less publicity than men’s. Professor Boston shows, for example, how when all women’s NCAA basketball tournaments games became televised, players’ NIL opportunities increased measurably.

She further demonstrates how schools have been, and can be, held responsible if athletes receive gender-disparate benefits and services from third-party sources. She argues that the relationship between Collectives and schools and the gender disparities arising from Collectives’ focus on men’s NIL licensing are actionable under Title IX.

Professor Boston offers regulators and Congress a menu of options to update Title IX from the easiest, most feasible options, to the most difficult but most desirable outcome, which would be legislative amendments to Title IX.

I learned a lot from this article. Like Professor Boston, I recognize some of the challenges involved in holding schools responsible for NIL licensing done between third parties and athletes. But, I agree with her that schools are not in a fully arms-length posture with these Collectives. I found most persuasive her argument about how potential NIL revenue is playing a significant role in recruiting in at least some sports. Since recruiting is an equal treatment factor, and gender-disparate potential NIL revenues fueled by Collectives’ practices have gender-disparate impacts on recruiting practices, that seems like a Title IX problem in need of a solution.

1. Recognizing that not all athletes so identify, for purposes of discussing NIL revenues from intercollegiate sports, Professor Boston’s article and this review are constrained by existing NCAA rules, under which athletes must select from its binary sports classifications, and must meet its requirements for those classifications.

Generative AI Meets Copyright

Jul 11, 2023 Pamela SamuelsonAdd a Comment

Pamela Samuelson

ChatGPT, Midjourney, and Copilot are among the numerous generative AI systems launched in the last year or so. They have attracted a huge number of users as well as several lawsuits. Among the lawsuits’ claims are that the makers of these systems are direct and indirect infringers of copyright because of their use of millions of in-copyright works as training data and because outputs of these generative AI programs are infringing derivative works.

At the core of these AI systems are foundation models on which the authors focus in their fascinating new article. They define this term as “large pre-trained machine learning models that are used as a starting point for various computational tasks,” including generative AI systems that may produce text, images, and/or software code in response to user prompts. The article identifies various actors who contribute to elements of these AI systems, including data creators, data curators, model creators, model deployers, and model users.

Those of us who are intent on understanding the legal implications of generative AI systems must, of necessity, be prepared to learn about the technology underlying these systems. Fortunately, these six Stanford researchers—some in computer science and some in law (our own redoubtable Mark Lemley among them)—have provided an essential guide for intellectual property and technology law scholars to the development and deployment of these systems. The article explores the extent to which developers and deployers of generative AI systems may rely on fair use to justify their use of in-copyright works as training data and how developers may limit their potential liability for infringements at the output stage.

For many copyright scholars, the article’s discussion of the fair use cases will be familiar, but the application of these precedents in the context of generative AI will be particularly useful. Yes, of course, the Authors Guild v. Google and iParadigms decisions suggest that computational uses of in-copyright materials can be fair use, but other decisions such as Associated Press v. Meltwater and Fox v. TVEyes suggest that much will depend on the particular uses that generative AI systems make of the in-copyright materials.

Foundation Models is not an advocacy article asserting that all uses of in-copyright works (or at least all that can be found on the open internet) as training data is fair use. Nor does the article argue that all outputs should be non-infringing so long as the outputs are not verbatim copies of the contents of specific training data. It offers a much more nuanced perspective about the challenges for system developers in understanding how to model computationally the degree of “transformativeness” that may be achieved by a second comer’s use of copyrighted works, as well as how to distinguish facts and expressions within those works.

The most novel section of Foundation Models is its discussion of technical strategies that AI system developers can employ to reduce the risk of copyright infringement when generative AI produces outputs in response to user prompts. These include data and output filters to detect similarities between the input data and outputs generated by the systems. Some technical mitigation strategies the authors describe must be done at the training data stage, while others, including data and output filters, can be done at the deployment stage.

The article discusses the Field v. Google decision for its recognition that Field had not used the “robots.txt” exclusion standard as a technique to stop Google from webcrawling his site. This consideration weighed against Field’s copyright claim that the search engine infringed by copying his content on that site. Foundation Models suggests that generative AI system developers and deployers would be well-advised to adopt one or more technical mitigation strategies to bolster their fair use claims.

