H. Brian Holland, We Are All Cyborgs Now: A Cognitive Theory of the Third-Party Doctrine
Once upon a time there were categorical protectsion for private papers against search & seizure. Courts then ruled that personal property was no longer categorically exempt from search and seizure. Distinction b/t papers and other effects was broken down. There’s a warrant requirement, though. But an exception if the gov’t’s acts aren’t a “search” under the 4A. A search occurs when the gov’t violates a reasonable expectation of privacy. Third party doctrine comes in to show there’s no reasonable expectation of privacy. Voluntary assumption of the risk that a third party would disclose to the gov’t—gov’t needs only subpoena or similar process.
Sotomayor: Ill-suited to digital age, given our sharing of more information with more third parties. “Sharing” means something different—not just intentional tangible exchange. Not necessary to commercial relationship; now baked into basic tech. Personal papers stored on third party servers such as those of Dropbox—covered by 3d party doctrine even though there may be no person on the other end! Does not comport w/reasonable expectations of privacy.
What is to be done? First, personal papers are not mere effects, which are separately mentioned in the 4A. Worthy of special, if not categorical, protection. Specifically, 3d party doctrine should be narrowed to protect personal memory artifacts that are created, communicated, and stored on tech designed to structure, replace or augment human memory. Looking at cognition and memory, linking function to doctrine. Related to the idea of embodied cognition and the extended mind. These cognitive memory artifacts are part of our cognitive systems; should be protected b/c of relations b/t 1A, 4A, 5A—they play a role in freedom of thought and freedom of conscience. Linked to idea of security of persons. Bodily integrity linked to autonomy, dignity, and identity—the extended mind is also an extension of the body.
Consistent w/other theories; passes Orin Kerr’s test of being tech neutral in the sense that you don’t gain a benefit as a criminal by using digital tech.
Many people agree that the 3d party doctrine should be curtailed but we haven’t had a theory w/a judicial impact. My solution: reinvigorate “papers” as a distinct object of protection w/their own theory. Problem: b/c of the history, there’s little literature on the protection of papers under the 4A; most is focused on the place where the papers were found—the home—and the papers as property. But that’s not descriptively accurate any more. Should we look at scholarship on creativity?
RT: Skeptical that would help b/c it’s not clear that papers are different—thinking about Erving Goffman & Peggy Radin, where it’s possible that other property will be equally important to personhood. Framers?
A: they said nothing about why they used both papers and effects! In England, the history is largely about searches for (seditious) pamphlets, which ended up being found in the home but were intended for broad distribution; in the US, the history is largely about searches for untaxed items/tax records.
Q: Recent case—police got pacemaker data—under 3d party doctrine. But if he wanted to get the information himself, DRM might have prevented him from doing that.
Stacey Dogan: what’s the line?
A: there is no absolute line—some people take pictures of everything they want to remember. I’d rather look at the 3d party instead of the particular use.
Dogan: but you may lose people who don’t believe that a song in your cloud storage is a personal appendage.
Amanda Levendowski, Fair Use for Fair Artificial Intelligence
Lots of scholarship on biased AI. There are lots of ways to introduce bias—the people creating the algorithms; the data selection; data classification. IP scholarship hasn’t addressed how © can channel AI. E.g., training facial recognition algorithms on white faces. Amazon’s same-day delivery rollout didn’t include Roxbury, predominantly black area—result that people in that area didn’t get the same access. ProPublica’s investigation of machine learning algorithm for reoffending after exiting the prison system; African-Americans were assigned higher risks for lesser crimes. This could change bail, sentencing, parole for the worse. Likelihood of arrest was related to overcriminalization/overpolicing in predominantly black areas. Word mapping: Google News showed gendered patterns. Neutral: king is to man as queen is to woman. Man is to computer programmer as woman is to homemaker—the algorithm picked up bias by reading human works. Training an AI on the Enron emails is going to be based on the habits of a bunch of white oil & gas guys from Texas. Latanya Sweeney: people searching “black” names aren’t going to get good job ads and are likely to get “search for a criminal record” ads instead. This isn’t how we should train our algorithms.
