Communications of the ACM

such regulations pass muster under the First Amendment?

Admittedly, these applications are novel artifacts not envisioned by the founders.

First Amendment protections are not without exceptions; for instance, they do not authorize people to violate intellectual property or reputational rights. U.S. citizens are not free to engage in libel, copyright infringement, or sedition. Obscenity is only partially protected.

None of these exceptions apply to the expressive activity involved in automated legal-assistance systems.

Alleged misinformation or harmfulness is not viewed as justifying suppression of books, except in extreme circumstances. Government is not appropriately in the business of judging the quality or content of speech. A landmark casep held that distributing the 1965 book How To Avoid Probate did not constitute the unauthorized practice of law.

One may be inclined to suggest that some automated systems are a form of “commercial speech” and thus deserve less protection. Commercial speech has generally been understood as the activity of beckoning business, not the substantive content of what is being offered. Selling a book does not render it any less deserving of First Amendment protection than giving it away for free.

An alternative way to avoid First Amendment issues is to conclude that programs are not “speech” at all but a form of conduct, analogous to the work of manual document preparers. This involves distinguishing between “pure” speech and “speech plus” that entails actions, as well as words. Sometimes speech-related action is not protected if it is physically dangerous. Does such a dangerousness rationale extend to communicative action?

Several legal scholars have tentatively concluded for the unconstitutionality of repressing online legal services under the guise of the unauthorized practice of law.q The following sections lay out an analytical framework that may support more definitive conclusions.

A typology of expressions. Figure 1

outlines one way to organize the vari-eties of expression a legal self-helper might access; blue boxes are categories, and green boxes contain examples.

Expressive conduct falls into two main categories: creating artifacts, or works of authorship, and “ performing,” or engaging in live, real-time communication with others. Artifacts in turn are either static (with fixed content in a fixed order) or dynamic ( programmed to present different content in different orders depending on external triggers (such as a user’s behavior interacting with it). Performances fall into two high-level categories: those in which communication is unidirectional, or one-way, (such as speeches) and those in which communication is bidirectional (such as one-on-one and many-to-many conversations).

Some features apply to multiple branches of the Figure 1 tree:

˲ Most modes of expression can be through either physical or electronic means; for practical purposes, programmed content and social-media interaction can be accomplished only electronically;

˲Electronically mediated expression can happen offline or online; that is, via electronic networks (such as the Internet) and protocols (such as the Web);

˲ Artifacts can include charts, diagrams, tables, flowcharts, decision trees, and other graphical elements; such things can also be used in most forms of performative expression;