Computational Intelligence and the Paradoxes of Legal Routine
Marc Lauritsen
(This is one of my favorite unfinished pieces, last touched in April 1990 when I was Director of Lawyering Information Systems at Harvard Law School. It later appeared in the widely read Publications 4/99 of the Judicial Academy of Northern Finland, University of Lapland.)
Summary: This article explores some of the complexities and paradoxes of routine activity in law, their implications for the possibility and desirability of certain modes of legal automation, and the ways in which perspectives suggested by theories in other fields may advance our understanding of these questions. Conjectures are made about the nature of legal routine that may have implications for artificial intelligence research in law.
I. Introduction
Routine activities in law — as in many other fields of human endeavor — seem ideal candidates for automation. By definition, they are repetitive and regular courses of behavior, entailing little conscious attention or exercise of creativity. Like the “routines” of computer programs, they seem predefined, mechanical, deterministic, and algorithmic. Lawyerly routines, for instance, make up the unremarkable patterns of everyday professional life, the tacit background against which the spontaneous, deliberative, unpredictable, and extraordinary figure. Likewise, the routine processing of business and family disputes obviates the need for official judicial attention.
In common conception, routine is the rule; irregularity is the exception. Routine is smooth; the non-routine is discontinuous. We are willing, and usually delighted, to entrust routine work (which is often boring, if not downright disreputable) to artificial devices. Humans have higher and better things to do with their natural intelligence.
There are senses of routine and related terms, however, in which these comfortable assessments lose their plausibility.
Routine is sometimes contrasted with formality, rationality, and legality. Routine may be patterned, but it is not necessarily governed by formal rules, nor derivable from axioms. Routine is often regularity without rules, order without an organizer. Routine can be substrate, emergent, sub-cognitive, sub-symbolic.
The experience of computer science has also been contrary to the intuitive equation of the everyday routine with the simple. The easiest problems to solve have turned out to include many that most humans find astonishingly difficult; common sense and everyday reasoning have been the most intractable. “[N]o amount of anticipation, planning, and programming can ever enumerate, a priori, all the variants of even a routine situation that may occur in daily life.”[1] “[A] computer cannot store enough information to approximate the vast amount of common sense knowledge that the average human utilizes daily.”[2]
So, is routine playing by the rules or not? Automatic or not? Formal or informal? Structured or unstructured? Systematic or not? Is it simple or complex? The way in which these antinomies break down, in which routine seems to straddle the poles of these supposed oppositions, in which this conceptual terrain seems to have become “loopified” [Hofstadter 1978, Kennedy 1987] parallels the erosion of time-honored distinctions in so many other fields: figure/ground in gestalt psychology and modern art; public/private and form/substance in contemporary legal theory; deterministic/chaotic in mathematics and physics. The resonances of these developments with aspects of the routine problematic cry out for a cross-disciplinary exploration, even at the risk of dilettantish superficiality.
Several related goals motivate this exploration: One is to improve our sense of what aspects of legal practice can and should be computer-assisted. How can we optimize the positive effects of computers as emancipatory dischargers of routine, rather than routinizers of humans? What are the social implications of increasing automation in the courts and lawyers’ offices? How might we head off dangerously oversimplified notions about the thresholds and limitations of computability in the legal arena?
More generally, what can law, computer science, and other disciplines dealing with dynamic systems tell each other about the paradoxical interplay of complexity and regularity? Can rules in an information system play a similar role to laws in a social system?[3] Might we simulate salient dimensions of social life in an artificial system — a virtual legal reality?
II. Needed: A Cognitive Jurisprudence of Legal Action
Most published research in AI and law has concerned itself with legal reasoning, as opposed to legal action; with the thinking rather than the doing of legal actors. Thinking, of course, is a particular kind of activity, and much of the discussion of legal argumentation straddles the blurry line between analytical consideration of arguments and their deployment in practical advocacy situations. But nonetheless there has been a dearth of illuminating discussion of practical lawyering activity from the standpoint of cognitive science and artificial intelligence.[4] The emphasis has been on ideas more than acts, concepts more than conduct.
One chapter in the cognitive science of legal action has got to be concerned with the nature and function of routine. And, routine or otherwise, intelligent behavior in real time raises fundamentally different questions from deliberative or analytical thought, and I attempt to characterize some of those differences.
III. Routine Words
An exercise in amateur etymology[5] yields a number of suggestive associations — and some surprising dissonances — among the words and concepts relating to routine. The English word routine derives from the French word route, which seems related (as illustrated by its modern cognate rote) to the Latin word rota and the Indoeuropean stem RET-, meaning run or roll. Yet it is typically defined in terms of regularity, which derives from the Indoeuropean stem REG-, meaning direct or straight, from whence also comes rule.