While this article is well worth reading on the merits, it is also a noteworthy contribution to an emerging literature in which computer scientists and lawyers collaborate to explore technology law and policy issues. While not written in perhaps the most scintillating prose, this article is an outstanding example of a successful collaboration to explore ways in which technologists and lawyers can work together to co-evolve practical ways to achieve socially desirable outcomes.

Can Neuroscience Fix Trademark Surveys?

Jun 7, 2023 Christopher J. BuccafuscoAdd a Comment

Christopher J. Buccafusco

In trademark litigation, consumer surveys can determine a number of important doctrinal questions, including the most important one: whether consumers are likely to be confused into thinking that the defendant’s product was made or licensed by the plaintiff. Recently, scholars have questioned the validity and reliability of standard trademark surveys, suggesting that they are easy to manipulate and biased in favor of one party or the other. Wouldn’t it be great, then, if there was a reliable way to determine whether a survey was biased or not? Using neuroscientific imaging, an interdisciplinary group of researchers (including law professor, Mark Bartholomew) has proposed just such a possibility in a new paper, From Scanner to Court: A Neuroscientifically Informed “Reasonable Person” Test of Trademark Infringement.

Trademark surveys can suffer from a number of flaws. They may be explicitly biased in favor of one party or another, for example, by describing the defendant as a “copycat” or the plaintiff as a “trademark bully.” They may exhibit more subtle biases in how they frame questions about similarity and confusion. And, finally, survey participants always know the nature of the survey they are taking, so participants may exhibit “demand effects,” providing what they anticipate are the surveyor’s desired answers rather than their true responses.

The researchers began by demonstrating the manipulability of survey instruments. With two different plaintiff products, they manipulated the survey language to create both “pro-plaintiff” and “pro-defendant” surveys, as well as a putatively “neutral” survey. For each plaintiff product, they included a list of other products that varied in their degree of apparent similarity. For example, Reese’s Peanut Butter Cups served as the plaintiff product in one group, and the other products included Toffee Crisp (an actual defendant in litigation brought by Reese’s) as well as less similar products like Snickers, Justin’s, and Ghirardelli. When they tested a group of participants recruited from Amazon Mechanical Turk, the biases had the expected effect. Participants thought the defendant’s product (Toffee Crisp) was more similar to the plaintiff’s (Reese’s) in the “pro-plaintiff” survey than in the “pro-defendant” survey, and the “neutral” survey produced intermediate results.

To generate a more objective measure of product similarity that does not rely on participant reports, the researchers exploited an intriguing feature of human perception and cognition known as “repetition suppression.” The idea is simple: when we are presented with a stimulus that is very similar to one that we have just seen, our perceptual response to it diminishes. This is effectively a visual heuristic. Having seen something once, our brains devote less cognitive capacity to seeing it the second time. The empirical strategy, then, compares the degree of neural diminution across various stimuli to objectively measure stimuli similarity. The more similar two stimuli are, the more participants’ neural responses to the second stimulus will be diminished.

The study entailed functional magnetic resonance imaging (fMRI) scans of 26 participants who viewed the various products described above. fMRI scanning measures relative changes in brain blood levels as a proxy for neural activity. By focusing on regions of the brain known to process visual stimuli, the researchers could measure the degree of diminished neural activity associated with the Reese’s-Toffee Crisp pair compared to the Reese’s-Snickers pair.

When the researchers compared the neural similarity measures detected by fMRI to the self-reported similarity measures from the prior surveys, they found a strong correlation between the fMRI data and the “neutral” survey but no significant correlations between the fMRI data and either the “pro-plaintiff” or “pro-defendant” surveys. This suggests that the “neutral” survey is, in fact, a good proxy for participants’ actual experiences of visual similarity.

To be clear, the researchers’ methodological contribution isn’t to suggest that all trademark cases should require incredibly expensive neuroscientific studies. Rather, by using techniques like this one, scholars can develop a set of “best practices” or “gold standards” for trademark survey research. The goal is to use neuroscience to validate much cheaper behavioral surveys.

Of course, this is just the beginning. There is much that the current study doesn’t tell us. It can tell us that participants think the overall visual impression of certain trade dress is more or less similar to other trade dress. But it cannot tell us, for example, whether the participants were paying attention only to the protectible aspects of the trade dress or not. Nor do we know if some degree of similarity is consistent with consumers being confused as to source. The Mercedes logo and a “peace” sign look very similar, but people may not be confused by them. But this study seems a step in the right direction, and I’m excited to see where this kind of research will go next.

