Dan Burk, University of California, Irvine School of Law
Racial Bias in Algorithmic IP
Unpacking bias: divergent meanings: statistical bias (sampling),
design (wrong type of model, model created for one purpose used for another), the
fact that “raw data” is an oxymoron, social bias (inappropriate social outcome—different
than the bias that designers often talk about, so they may talk past each other
even if there’s overlap). This can make some responses beside the point:
transparency, auditing, “human in the loop”—those are mostly attempts at better
technical accuracy. That’s a red herring. Example from criminal justice: parole
determinations. Zip code is a major factor in these, and that’s correlated with
poverty and race. It’s a really good predictor, but that’s the wrong question:
why do we put up with that connection/digital redlining? An AI patent examiner
that rejects/narrow claims based on race of inventor, that’s not inaccurate to
past practice. It’s wrong in the sense of being immoral and [supposedly]
socially unacceptable.
Machine bias may be the wrong question. Let’s deal with the
bias in the system; it doesn’t matter whether it comes from a human or from an
AI trained on the past practices of humans. But there are differences in
practice: Illusion of objectivity; humans put too much faith in numbers, Donna
Haraway’s “God trick” in which the human is no longer visible in the picture. E.g.
overweighting of numerical evaluations. Algorithmic performativity: they enact
what they assume and create their own social facts. The yearly ranking of law
schools by USNWR. Wendy Chun suggests: we should use these systems as
diagnostics. We don’t think a weather forecast is diagnostic or that it can be
used to create better weather; we can use AIs to find out what our biases are.
Ann Bartow, University of New Hampshire Franklin Pierce
School of Law
Ruth Bader Ginsburg’s Copyright Jurisprudence (with Ryan Vacca)
Strong copyright and liberal politics. As a DC Circuit judge
and SCt Justice, she authored sixteen opinions in © cases, ten majority, four
concurrences, two dissent, and joined 11 others, ten majority, five of which
were unanimous, and one concurrance. Mostly owner-favorable and Goliath over
David, which may seem surprising. Five opinions favor authors: Tasini, Eldred,
Golan, Reid (WFH), and arguably Muchnick, the followup to Tasini. Many were in
favor of owners: Petrella, Nat’l Cable TV Ass’n v. Copyright Royalty Tribunal,
Atari v. Oman (2x). Concurred in Grokster & Star Athletica; fiery dissent
in Kirtsaeng. But did favor accused infringers sometimes: Fourth Estate v.
Wall-Street.com & OddzOn v. Oman as well as concurring in Quality King. She
joined Campbell, Fogerty v. Fantasy, Feltner, and several others.
Unpersuasive explanations: widely known love of opera and
art (other Justices have those too); Jane Ginsburg’s uncontested expertise (Ginsburg
fille isn’t particularly highly cited by Ginsburg J. and Breyer cited her against
a Ginsburg opinion).
Better: Incrementalism—if judges reach unnecessary issues,
this creates instability/undue stress on judiciary. Evident in gender equality
work and critique of reproductive rights litigation based on lack of incrementalism.
Intergovernmental deference: evolution/interpretation of law shouldn’t be
diatribe against Congress, President, admin agencies, or states, but as dialogue.
Courts have instiuttional capacity constraints and must do this to be effective.
Lily Ledbetter Fair Pay Act after loss at Scotus; Petrella opinion recounts
back and forth with Congress and courts. Atari v. Oman: Sent it back to Office
twice and told them to look at Feist.
Eldred and institutional capacity: worried about whether the
courts were the right institutions to decide duration. How does Breyer know his
economic analysis is right? [Framed as “where is his economics degree from? And
Lemley in chat says “Oxford.”]
Some divergence in © jurisprudence: seemed to favor
alternative remedies in ©--In Tasini, suggested court shouldn’t issue
injunction preventing inclusion of the disputed articles, but instead compulsory
licenses/consent decrees. She was much more skeptical of alternative remedies,
like a lesser military school for women instead of VMI, for gender
discrimination.
Ryan Vacca: responding to chat point that citing Jane Ginsburg
isn’t a great proxy for being influenced by her, and they acknowledge this.
Dalindyebo Shabalala, University of Dayton School of Law
Solomon Linda, Traditional Knowledge Pirate? Mbube, “The
Lion Sleeps Tonight” and Traditional Knowledge
Challenging the originality of Linda’s work as potential cultural
threat. The story: popularized in the 1950s including by Pete Seeger from a pop
music record in South Africa. That song, story goes, was composed by Solomon
Linda, who died penniless while millions were made. 2000 article resurrected
the story of misappropriation and injustice; led to a suit against Disney.
Settlement elided the basic question of who really created it. He argues it
should be tested in court because it’s a key issue in TK: What happens when a
traditional piece of music gets translated, derived from, built on, recorded,
and moved into the © system? The classic TK story is an outsider coming in and
taking it into the © system. Story of Linda is, he suggests, a much more
traditional trajectory: an insider who traverses the boundary of the community
to the outside and acts as the translator, then lays claim themself to the TK.
Musicologically the origin story is more complicated, and the question of what
rights insiders should get is much more complicated than misappropriation by
record companies.
Linda was an originator of a style. Urban music/dance, built
on tradition of rural Zulu style. He took a wedding song sung by village girls.
