Least Cost Expectation. ⢠Factual determination ... A complete poem if less than 250 words and if printed on not more than two pages or from a longer poem, ...
Algorithmic Fair Use Dan L. Burk University of California, Irvine 2017-18 Fulbright Cybersecurity Scholar
Project Goals • Extend “Infrastructure” Work (Julie Cohen) • Connect to Broader Literature • Technology changes • Humans/human reactions don’t
• Push Back on “Techno-utopianism” • Case Study for Algorithmic Law • Re-focus Analysis • Not whether fair use can be automated • Effect on the law from attempting automation
Copyright Law • Exclusive Rights • • • •
Expressive works Vested in author Transferable Pecuniary incentive
Algorithmic Enforcement • Across Business and Legal Domains • Digital Copyright • • • •
Filtering Policing Notice and Takedown “Fingerprint” Matching
• De Facto Arbiter • Asymmetric costs • Altered Expectations
Exceptions and User Privileges • Fair Dealing • Laundry list • Discrete exceptions • Closed canon
• Fair Use • • • •
Balancing test Four factors Notoriously indeterminate Fact-specific
Rules and Standards • Classes of Legal Imperatives • Definite and discrete • Indefinite and open • Contextual or static
• Reciprocal advantages and disadvantages • Flexibility versus certainty • Enforcement and compliance • Ex ante or ex post designation
“Crystals and Mud” • Modulation in Law • Exceptions or per se shortcuts • Procession between two extremes • Continuous rather than discrete
• Least Cost Expectation • • • •
Factual determination General or contextual Institutional competence Legislative or adjudicatory
Good Faith Belief of Infringement “We note, without passing judgment, that the implementation of computer algorithms appears to be a valid and good faith middle ground for processing a plethora of content while still meeting the DMCA's requirements to somehow consider fair use.” -- Lenz v. Universal (9th Cir. 2015) (passage withdrawn)
• Big Data/AI Alternative • Machines that learn, not machines that think • Patterns/correlations in massive data sets • Self-defining
The Magic Worldview “Human decision makers are flawed and biased. The biases and inconsistencies found in individual judgments can largely be washed away using advanced data analytics. . . . Even if a machine-produced law is not perfectly unbiased, as long as it is less biased than a law produced by individual humans, the result will be net beneficial.”
Relevance of Algorithms (Gillespie, 2014) • Patterns of Inclusion • Cycles of Anticipation • Evaluation of Relevance • Illusion of Objectivity • Patterns of Entanglement • Production of Calculated Publics
Inherent Data Biases • “Raw Data is an Oxymoron” • Data is always “cooked” • Useless otherwise
• Data Curation • Formatting • Selection • Availability
Patterns of Entanglement • Reflecting Back • Social negotiation • Recursive feedback
• Algorithmic Recognition • Hidden and visible
• User Habituation • Accomodating implicit logics • Self-affirming logics • Reshaping user practices
1976 Fair Use Guidelines A complete poem if less than 250 words and if printed on not more than two pages or from a longer poem, an excerpt of not more than 250 words.
Either a complete article, story or essay of less than 2,500 words, or an excerpt from any prose work of not more than 1,000 words or 10% of the work, whichever is less, but in any event a minimum of 500 words. (Each of the numerical limits stated in "i" and "ii" above may be expanded to permit the completion of an unfinished line of a poem or of an unfinished prose paragraph.)
Calculated Publics • Fair Use Guidelines • Popular adoption • Judicial adoption