I have just posted Speech In, Speech Out, one of several scholarly responses included as part of Ronald K.L. Collins and David M. Skover’s new book, Robotica. The book is their take on how the First Amendment will adapt to an age of robots. To quote from the publisher’s description:
In every era of communications technology – whether print, radio, television, or Internet – some form of government censorship follows to regulate the medium and its messages. Today we are seeing the phenomenon of ‘machine speech’ enhanced by the development of sophisticated artificial intelligence. Ronald K. L. Collins and David M. Skover argue that the First Amendment must provide defenses and justifications for covering and protecting robotic expression. It is irrelevant that a robot is not human and cannot have intentions; what matters is that a human experiences robotic speech as meaningful. This is the constitutional recognition of ‘intentionless free speech’ at the interface of the robot and receiver. …
And here is the abstract to my response in Speech In, Speech Out:
Collins and Skover make a two-step argument about “whether and why First Amendment coverage given to traditional forms of speech should be extended to the data processed and transmitted by robots.” First, they assert (based on reader-response literary criticism) that free speech theory can be “intentionless”: what matters is a listener’s experience of meaning rather than a speaker’s intentions. Second, they conclude that therefore utility will become the new First Amendment norm.
The premise is right, but the conclusion does not follow. Sometimes robotic transmissions are speech and sometimes they aren’t, so the proper question is not “whether and why?” but “when?” Collins and Skover are right that listeners’ experiences can substitute for speakers’ intentions, and in a technological age this will often be a more principled basis for grounding speech claims. But robotic “speech” can be useful for reasons that are not closely linked to listeners’ experiences, and in these cases their proposed “norm of utility” is not really a free speech norm.
Robotica also includes Collins and Skover’s reply to the commenters. In the portion of their reply that discusses Speech In, Speech Out, Collins and Skover say that I have misunderstood their argument by conflating First Amendment coverage (what qualifies as “speech”) and First Amendment protection (what “speech” the government must allow). Intentionless free speech (IFS), discussed in Part II, is a coverage test. Only once it is resolved does the norm of utility, discussed in Part III, take the stage to answer protection questions. Their example about a “robotrader” whose algorithmic buy and sell orders qualify as covered speech comes near the end of Part II, so when I use it to criticize the norm of utility, I am mixing up two distinct pieces of their argument.
This reply is clarifying, though probably not in the way that they intend. They are right that I misunderstood the structure of their argument. But in my defense, Collins and Skover misunderstand it too.
Speech In, Speech Out applies the norm of utility to coverage questions because that is how Robotica applies it. For the robotrader’s messages to other robots to qualify as covered speech, one needs a theory that is not just intentionless but also to some extent receiverless. Collins and Skover claim that there is covered speech when robots exchange information “at the behest of and in the service of human objectives.” (46) But this is an appeal to the purpose and value of the message, not its meaning. If there is a nexus to the reception of speech by a human listener here, it is coming from the norm of utility, not IFS. So if my response anachronistically smuggles Part III’s discussion of protection and the norm of utility back into Part II’s discussion of coverage and IFS, I am only following their lead. The “[c]onfusion over this dichotomy” (113) they perceive in my response is just a look in the mirror.
Although I stand by the criticisms in Speech In, Speech Out, I regret the process that produced it. The principal reason that Collins and Skover and I have spent so long stumbling around in this scholarly hall of mirrors is that our exchange has taken place entirely through completed manuscripts: they wrote their book, we commenters wrote our responses, they wrote their reply. Questions for each other were off the table. They picked the format to promote “[d]ialogic engagement” (111), and there is indeed more dialogue here than in a standard monograph. But compared with the lively, back-and-forth, interactive qualities of a good workshop, conversation, email chain, or even blog-comment thread, one and a half rounds of manuscript exchange aren’t much at all. Fixed-for-all-time publications are a good way to memorialize one’s best understanding of a subject, but there are better ways to converge more quickly on mutual understanding in the first place. In the future, I will look more skeptically on invitations to take part in scholarly projects where ex parte contacts are off-limits.