Developed in a sharing/building/borrowing culture. How should that affect what
we think the world owes Linda? Community rules may continue to bind insiders
when they try to exploit works outside those communities. We should not think
of this as a new problem. © has always been bound up with nationalism. The
problem to be solved was misappropriation across borders. The problem for
traditional communities is very much the same. Should think about solutions for
insiders who flee.
Bita Amani, Queen’s University Faculty of Law
Law, Race, and Alchemy: Exclusion(s), Existential Crises,
and the Transformative Possibilities of Intellectual Properties
Operationalizing privilege through law presented as formally
equal. Curricula/cultural materials are an important part of naturalizing
privilege. “Exclusivity” of rights: IPRs are part of broader colonial legal
regimes, so why would we expect anything other than disparate impact in rights
acquisition and enforcement? Authorship in fact v. authorship in law is gendered/raced;
work v. play distinction means some of us are disenfranchised from labor/means
of production while others become owners.
Carys Craig, Osgoode Hall Law School, York University (with
Anupriya Dhonchak)
A Feminist Theory of Moral Rights: Creative Agency and
Voices from the Margin
Not surprising that feminism would have something to say
about a right conceived of as intensely personal. Dichotomies: public/personal;
economic/moral; rational/emotional; masculine/feminized; valued/devalued: so
moral rights are the feminized Other. Personhood rationales have therefore
tried to masculinize/propertize the justifications. The image of the artist is
that of the solitary male genius, individualized author. Hegel, proponent of
moral rights, held that women were not capable of art. Kant likewise thought
that knowledge unfitted women for their places. Not clear why a feminist would start
from Kant and Hegel! Moral rights rhetoric is also filled with references to
paternity; patriarchal metaphors of birth without women, right to control
offspring. We need a better ontology of authorship.
A feminist relational theory of authorship locates
creativity in cultural situations and social relationsh; creative capacity is
enabled by relational web. Relational autonomy, capacity to make meaning is the
foundation of authorial rights. Authorship as dialogue/relationships of
communication. Every text is multivocal; for feminists, this invites
exploration and activating of the unvoiced, exiled world of women (Mary O’Connor).
Could say that integrity right ensures dialogue is a real
one, but we think that clings to the romantic idea of there being a single
meaning bestowed on work by author, and dialogism refutes that at the core. Risk
of weaponizing moral rights against those who critique the dominant culture. Integrity
right misundersands relationship b/t author, text and public in way that casts
disruptive dialogic engagement as moral and legal wrong; not consistent
w/feminist politics of confrontation, resistance, and social reform. Example: Fearless
Girl attacked as violation of moral rights of bull sculpture’s sculptor.
But: attribution right deserves attention through feminist
lens. Tracing who is speaking, from where, on behalf of whom can be about
feminism and authorship. De-Kanting acknowledgement: Kant thinks that the harm
is compelled speech, but feminist theory is that the harm is the silencing, the
refusal to acknowledge that someone has spoken. The power to make knowledge
claims v. the people who have been erased from/made invisible in our narratives.
Attribution is a call against erasure.
Caveat: it doesn’t follow that © is the right space to
achieve political goal of amplifying marginal voices given the bluntness of
legal tools and power needed to wield them. Empowering communities of practice
is the goal. Moral rights don’t occupy a moral high ground; they’re based on
the same patterns of exclusion and control as ©, but can benefit from feminist
reeimagining.
Moderator: J. Glynn Lunney, Texas A&M University School
of Law
Lemley: AI does expose some fundamental contradictions in
what we think of as inequality: a broader problem of how we measure equality
and how we want to balance procedure and substance. Making that explicit can be
useful.
For Bartow: Don’t shy away from the harm she did in throwing
out Sony in her Grokster concurrence, Petrella and its authorization of a wave
of © lawsuits forever, etc. There are interesting agency/courts/etc. metrics to
all of these, it is still worth noting that she ends up on the side of the ©
owner and he doesn’t think that’s a coincidence.
Burk: all human tech may be prosthetic: cars are feet
prosthetics; AIs may be cognitive prosthetics, but bring all human baggage.
Amplifying and disclosing may be the useful parts—using them as diagnostics to
figure out where we’ve been screwing up. But they are our messes to solve.
Bartow: yes, and that’s painful about this project.
Betsy Rosenblatt: For Shabalala: Norms are good at governing
in communities, but they are really bad at governing outside of communities.
Natural takeaway might not be that everyone owns TK but that no one does, but
that may be imposing a norm on the group from outside. Maybe that’s exactly
where we need laws (when it moves outside the community) but then whose rules
should we adopt as the law?
Shabalala: This is the key part of the project. The premise
has to be cultural sovereignty: the only communities with a strong claim have
to build on a preexisting political sovereignty, b/c the right to regulate
citizens’ lives is built on sovereignty. Rosenblatt says that’s a very American
way of looking at sovereignty, but Shabalala responds that African nations have
wholeheartedly adopted it, although there are sub-sovereign, sub-metropolitan
communities that are clearly left out of this formulation, and that is a real
problem. Reaching outside the community is required to regulate exploitation,
but TK may not be the answer.
Amani: Note that in the US the definition of an Indian comes
from the federal government; the problem is complementarity of structural
inequality that is imposed.
RT: for Craig: Is this a right/interest of the author or a
right/interest of the audience? Consider anonymity; women writing as men; whites
writing as natives.
No comments:
Post a Comment