Low-friction data (easy to acquire) are biased. Twitter firehose: API allows most tweets to be used in research. Userbase skews young, white, and urban—made Hurricane Sandy look like it was centered in Manhattan, not Rockaway where they lacked electricity. Wikipedia: largest CC source of data: fewer than 10% of active Wikipedia editors are women. If you’re using Wikipedia for low-risk natural languages training, you get the bias of the people who edit Wikipedia. Same thing happens with public domain books, and judicial opinions (largely written by old white guys w/degrees from Harvard and/or Yale).
Solution: Google Books. Messy ethical question: do we actually want our robots to be really good at recognizing people, or talking to us?
Q: Copyright Compendium’s argument that copying by a machine isn’t expressive at all?
A: Compendium’s focus is whether computers can author. It’s not copying, it’s scraping; there are different words for the same thing. But Judge Koh in Meltwater was very concerned about scraping that ignored robots.txt. It wasn’t a true CFAA case, but the idea of scraping can create issues with CFAA.
Alex Roberts: what’s the protected by copyright but unbiased corpus that could be used instead?
Q: works by feminist authors or those w/intersectional approach, post-1923. If you want a robot that doesn’t code homosexual as a slur, you need recent books.
Tiffany Li, Robots vs. Animals: Toward a Unified Theory on Intellectual Property Rights on Non-Human Creators
Elephants paint; the monkey selfie; computers or algorithms can create art or new discoveries. No consensus on applying IP principles to those two categories.
Why does this matter? Today’s science fiction is tomorrow’s science fact. What incentivizes non-humans? Should we try to do so? Do we only want to incentivize human creativity? Fairness theory: should trees have ©? Who determines who the creator was if there’s no human author? Personality theory: does a nonhuman have any moral rights?
Monkey selfie: is the author the person who pressed the button on the camera? Or the person aware of the work being created? David Post says pressing a button b/c it makes a funny sound isn’t authorship b/c authorship requires an awareness that a work is being created.
Algorithms don’t work the same way computer generated art does. Right now our computers are not aware of an act of creation. If we consider authorship as only something human, there’s no point in discussing this, but I’d argue that what creativity is shouldn’t be limited to humans—there will almost certainly be AI that can decide to create, or will create what looks like the same mode as human. Ultimate Q: what does it mean to create, produce, or author a work? Comparing computers to animals can lead to interesting conclusions about envisioning a photo v. clicking a button.
RT: Why care? Compare this question to 3D printing, which also to me raises “why care” questions insofar as the principles seem pretty similar even if the outputs are different.
Holland: does it matter whether it’s the animal incentivized by the click of the camera or the computer incentivized by its programming? If the point of © is more works, which is a big if, then would it matter whether it was a human?
Cathy Gellis, Testing the Limits of the Section 230 IP Exemption
Stealth provision from Sen. Wyden to counteract the rest of the CDA—to make sure that censorship wouldn’t be too easy.
Trademark, not part of the Progress Clause—why does it get shoved under the rubric of intellectual property? Paper tries to explore that. How do we figure out what IP is? Why do we think that, and can we push back on it? One reason this is important: lots of people use workarounds. Lots of people frame their problems as © problems; Google v. Garcia was not about the film’s distribution being harmful for reasons that had anything to do w/©. Registrars are willing to break anonymity too easily—anonymous speech is protected by the 1A. So she wants to get TM out of 230 to avoid evasion.
There are other things that go wrong. In the 9th Circuit, the IP exemption means federal IP; other circuits, and even California state law, include state based IP claims.
Dogan: Need to argue on two levels: if you’re making an argument that this statute should be interpreted to exclude TMs, you have to look at the legislative history to figure that out. Even though there’s been pushback against use of IP as a label more broadly, it’s undeniable that in common parlance lots of people include TMs as IP.
A: there doesn’t seem to be legislative history on this.
Dogan: can you fix this if © workaround still exists? You generally want there to be less liability for intermediaries and this is a way you think you can get there—more persuasive to give a principled reason.
A: one of these things is not like the others.
Silbey: since 230 was supposed to be easily operable immunity, the idea that it was exempting federal criminal law and IP should also be easily identifiable as such. One might think registered IP could be included. What is IP is too hard a question to answer, but which IP for easy on/off immunity switch might be easier. Could also be how people claim the IP, rather than preemption.