The Complex Effects of Piracy on the Movie Industry

May 9, 2023 Christopher J. SprigmanAdd a Comment

Christopher J. Sprigman

Does piracy of creative goods such as movies, books, or songs reduce paid demand for those goods? This seemingly straightforward question has proven surprisingly difficult to answer in the real world.

Piracy may draw away customers who might otherwise have paid. But it’s also possible that consumers of pirated copies are, by and large, not people who would have paid to consume if they couldn’t get access for free. Piracy may also help spread the word about a good movie, book, or song. This sort of informal advertising might drive up paid consumption, even if some people who would have paid are lost to piracy. It’s also possible that some combination of all these things might happen, with uncertain net results.

In a new empirical paper, titled Avengers Assemble! When Digital Piracy Increases Box Office Demand, Klaus Ackermann, Wendy A. Bradley, and Jack Francis Cameron offer a nuanced and interesting study of the effects of piracy on the movie industry. The effects of piracy, as it turns out, have a lot to do with what kind of movies we’re talking about.

The authors built a novel dataset that identified the existence and the timing of the earliest upload of a high-quality pirated copy for every U.S. movie release. The authors did this with data on the appearance of movie piracy “torrents” between January 2004 and January 2020 from online piracy site The Pirate Bay (TPB). The authors then matched this data with movie release information during the same period from the well-known IMDb database.

Merging these two streams of data allowed the authors to match up release dates and “piracy dates.” They measured changes in box-office revenue for pirated movies in the first 48 days after their releases in theaters, relative to the preceding period following releases before the movies were pirated. They then used a formula to adjust for the general fall-off in movie box-office revenues over time. If piracy was substituting for paid demand, the authors could pick up that effect by comparing (time-adjusted) pre-piracy vs. post-piracy box office revenues across many films.

The authors hypothesized that piracy has different effects on different types of movies. Specifically, they theorized that “spectacle” movies—the kind of movies that people want to see in the movie theater—may be less affected by piracy than “story” movies that people are more content to watch on their computers. In other words, spectacle movies may benefit more from word-of-mouth advertising that piracy may provide while losing fewer customers to demand substitution, compared to story movies.

To aid this assessment, the authors constructed two measures of movie “spectacleness.” One used a movie’s release in 3D or IMAX formats as a proxy for that quality (because “spectacle” movies are the kinds of movies that people want to see in these especially immersive formats). A second categorized movies into genres associated with spectacleness and story-focus by measuring the number of movies in various genres nominated for the “best visual effects” Oscar (associated with spectacleness) as opposed to the “best original screenplay” Oscar (associated with story-focus).

Based on data for more than 400 movies, about half of which have been pirated within the first 48 days of release, the authors concluded that piracy had the mixed effects they predicted. For films for which in-theater viewing adds value (“spectacle” films), there is a 13% increase in average daily box office revenue after the appearance of a high-quality pirated version of the film online. For story-focused films, on the other hand, there is as much as a 30% decline in average daily box-office returns after the appearance of a high-quality pirated version of the film online. This is consistent with the idea that piracy acts as a substitute to films focused on story, where the full value of the film can be consumed at home.

The authors’ findings shouldn’t be too surprising. Think for a moment about the music industry. Recorded music is more vulnerable to piracy than live music because a big part of the appeal of live music is the immediacy and communal experience of the concert. Such experiences cannot be replicated in a pirated recording.

So we might expect that during the post-Napster but pre-Spotify/Apple Music era when online piracy was driving down revenues for recorded music, there would be an industry shift toward more focus on live music. There was indeed a very rapid growth in that period of big live music firms such as Ticketmaster and Live Nation. Moreover, during that period the rise of live music revenue very closely mirrored the decline of recorded music.

In 2000 (just after Napster’s debut), recorded music represented 53% of the global music industry. By 2017 (when paid streaming started to restore lost recorded music revenues), recorded music’s share of total music industry revenues had dropped to 38%, while live music went from 33% to 43% of the industry.

