The Journal of Things We Like (Lots)
Select Page
David E. Pozen, Eric L. Talley & Julian Nyarko, A Computational Analysis of Constitutional Polarization, 105 Cornell L. Rev. 1 (2019).

The moment is at hand. No longer self-consciously experimental, computational analysis now comes to constitutional law in an ambitious effort pursued by David Pozen, prominent in the field at Columbia; his Van Halen-like colleague Eric Talley, known especially perhaps for his law and economics theorizing and his corporate and contract law investigations – and Julian Nyarko, newly arrived at Stanford as a practicing computationalist interested first of all in contract law. Their project takes up the question of polarization and its pertinence constitutionally, an issue that has become more prominent in the past few years.

The effort at one level is straightforward. It turns out that all remarks made by members of Congress, on both the House and Senate side, dating from 1873 to 2016, now exist in a collected, computer analyzable form. What might we learn if we read all these speeches? We can’t, of course – too much to know, too much information to acquire and think through, even if we read quickly rather than carefully. Machines read too, assimilating many more documents much faster. But they read their way – sorting words, counting uses, noticing conjoint and disjoint patternings, and the like. We need to know what we know, therefore, when we read machine-reading results. Pozen, Talley, and Nyarko show us how they assembled their mechanism and the decisions they had to make in order to ready their computer for work. These decisions become a sort of pedigree.

Word frequency lists as such are not the main focus. Instead, the idea is first to check legislator vocabularies against constitutional dictionaries. These last are collections of words that count as “constitutional” terms. The word “constitution” itself and its immediate modifications like “unconstitutional” and “nonconstitutional” generate a first dictionary. Words included in the U.S. Constitution fill up a second collection. Words seemingly part of ordinary discussions of immediately constitutional terms become a third grouping. The computer is instructed to search out, accumulate, and report incidences of these constitutional usages in congressional talk.

Because the speakers’ party affiliations are known, the computer can determine within given periods whether members of one party or the other are more prone to constitutional usages, and whether given usages vary or not with party membership. Pozen, Talley, and Nyarko put together training lists of constitutional usages in recent periods in order to suggest versions of “right” and “left” groupings, tuning the lists to match very well with Republican and Democratic affiliations of speakers. These groupings are made available to the computer to predict the party affiliations of congressional members within the larger sets. This in turn allows the computer to assess degrees of orthodoxy among party members.

All this effort (and more – including comparisons with Wall Street Journal and New York Times editorial usages), all parts carefully designed, executed, and reported, allow the study to state findings concerning constitutional polarization. Especially provocatively, it appears, both Republicans and Democrats made frequent use of constitutional terms in the most recent period (not so in other ranges). But the terms differed considerably depending on party affiliation. This may be evidence, the authors suggest, of strong constitutional polarization. It is as though two ships “Peerless” pass in the night (as Professors Talley and Nyarko might have put it).

This is too terse. It is clear nonetheless that the work here is very good indeed. What should we make of it? Should we want to read exercises of this sort? Assemble them ourselves? “Might as well jump?” The great value, I think, is old-fashioned. Observing closely what this effort involves, we are driven quickly from initial skepticisms to thinking about helpful differentiations and elaborations.

First, we notice that constitutional usage is reduced to piles of words. Sorting words is good work for computers. But we “not-computers” tend to believe we proceed otherwise. Of course, “we” may not be a homogeneous set. Are members of Congress engaged in the same efforts as judges or lawyers or academics? Consider: Sometimes constitutional terms — words, sentences, passage, entire documents – are considered closely, as elements in efforts to think intensely and independently about seemingly difficult and important matters. On other occasions, constitutional terms are treated as “doctrine” – standard formulas framed as rules and exceptions or levels of scrutiny or test steps – in order maybe to achieve institutional uniformity or integrity. In other instances, terms are pretty much emoji. They communicate attitude or intensity, mark normative postures, stand up or stand down (we might say).

