iSpeak: Automated Authorship and Accountability in the Digital Age
Posted: 13 Jul 2018
Date Written: February 5, 2018
Automated authors generate content that impacts nearly every part of civil society. Children watch automatically–generated YouTube videos to learn language and social skills. Corporations use automated authors to create logos, slogans, poems, artwork, and even entire books in a multibillion dollar industry. And federal courts increasingly rely on reports generated by automated authors in making life and death decisions for criminal defendants. Despite the growing ubiquity of automated authorship, courts and commentators have yet to successfully theorize attribution for works generated by automated authors. Can a work have a nonhuman author? If so, should the law treat its works differently than works created by humans? The majority of courts simply avoid the question; others simply assume that only humans can be authors; still others do not even recognize a question exists. The result is a dangerous jurisprudential blindspot that rests on unstated, untheorized, and unscrutinized assumptions about speech produced by automated authors.
This Article attempts to correct that problem by proposing the first comprehensive theory of automated authorship. While other scholars have analyzed particular forms of automated speech (such as reports generated by a computer) or particular areas of law (such as First Amendment law), none have offered a comprehensive theory of attribution for automated authorship. This Article fills that gap by creating an analytical framework of criteria to categorize automated authorship in one of three ways: fully attributable to the automated author, fully attributable to a human contributor, or partially attributable to both. For the latter case, this Article borrows from existing principles of copyright and patent law to articulate three novel theories of coauthorship — self-replicating statements, extended statements, and joint statements.
The automated authorship framework has significant implications for copyright law, constitutional law, and evidence law. For copyright law, the scope of automated authorship controls the scope of the copyright protection for an automated content industry that generates billions of dollars every year. This Article argues that copyright law should protect works created by automated authors as a means to encourage investment in improving the quality of works generated by automated authors. If the fruits of an automated author’s labor are subject to unlimited free-riding, firms have significantly diminished incentives to invest in developing intelligent automated authors in the first place. Second, automated authorship impacts both the First Amendment right to freedom of speech and the Sixth Amendment right to confront witnesses in criminal proceedings. Rather than focusing on whether automated authors deserve First Amendment protections (or require Sixth Amendment confrontation) as other scholars have done, this Article recenters the inquiry by asking when an automated author is speaking in the first place. Shifting the question from “what kind of computer speech is protected” to “when is a computer speaking” offers both descriptive and normative advantages in First and Sixth Amendment contexts. Finally, in an inquiry related to the Sixth Amendment question, this Article discusses whether works of automated authorship should be considered hearsay and thus presumptively subject to cross-examination for litigation purposes.
Keywords: automated authorship, digital evidence, software, computer programs, confrontation clause, copyright, author, artificial intelligence
Suggested Citation: Suggested Citation