Automating the Uncertain Judge
Courtney M. Cox
Ignoring uncertainty is wrong. An ideal judge aims at proper
outcomes, and if they ignore uncertainty they ensure error, sometimes serious
error. Ignoring strength of belief won’t always result in error but sometimes
it will. Rule-driven and data-driven models both need to grapple with
uncertainty. Jurisprudential tools & doctrines exist, along with rational methods
for coping with uncertainty. More research is needed.
Q: can a data-driven machine be a legal realist? If the idea
is to take into account sufficient social context, the training data is the beginning
but not the end of decisions, won’t it converge on rule-based formalism?
A: Probably a machine can respond to whatever it had for
breakfast. Underlying the Q are the difficulties in the data: what judges had
for breakfast causing variations is baked in, which is commonly discussed in
the context of bias. We at least have a sense of what we want to control for or
what they’re doing wrong, but work has been started on that; here, with
uncertainty, there’s not a lot of work figuring out when a judge is doing
something for jurisprudential reasons or because they’re uncertain about which
method of textualism to use.
Silbey: What’s the difference between uncertainty and highly
contextual analysis?
A: Probably related.
Q: why not program different versions, one that tries to
minimize transaction costs, one that tries for legal realism, etc. and pit one
against the others?
Some discussion; she showed a diagram about judges who were
committed to different theories of jurisprudence (e.g. originalism, pure
natural law) with different degrees of confidence, and used that to show a
model of how those judges would evaluate overruling Plessy/deciding Brown
as it was decided. I didn’t really end up convinced that this was doable or
descriptive of uncertainty in judging at the jurisprudential level as opposed
to the factual level.
Schedule A Defendants
Eric Goldman
Complaint doesn’t list actual defendants; schedule A. Often
sealed; often contents are Chinese manufacturers who are alleged counterfeiters.
Similarity to Doe defendants—one advantages is plaintiff can use the revelation
of identities of defendants as leverage to get settlement, as in the porn/©
cases. Another option: list them in the caption; making joinder concerns more
obvious. No confidentiality of Ds’ identities—defendants can coordinate; asset freezes
may not work.
Location by state, per Bloomberg Law: ~3000 in Illinois, ~500
in NY, 374 NJ, long tail. About 3000 default judgments, 688 voluntary/joint
dismissal, 44 admin dismissal, 33 trial/judgment entered, 15 contested
dismissal, 15 summary judgment, 2 other. Where unsealed, about 218 defendants
per case, meaning that Emojico sued over 10,000 defendants in one year. Most of
the cases are TM (3000).
Concerns: Robo-pleading; problems with service/how do we know
it’s working properly; personal jurisdiction—we know that harm in jurisdiction is
not enough, and addresses may not be listed; joinder/filing fee issues—how are
the parties related enough to all be on the same complaint? Very few judges
raised it on their own. Sealed defendant
lists: a red flag b/c we can’t track what’s going on in courts that way—the rejoinder
is that we need to seize their assets, but we don’t get unsealing even when
there is no concern/seizure is accomplished. Emojico dismissed defendants who
showed up to fight back. All this is ex parte. Judges used to adversary process
can allow lots of errors to creep in if the other side doesn’t show up.
The point is not to litigate, but to get the court order to
take to a place like Amazon, which will shut down the entire business and not just
the targeted conduct. The judge doesn’t know this and things look like they’re
going pretty well to the judge.
What do we do? Judicial education? FRCP?
Emojico targets people selling mugs that have emoji on them advertised
as “Emoji Gifts Birthday Present”—rubber-stamped as TM infringement every step
of the way in court and by Amazon. This is worth fighting back against.
Lisa Ramsey: Lifeguard licensing also claims to own rights
in “Lifeguard” and cross. When someone fought back, they dropped the case and
the court denied an attorneys’ fee request. So we need something to be done.
Lemley: Defendant class actions do exist—there are
protections for the class, judges have to sign off, could be expanded.
A: loves that idea b/c it would force evaluation of
commonality between the parties, which is worth looking for.
Felix Wu: how much do substantive claims matter? Troll usage
is bad, but can imagine legitimate usage by mass infringers.
A: needs to consider more—horrified by the procedural aspects.
The Class Action as Licensing and Reform Device
Xiyin Tang
Google Books settlement was rejected; quasi-legislative
provisions were more for Congress. Have there been other settlements in class
actions that tried to obtain a release not just for past harms but for forward-looking
uses. Relatively few 1938-1990: 7. Three filed by the same person, head of
ASCAP; another performance rights; composers. From 1990-2000, 22 class actions.