All three modes are sometimes apt. But in different settings each might be more or less salient. We might imagine members of Congress who think hard and originally about basic constitutional questions. Regrettably, John C. Calhoun comes to mind. Bruce Ackerman suggested that the remarks of Hubert Humphrey and Everett Dirksen in the course of congressional discussion of the Civil Rights Act of 1964 should rank high. Often, it is easy to suppose, legislators proceed within established doctrinal fences. Or—maybe even more frequently—they use constitutional references and resonances to signal stances, to evoke ideological allegiances and intensities – to cast “emojis” as though fly fishing. This last use is formally a kind of mirroring or borrowing, or maybe translation, of attitudes or commitments or aversions already in play in society at large. These captures may be politically sharply relevant – and if reused often enough may become constitutional content by association. Effective trademarking maybe, purposefully derivative, these constitutional cartoons are in themselves neither analyses nor close “doctrinal” maps. But none of these three modes are, in any general way, inappropriate.

Second, we also readily recognize that polarization might be pictured from a dramatically different angle. It is not surprising if political parties understand their own agendas in differing, sharply distinct terms. We remember 1800 and its very hard hardball. “Popular sovereignty” and “freedom national” were at bottom opposing constitutional notions in the 1850s, ultimately fiercely held. Cataclysm and constitutional re-grounding followed, we know, unresolved controversies were renewed within new terms, fought in new ways even now.

It is also plausible to think that some polarizations involve refusals to recognize new constitutional understandings as constitutional at all. The dramatic devices of the sit-in sparked civil rights movement often found only awkward constitutional protection, and opponents (notably, otherwise low-key Lewis Powell) denounced that movement as anti-legal and best forgotten. We remember John Lewis well as an individual, bravely hard-headed. But he is less often celebrated as one participant among many other animators of a remarkable democratic leviathan.

We know all this, of course. We worry whether we ought to be able to take steps – discover our inner Henry Clay, as it were – to manage constitutional crises in advance of full out collision. Pozen, Talley, and Nyarko offer us a diagnostic device. The groupings of constitutional dictionary entries they are able to discern are accumulations of instances. These gatherings might be useful preliminaries if we try to impute interior thematics. This is work on our part, not the computer’s. These thematics are the jurisprudential equivalents of literary criticism, plainly not themselves computational, but rather re-visioning computational heaps as harbingers, maybe pointing to Dworkin-like enterprises in the end.

Against this backdrop, we remember Eric Talley earlier on. Writing with Ian Ayres, Talley restated the notion of Solomonic bargains as a question of dividing entitlements, suggesting that such divisions ex ante — more precisely, our responses to them – sometimes worked later to facilitate improving rearrangements.1 The beginnings of bargaining begin in the disorganization of law. Arbitrageurs read jurisprudence. This article continues to reward close study (not just snap-shot sloganeering) – maybe especially as constitutional analysis one-step disguised. For present purposes: Sometimes as we judge congressional constitutional polarizations, we may search out possible cut constitutional or other legal entitlements as building blocks and begin to sketch bargaining paths as responses. Computational mechanics looks to be part of the beginning stuff of law itself, a first lens in recognizing structural breaks and bends and thus potential rapprochements. Sometimes, anyway. In other instances, the cut entitlement will look to be profound, its remedy not necessarily some sutured recombination.

Not only congressional remarks: constitutions, statutes, judicial opinions, all the rest – these are just words, aggregates and arrangements open to study. Computational analytics frame the window, plot paths into and around or through legal complexities we otherwise see as dense to the point of darkness. Pozen, Talley, and Nyarko touch down on law’s documentary substrate. This is law-at-first they are writing.

Download PDF
  1. Ian Ayres & Eric Talley, Solomonic Bargaining: Dividing a Legal Entitlement to Facilitate Coasean Trade, 104 Yale L. J. 1027 (1995).
Cite as: Pat Gudridge, “I have seen the future…” “And it works?”, JOTWELL (January 20, 2021) (reviewing David E. Pozen, Eric L. Talley & Julian Nyarko, A Computational Analysis of Constitutional Polarization, 105 Cornell L. Rev. 1 (2019)),