ARTIFICIAL MEANING(fn*)
Lawrence B. Solum(fn**)
INTRODUCTION: FROM ARTIFICIAL INTELLIGENCE TO ARTIFICIAL MEANING
As artificial intelligences (AI) become more powerful and pervasive, communication by, with, and among AIs has become a common feature of everyday life. Early in the history of AI, there was ELIZA-a simple program that utilized simple pattern-matching algorithms to simulate a psychotherapist interacting with the user of the program.(fn1) Human communication with AIs has been depicted in film and fiction, from the iconic confrontation of humans with HAL in 2001: A Space Odyssey(fn2) to the very human Theodore who falls in love with an artificially intelligent operating system named Samantha in Her.(fn3) But complex communication with AIs is part of everyday life in modern technological societies, including Apple's Siri, automated telephonic service systems for airlines, online ordering systems for merchants like Amazon.com, and characters in video games who converse with human players. All of these contexts involve artificial meaning, which we can contrast with communications by natural persons (human beings)-which we can call natural meaning.
This Essay investigates the concept of artificial meaning, meanings produced by entities other than individual natural persons. That investigation begins in Part I with a preliminary inquiry in the meaning of "meaning," in which the concept of meaning is disambiguated. The relevant sense of "meaning" for the purpose of this inquiry is captured by the idea of communicative content, although the phrase "linguistic meaning" is also a rough equivalent. Part II presents a thought experiment, The Chinese Intersection, which investigates the creation of artificial meaning produced by an AI that creates legal rules for the regulation of a hyper-complex conflux of transportation systems. The implications of the thought experiment are explored in Part III, which sketches a theory of the production of communicative content by AI. Part IV returns to The Chinese Intersection, but Version 2.0 involves a twist-after a technological collapse, the AI is replaced by humans engaged in massive collaboration to duplicate the functions of the complex processes that had formerly governed the flow of automotive, bicycle, light-rail, and pedestrian traffic. The second thought experiment leads in Part V to an investigation of the production of artificial meaning by group agents-artificial persons constituted by rules that govern the interaction of natural persons. The payoff of the investigation is presented in Part VI. The communicative content created by group agents like constitutional conventions, legislatures, and teams of lawyers that draft complex transactional documents is artificial meaning, which can be contrasted with natural meaning-the communicative content of those exceptional legal texts that are produced by a single individual. This insight is key to any theory of the interpretation and construction of legal texts. A conclusion provides a speculative meditation on the implications of the new theory of artificial meaning for some of the great debates in legal theory.
I. MEANING: WHAT AND HOW?
First the what, then the how. We begin with a general inquiry into the meaning of "meaning."
A. What Is Meaning?
This Essay is about meaning, but what does "meaning" mean? The word is ambiguous, as illustrated by the following examples:* Do you think that those clouds mean it is going to rain?* I didn't mean to hurt your feelings.* Do you know how much you mean to me?* I got the letter translated! Now, I know what it means. The ambiguity of "meaning" afflicts the law as well. When we ask about the meaning of a constitutional provision, statute, rule, regulation, or judicial opinion, we could be asking about the kind of meaning that translators provide-roughly linguistic meaning, but more precisely communicative content. But we might be asking about the implications of a text for legal doctrine or for a particular case. Or we might be asking about the purpose or function of the text. These are all distinct senses of "meaning," and before we proceed further, we need to sort them out. There are at least four senses of "meaning" relevant to legal texts:
Communicative Meaning. Suppose someone asks a question about what the Framers meant by using the phrase "arms" in the Second Amendment: were they referring to weapons or the upper limbs of the human body? This sense of meaning refers to communicative content- the content conveyed by the legal text given the context in which it was authored. In this Essay, I will use the word "meaning" to refer to this sense of meaning, and for greater clarity I will sometimes use the phrase "communicative meaning."
Legal Meaning. The communicative content of legal text is not the same thing as the legal content associated with the text. For example, the communicative content of the phrase "freedom of speech" is quite sparse, but the legal content associated with the phrase is very rich indeed: it encompasses a complex set of doctrines, addressing topics like prior restraint, defamation of public officials, and even the regulation of billboards. We can call meaning in this sense "legal meaning."
Application Meaning. Sometimes when we ask about the "meaning" of a legal text, we are asking about the implications that it will have, usually in a particular context. For example, we might ask, what does the First Amendment freedom of speech mean for my defamation suit? Does it provide me a defense? When "meaning" is used in this sense in the context of a legal text, we are concerned with the application of the text to a particular case or to some set of cases. Let us call this "application meaning."
Purposive Meaning. Meaning is also used to refer to the purpose or motive that produced a particular legal text. For example, we might refer to the aim of a constitutional provision by saying, "What did the drafters mean to accomplish through the Privileges or Immunities Clause of the Fourteenth Amendment?" We can call this "purposive meaning."
So, there are four senses of "meaning" in legal contexts. Our focus will be on communicative meaning, but before we begin to explore artificial meaning we need to say something about how communicative content is produced.
B. How Is Meaning Produced?
Theories of meaning are the terrain of the philosophy of language and theoretical linguistics, but in a short essay, even a short and dirty tour is impossible. Instead, we will examine one important approach to theorizing about theories of meaning. That approach is associated with philosopher Paul Grice, but this is a simplified version that differs from his account in several ways.(fn4) We can begin with Grice's idea of speaker's meaning,(fn5) although we will primarily be concerned with legal texts, so our version of the idea could more precisely be called "drafter's meaning."
We can begin with a definition: Speaker's Meaning: The speaker's meaning (or utterer's meaning) of an utterance is the illocutionary uptake that the speaker intended to produce in the audience on the basis of the audience's recognition of the speaker's intention.
Quite a mouthful! Grice's idea is quite general; it can be illustrated with an example in which there is neither speech nor text, but communication occurs nonetheless:Imagine that you have stopped at night at an intersection. The driver of another car flashes her lights at you, and you make the inference the reason for her doing this is that she wants to cause you to believe that your lights are not on. And based on this inference, you now do, in fact, realize that your lights are not on.(fn6) Communication succeeds because the other driver recognizes your communicative intention. In the language of game theory, you and the other driver have common knowledge of this communicative intention. Of course, drivers recognize many communicative intentions when they are on public roads and highways. In the United States, octagonal red "stop" signs communicate a legal command, roughly "halt before proceeding."
Here is another example, from the California Driver Handbook:(fn7)
Traffic signs are a particular kind of legal communication. They convey communicative content, and some of them (stop signs, for example) have associated legal content. If you run a stop sign, you can get a ticket for an offense punished by a fine.
In conversation, speaker's meaning can rely on a rich communicative context. If Ben and Alice are having a conversation at a coffee house, there are multiple channels for producing uptake-the recognition of communicative intent. Imagine the following dialog: Alice: Another?Ben: Black this time.Alice: Sugar? [rising inflection] Ben [shaking his head from left to right]: Uhhh. Alice: OK.
The communicative content is far richer than the literal meaning of Alice and Ben's utterances. "Another" meant "would you like another cup of coffee." "Black this time" meant "I would like the coffee without milk or cream." "Uhhh" plus the shake meant "I would like the coffee without sugar." If we take away our knowledge of context, the same words might mean something totally different. Try your hand at inventing different contexts for the same words and you will see what I mean.
Some legal communication occurs in information-rich contexts. Face-to-face negotiation of...