Something analogous is happening in the motion picture industry, although the effect is probably not as pronounced. That is, the industry’s product mix may have shifted toward “spectacle” films because these sorts of film tend to be more resistant to piracy. Indeed, the authors gesture in this direction, stating that their findings suggest because the value of a film is linked to its “spectacleness,” the industry would be wise to adjust its creative output on the margin—i.e., to produce more spectacle films—to blunt piracy’s effect rather than investing in the law enforcement efforts that would be required to reduce piracy by any substantial amount.

Alternatively, movie studios may seek to insulate story films against piracy by, for example, releasing them to streaming channels simultaneously with theatrical release. Or, maybe movie studios could invest directly in upgrading theaters for these story-focused films to enhance the in-theater viewing experience in other ways, such as by making the theater a place for fun and social interaction. (Theaters such as the Alamo Drafthouse are already offering this kind of experience).

If so, then the principal effect of movie piracy may not be to lower the overall demand for movies or the number of movies produced. It may be to shift the kind of movies produced, or, more subtly, to shift the way that movies are presented to the public. Unlike the relatively simple framework in which piracy leads to fewer movies, the real effect of piracy may be more subtle, and the case for investing significant resources (especially public resources) in anti-piracy efforts less clear.

Policy Experimentation to Address Inequality Among Innovators

Apr 4, 2023 Lisa Larrimore OuelletteAdd a Comment

Lisa Larrimore Ouellette

Inequality among innovators is a substantial social problem in terms of both equity and economic growth. For instance, Raj Chetty’s Opportunity Insights group has documented that if women, racial minorities, and low-income Americans invented at the same rate as high-income white men, then the rate of U.S. patenting would quadruple. They also note the glacial progress toward closing these gaps, such as the 118 years it will take to reach gender parity at the current rate.

These inequalities affect not only the rate of innovation, but also what kind of innovations are created—for example, all-female inventor teams are more likely to focus on women’s health. Unfortunately, the evidence base for policy interventions to reduce these innovation gaps remains depressingly shallow. Most policies are tested without a rigorous evaluation strategy or control group, making it difficult to determine whether they had any effect.

A new paper from the U.S. Patent and Trademark Office (USPTO), Closing the Gender Gap in Patenting: Evidence from a Randomized Control Trial at the USPTO, is a remarkable addition to this literature. For the first time ever, the USPTO has tested a policy intervention as a randomized experiment, allowing a credible evaluation of its effectiveness. Changes in patent policy have rarely been tested with any element of randomization and have never been tested previously by the USPTO itself. Even if this experiment had yielded null results, the effort would still have been laudable as a model for how agencies can assess the impact of a new policy and publicly disclose the results. But the experiment also documents that the intervention—a new program to help patent applicants without legal representation—led to a sizeable decrease in the gender patenting gap.

The USPTO’s experiment began in 2014, when it created a new “Pro Se Pilot Examination Unit” to help pro se inventors (those without professional assistance) through the patent examination process. Obtaining a patent is not a user-friendly process, with most patent applications receiving a “rejection” or even a “final rejection” (which is actually more akin to a “revise and resubmit”) before eventually being allowed.

One study suggests that around half of the patent gender gap is due to women being more likely to abandon their patent applications after these discouraging replies rather than persisting in this back-and-forth process with the patent office. To address a concern that pro se inventors may be particularly disadvantaged in this process—for reasons unrelated to the merits of their inventions—patent examiners in the Pro Se Pilot received training on strategies to assist these inventors. For example, examiners would encourage applicants to call them with questions and would proactively help applicants draft better patent claims.

In the same way that promising new medicines are rigorously tested in randomized controlled trials that assign patients to either the new treatment or a control group, the USPTO decided to test this new examination unit by randomly assigning pro se applicants to either the Pro Se Pilot or to the regular examination process. By comparing outcomes across the two groups, they found that the Pro Se Pilot increased the likelihood of receiving a patent for all pro se applicants, and that it had a particularly striking effect for women. The likelihood of receiving a patent increased by 6.1 percentage points for men compared with 16.8 percentage points for women. The gender effect was even larger among first-time U.S. applicants: the likelihood of receiving a patent increased by 5.8 percentage points for men and a remarkable 23.5 percentage points for women. These results provide strong causal evidence of the new program’s value in closing the patent gender gap for pro se applicants.

