How do we fight fake news? Promoting free speech can, at the same time, promulgate false speech and the more vigorously we protect free expression, the more we inadvertently permit deception. The problem is hard. Legislatures across Asia, North America, Europe, and Latin America grapple with it. It touches campaign finance reform, already a thorny issue. As a computer scientist, I argued frequently with my law professor father over ways to solve it. Many of the technological methods we might use to curb false speech run afoul of current law and the very idea of free speech. Yet, the space between law and technology is where we might find better answers.
The problem runs deeper than the technical challenges of machine learning. At one level, reducing Type I errors simply invites those of Type II while training one filter to recognize fake news simply invites adversaries to train other filters to write it.3,4 No, at a different level, courts question the desire to regulate fake news at all: they bar intervention in cases of politically protected speech. If fake news is political, it should not be regulated. From a legal theory perspective, there are elements of this policy that are wise—courts should avoid judging political truth—at the same time there are elements that are unwise—courts should dismantle systems that prevent us from hearing truth. At present, court decisions stifle competing truths and it is here that an old idea from computer science, the Church-Turing thesis, suggests the ban on some forms of intervention should be lifted. Computability theory has much to add to legal practice in the design of better systems.