Tech was a driver. Many were attempting to address new tech uses. Cahn v. Sony,
SDNY 1990, was about audio tapes; settlement wasn’t monetary but promise among
settling parties to pursue subsequent legislation, resulting in Audio Home
Recording Act. Frank Music v. Compuserve, SDNY 1993, for liability for ISP for
infringement by users. First settlement in copyright class action providing for
future royalties. Pre-512; Compuserve testified in Congress multiple times about
the need for safe harbors, and their class action experience alerted them to
the uncertainty in law/need for legislative resolution.
Ferrick v. Spotify, SDNY 2018: licensing the long tail of
songwriters/publishers; claim was that they shouldn’t have used the NOI
compulsory licensing process, esp given that Copyright Office couldn’t keep up
with the volume—work by work process wasn’t tenable for music streaming;
settlement provided a lot of framework for Music Modernization Act.
Mass tort settlement as regulation: global peace. Here,
though, the rights aren’t necessarily recognized in substantive law, which
leads to the filings in the first place—the class actions are gap-filling. Will
continue as new tech uses challenge ©.
Q: CAFA?
A: Proceduralists worry about Ps getting too little and ©
folks worry about Ps getting too much. But courts are supposed to scrutinize
settlements more carefully, which is significant.
Sprigman: Public info: Spotify relied on Harry Fox early on,
before NOI—Harry Fox would match what it could. Over time they would rematch
and pay out arrears for new matches. The plaintiffs said paying arrears wasn’t
allowed. What’s the reaction? Harry Fox lost its place—it’s not the central player
now that it was then. Statutory damages made this messy: damages were either
very payable or astronomical, and it was very difficult to know the range.
Another source of uncertainty: there was a claim that for normal streaming (no
copy) there was no digitial phonorecord delivery, which meant there was no license
needed, which created uncertainty on the other side. So both parties had an
incentive to get the process fixed and NOIs rationalized.
A: HFA is owned by the music publishers so the fact that
they could not actually get it together to figure out who owned what was also
significant.
The Solicitor General’s Mixed Record of Success Before the
Supreme Court in Copyright Cases Pam Samuelson
Started w/significant contrast b/t SG’s brief in GvO and the
outcome. Likewise the SG supported Georgia against PublicResource.org. Compared
to the patent side the SG does not have a great record of success on substantive
© issues. OSG has won two cases for the US (Eldred & Golan) & lost one
(Dowling). Justices have more confidence in how they thought about © for themselves,
especially with Breyer and Ginsburg. SG
filed only 2 amicus briefs in 20th c private litigant cases (CCNV, Quality
King), but amicus in all but 2 of 21st century cases—pretty big
shift. Procedure/remedies, SG analysis or close prevailed in 5 of 6 cases. Substantive:
didn’t agree with SG 9 times. In 4 cases agreed with SG about who should prevail
but not about why, different/narrower. 4 cases: disagreed with both who and
why. One 4-4 split in Costco v. Omega, where SG failed to persuade SCt to rule
in Omega’s favor. Also did not prevail in 2 remedies cases. Where the
divergence was only on analysis, not outcome: Grokster, Aereo. Divergence, but
not huge: CCNV, Star Athletica, Kirtsaeng II (atty fees).
In only 2 of 18 majority opinions did majority cite
approvingly to SG arguments—it was Ginsburg. In 3, explicitly criticized SG: Quality
King, Kirtsaeng I, and Star Athletica. But mostly just ignored SG, especially
in substantive interpretation cases. Contrast w/other studies of SG influence
and citation rates.
On average, SG espoused high protectionist views. One
reason: growing influence of Copyright Office lawyers on SG briefs. CO lawyers
are on 5 substantive briefs and 3 procedure/remedies cases, and also advise the
SG.
Why ignore SG? SCt likes its own decisions, including older
ones like Bobbs-Merrill.
When should the SG file in private cases? Not as often,
absent CVSG or direct implications for CO procedures in registration—Star Athletica
or Fourth Estate—or effects on foreign relations/border enforcement—Quality King/Kirtsaeng.
Constitutional challenges (though they didn’t file in Felner or Allen v. Cooper).
But they shouldn’t file in substantive cases like GvO, Warhol v. Goldsmith.
They too often ignore the public interest.
Dan Burk: increased unwillingness to defer to the executive,
which might be a bad thing overall.
A: note that the CO is not an executive agency, which should
complicate things. And it stayed out in the 20th c. IPEC, USTR, and ©
group in the USPTO all reinforce the CO’s high protectionism, so it would be
good if they reached out to other entities. No real federal interest.
Burk: EPA’s views on how statute should work should be
weighed even in private litigation.
Silbey: cynical explanation—where they practiced before?
Posts
No comments:
Post a Comment