Of course, this intervention is only one small step toward addressing the innovation gender gap more broadly. Future research should investigate whether similar changes in examiner training could help reduce the patent gender gap for broader groups of applicants. Less than 1% of U.S. patent applicants are pro se, but the additional guidance provided through the Pro Se Pilot might also help a larger group of inventors, such as those at small and micro entities who are currently disadvantaged by lower-quality legal representation.

In addition, the USPTO should study whether the reduced gender gap persists beyond patenting. Receiving a patent is worth little in isolation; financially benefiting from patents depends on other institutions with their own gender biases, such as corporate rent-sharing and venture capital. The Pro Se Pilot increased the likelihood that a pro se applicant would receive a patent, but it is worth examining longer-term outcomes such as assignments of these patents, new patent applications from these inventors, and non-patent outcomes gathered by survey or by linking to other datasets.

But the need for further research should not detract from the monumental nature of this study, which has simultaneously tackled two problems of bipartisan interest: inequality among innovators, and the need for better evidence to improve government effectiveness. In 2018, President Trump signed the SUCCESS Act of 2018, which tasked the USPTO with studying and recommending solutions to the problem of inequality among innovators. And the USPTO’s current Learning Agenda—developed pursuant to the Evidence Act of 2018—commits the agency to develop evidence on how to improve the effectiveness of patent examination in general, and with assessing participation in the patent system by underserved populations. The success of the first randomized controlled trial run by the USPTO on both of these fronts will hopefully inspire the use of rigorous experiments to test other policy interventions, both within and outside the patent context.

Do Sex and Drug Associations Tarnish or Burnish Trademark Reputations?

Feb 23, 2023 Sarah FackrellAdd a Comment

Sarah Fackrell

Federal trademark law now protects certain marks against “tarnishment.” If a mark is associated with “bad” things such as drugs or sex, the theory goes, that may harm the seller’s reputation and dilute the mark’s “commercial magnetism.” The theory sounds plausible enough, in theory. But what if that is not how it works in practice?

In Trademark Tarnishmyths, Linford, Sevier, and Willis add to the growing literature that empirically tests this theory of trademark tarnishment. The authors conducted two experiments in which famous marks were associated with sex, drugs, or sacrilege. The authors assert that theirs is “the first test of whether drug-related and sacrilegious uses tarnish appropriated marks, in two separate studies.” (P. 9.)

In the first study, “subjects were shown images of target marks used to sell cannabis products or in off-color, sexual contexts.” (P. 1.) They found that, instead of decreasing the desirability of the target marks, these exposures actually increased or “burnished” the desirability of those marks.

In the second study, “subjects were shown banner ads with cannabis-infused Skittles and satanic-themed Sunday sales of Chick-fil-A sandwiches.” (P. 2.) They found that highly religious respondents “reported Chick-fil-A was less tasty” after being exposed to the sacrilegious imagery. But the same was not true for cannabis; “conservative respondents exposed to the drug-related stimuli reported Skittles were more wholesome compared to the control – another burnishment effect.” (P. 2.)

The authors conclude that “the case for tarnishment might be weak in circumstances where courts have been most willing to presume tarnishment occurs”—i.e., when a famous mark is associated with drugs or sex. (P. 2.)

Insofar as these studies cast doubt on the conventional theory on tarnishment, this has some important implications for trademark law and policy.

First, as the authors note, to the extent their results are generalizable, “courts should require some evidence of likely tarnishing effect before granting relief on a claim of dilution via tarnishment” instead of just assuming such an effect. (P. 9.)

Second, the authors suggest that “anti-tarnishment protection might well be due for congressional reevaluation or vulnerable to constitutional challenge on First Amendment grounds.” (P. 9.) Claims of trademark dilution, and tarnishment in particular, have always been in considerable tension with the First Amendment, as other scholars have noted. If the types of uses that courts have, to date, treated as presumptively tarnishing don’t actually tarnish—and may even burnish—the mark’s reputation, it’s hard to justify this (relatively recent) federal cause of action.

In light of expanding legalization of marijuana at the state level, I wonder whether cannabis would be viewed as “bad” by enough people to make it the best testing reference. After all, older cases dealt with things like cocaine. But considering the public reactions to the October 6 “Statement from President Biden on Marijuana Reform,” it appears that at least some politicians continue to think—or at least, think that their voters think—cannabis is, in fact, a seriously bad drug. It would be interesting to see what would happen if these types of experiments were run using other drugs, such as methamphetamine.

Overall, this is a very valuable addition to the trademark dilution literature, and I highly recommend it.

How Do Patents Influence Cumulative Innovation?

Jan 24, 2023 Laura Pedraza-FariñaAdd a Comment

Laura Pedraza-Fariña

Ask any patent law student why we have a patent system, and they are likely to answer that patent law addresses a fundamental market failure: the free-riding by non-inventors on the inventions of others. A patent holder’s right to exclude others from making and using her patented invention addresses free-riding directly, restoring ex ante incentives to invest in innovation. But in solving the free-riding problem, patents create a second-order problem—one that is inextricably linked to the dynamics of innovation itself. Because all knowledge, and therefore all innovation, is cumulative, patents make innovations that build upon a patented feature more costly for parties other than the inventor, who must license an invention if they are to build upon it.

The problem of “follow-on” innovation has long preoccupied both economists and legal scholars. In their excellent paper, Measuring Follow-On Innovation, Janet Freilich and Sepehr Shahshahani contribute to this debate by bringing together both a deep understanding of patent law doctrine and precise econometrics research. In so doing, they make important contributions not only to the empirical literature, but also to our current theoretical thinking about the impact of patents on follow-on innovation.

Freilich and Shahshahani’s key empirical contribution is a refinement of the measure to capture the universe of follow-on innovations that are plausibly influenced by a patent. Two dimensions of patent law doctrine impact follow-on innovation. The first, patent breadth, determines how wide a net a specific patent casts—the more expansive the metes and bounds of a patent entitlement, the more follow-on innovations it will ensnare. The second, the collection of patent infringement doctrines, determines which types of activities in fact encroach upon the patent right. Their paper largely cabins the issue of patent breadth by focusing exclusively on measures of patent infringement.

Their article’s core claim is that existing empirical analyses of the impact of patents on follow-on innovation include activities that are non-infringing—and therefore, not the type of “patent related” follow-on innovation that those studies attempt to measure. The effect of this incorrect inclusion is large; in fact, Freilich and Shahshahani conclude that “little follow-on innovation is directly affected by the patent, with implications for theorizing the tradeoff between initial and follow-on innovation.” (P. 4.)

Freilich and Shahshahani focus largely on replicating and refining results in one important paper by Bhaven Sampat and Heidi Williams. In that study, Sampat and Williams measure follow-on innovation on patented and unpatented genes, using creative strategies to eliminate selection biases and concerns about claim scope.³ Freilich and Shahshahani’s refinement consists of eliminating non-infringing activities from the pool of follow-on innovations. Replicating Sampat and Williams’ methodology, they rely on articles published in a scientific journal that discuss research on the patented genes as their measure of follow-on innovation. Specifically, their refinement removes from the pool of follow-on innovations those scientific publications that fall under the following well-defined non-infringement categories: (1) extraterritorial activities; (2) activities by persons affiliated with the state government; (3) activities by persons affiliated with the federal government; (4) activities generating information for submission to the Food and Drug Administration (FDA) under the 35 U.S.C. § 271(e)(1) safe harbor provision; (5) and using or importing results generated from a patented technique.

Sampat and Williams found a small but statistically significant effect of gene patents on follow-on innovation. In light of other studies on gene patents,⁴ one might have expected that this paper’s correction would have uncovered a larger impact of gene patents on follow-on innovation. But the opposite was in fact the case. While Sampat and Williams found that gene patents had a small negative effect on follow-on innovation, Freilich and Shahshahani’s refinement found an even smaller effect. In short, with the new correction, patenting genes has an almost negligible impact on follow-on innovation.

Their refinement is important in no small part because it requires deep knowledge of patent infringement doctrines, including understanding which infringement doctrines produce predictable results ex ante that are unlikely to lead to litigation. This last point is critical: because the universe of follow-on innovations should include any invention for which a license would reasonably be required, their measure should only exclude those innovations for which we can be certain that a license would not have been sought ex ante. The importance of identifying clearly non-infringing activities also puts pressure on their empirical methodology. In this regard, one particular infringement area—the safe-harbor provision under 271(e)(1)—is worth discussing further.

The experimental use exception proves to be an important filter in their corrected measurement, eliminating 388 publications from the original pool of 2,771 follow-on publications. The 271(e)(1) safe harbor provision is meant to exempt from infringement activities “reasonably related to the development and submission of information”⁵ to the FDA. Courts have interpreted this provision broadly to include many drug-development activities, including early-stage research. Recent caselaw, as Freilich and Shahshahani point out, excludes two key types of activities from this safe harbor: basic scientific research and patented technology that is not itself subject to federally regulated approval (also known as the research tools exclusion).

To identify publications that fall under the 271(e)(1) exception, the authors use a proxy: whether the publication discussed a potential therapeutic application. This proxy is cleverly designed but both potentially under- and over-inclusive, as the authors recognize. Some exempt research projects may not explicitly address therapeutic applications. On the other hand, it is plausible that scientific publications engaging only in basic scientific research mention therapeutic applications for the purpose of seeming more attractive to funders or publishing venues. The research tool exclusion could also lead to over-inclusivity concerns. Because correctly identifying research that is exempt under 271(e)(1) is likely to be crucial in future studies of follow-on innovation in the biomedical sciences, further refining this measure would be a fruitful avenue of future research.

Freilich and Shahshahani next turn to exploring how their refinement might help reconcile other studies on follow-on innovation—many of which found larger effects of patents on downstream innovation. They begin with an explanation rooted in sociological factors. University researchers, who are likely overrepresented in the sample of follow-on innovations measured by published scientific papers, may be largely unaffected by details of patent law doctrine. This effect can be explained by a robust underlying social norm that fosters the free sharing of information and materials, and a disregard for patent entitlements. The situation might be quite different for communities embedded in biotech start-ups and established pharma companies, where uncertainties about the content of 271(e) and doctrines designed to police claim breadth may lead to choices of research projects away from those that require the use of patented materials.

Their explanation reveals another interesting mechanism worth exploring in future studies on follow-on innovation: the impact of patents is likely to be mediated by communities’ social norms. This makes it crucial to understand those social norms and the likely distinct mechanisms by which communities of university researchers, biotechnology startups, and established pharmaceutical companies (among others) decide which follow-on research projects to pursue and which ones to abandon.

Finally, Freilich and Shahshahani consider how their measure may inform existing theories of innovation. As they recognize, “the proper measure of follow-on innovation depends on the hypothesized mechanism through which a patent might affect downstream innovation.” (P. 29.) Two quite different theoretical perspectives on the role of patents in follow-on innovation have emerged from the literature. The first one conceptualizes patents as increasing incentives for patent holders to invest in follow-on innovation. Under this theory, patents do not diminish follow-on innovation; rather, patents tend to concentrate follow-on research in the hands of fewer patent-holding inventors. In contrast, the second theory emphasizes the access costs that patents impose on follow-on innovators other than the patent holder.

Freilich and Shahshahani’s combination of detailed legal analysis with econometric research opens the door for testing a number of additional questions about the mechanisms of follow-on innovation. For example, one interesting question that emerges from these two theories is whether innovation by multiple parties is likely to be qualitatively different from concentrated innovation by a few patent-holding pioneers. Here, it would be interesting to further parse the authors’ refinement to test whether patents influence the kinds of follow-on innovation that take place. More specifically, do patents change the balance between incremental and breakthrough innovation? Network theories of innovation suggest that breakthrough innovation benefits from knowledge recombination across firm boundaries. By increasing access costs outside the firm, patenting may lead to more incremental and less breakthrough follow-on innovation.

A single paper cannot settle the debate between these two theories of follow-on innovation. But by emphasizing the importance of infringement doctrines in measuring follow-on innovation, and by showing how this measure modifies the experimental results of several prior articles, the authors move us towards a more precise answer, an answer that is likely to vary by industry and to impact basic and applied research in different ways.

1. Bhaven Sampat & Heidi L. Williams, How Do Patents Affect Follow-On Innovation? Evidence from the Human Genome, 109 Am. Econ. Rev. 203 (2019).
2. Fiona Murray & Scott Stern, Do Formal Intellectual Property Rights Hinder the Free Flow of Scientific Knowledge?: An Empirical Test of the Anti-Commons Hypothesis, 63 J. Econ. Behav. & Org. 648 (2007).
3. 35 U.S.C. § 271(e)(1).