This jumble of notions of straightness and circularity[6] may seem less odd when one considers that a circle is a straight line bent at a consistent angle, and that a wheel tends to roll in a straight line. And rotation of an object about a center is, after all, a controlled, regular kind of behavior.[7]
These word pedigrees are of limited relevance. But the metaphors underlying the terms strongly influence our concepts [Lakoff and Johnson, 1980], and cataloging some of the usages is one way of excavating those metaphors. (A chart mapping approximately 100 words and phrases synonymous and antonymous with routine will be included in the full paper.)
IV. Computability and logic
Insofar as routine activity is characterized by rule-following, one is tempted to regard the concept of routine as co-extensive with that of algorithm, and thus with computability. As noted above, there is a lively presence of machine words in the terminology of routineness. We think of routine actors as being “on automatic pilot,” as acting mechanically, as being “programmed”. Robots are presumably good at routine.
It would be convenient if the routine turned out to be co-extensive with the computable. Rigorously developed theories of computational complexity, consistency, completeness, and effectiveness might then be directly applicable. Finitistic metamathematical procedures might be adduced to validate proposed formalizations of legal routines. [Boolos and Jeffrey, 1974]
The effort to equate routineness with computability founders, however, on several counts. Among other things, many undeniably routine activities ‑‑ such as using written and spoken language ‑‑ appear dependent upon forms of subcognition (e.g., pattern recognition) that seem unlikely to be Turing-computable. [Hofstadter 1985] Formalism is on the retreat in many quarters [Leith 1990], and it has little hope of finding refuge in routine.
V. Routine in the practice of law
Large systems of legal administration such as criminal justice or landlord-tenant relations tend to fall into routinized behavior. Routine processing of disputes most often takes the form of settlement. One manifestation of routineness in legal practice in these settings is the conceptual taxonomy of cases. Sudnow [1964], for instance, identifies the “normal” (as opposed to statutorily defined) crimes with which public defenders deal — proverbial characterizations, prototypes, recipes, that are specific to a given practice community (or ecosystem). “Offenses whose normal features are readily attended to are those which are routinely encountered in the courtroom.” (Id. at 261) Fluency in these normal crimes “constitutes the mark of any given attorney’s competence.” (Id.) “[The lawyer’s] grasp of that knowledge over the course of time is a key indication of his expertise.” (Id. at 262)
Defense “lawyer regulars” appear again in Blumberg’s The Practice of Law as Confidence Game. In the bureaucratic setting of routine criminal justice, the client becomes a means to other ends of the organization’s incumbents. “Assembly line justice” is the result. All significant participants are bound into an organized system of complicity. [Blumberg 1967, at 22.] Rational, impersonal elements involving economies of time govern. Routine here is equated with cooptation. Defendants are “processed.”
Similar patterns in the delivery of civil legal services are discussed in [Bellow 1977]. “Despite their commitment to avoid the kind of cautious, detached, client controlling service that so many public bureaucracies — public housing authorities, welfare departments — seem to provide, my guess is that, if one looked carefully, one would conclude that this is precisely the kind of service our clients are receiving.” (Id. at 108.) “Routine processing of cases” results in perfunctory service. Relationships with clients are dominated by these routines. The vast majority of cases result in settlements, the prevailing pattern of which generates its own expectations and makes higher settlements difficult to obtain. (Id. at 109.)
All of this occurs despite the legal aid attorney being sympathetic and concerned about the plight of clients. There is a sense that the system is unchangeable. Cases are handled superficially. Bellow discounts the “caseloads are just too high” explanation, and explains the bureaucratization of client service in terms of the system’s demands for accommodation, and the vulnerabilities of inexperience. There is a short step from emotional detachment characteristic of legal professionalism to impersonal routines. Routinized case handling “receives validation from the widely held view that individual client service bears little relationship to efforts to use the law for political or social change.” (Id. at 119.) Routine is seen as simple, dull; the routine case is the opposite of the “test case”. Routine is low status.
Bellow sees personal involvement and political orientation as essential to avoiding the further bureaucratization of legal aid work, and argues that the importance of day-to-day work must receive recognition. An appreciation of the complexities of routine, and an understanding of how some of its undesirable incarnations might be minimized, are two possible payoffs of the investigations suggested in this paper. The ways in which computers can both exacerbate these problems and contribute to their solutions will be increasingly central to such discussions.
VI. Ranging far afield
A. Chaos and the Fractal Anatomy of Legal Thought
Chaos theory teaches us that the successive iteration of even simple, deterministic procedures can produce chaotic complexity and that, conversely, system dynamics seemingly not traceable to discrete generative rules can exhibit startling regularities (in state-space trajectories.) Thus any ascertainable or fixed boundary between deterministic behavior and nondeterministic behavior is dubious. [Gleick, 1987]
Goldberger [1989] points out that fractal structures (such as the human brain) can greatly amplify the surface areas available for such functions as information processing, while being robust and resistent to injury and disturbance. The routine behavior of the heart and related organs is erratic and fractal.[8]
Based on these and related borrowings, I suggest that legal doctrine, legal consciousness, and structures of lawyering behavior can be viewed as fractal structures left behind by chaotic processes.[9] This viewpoint may incline us to be more appreciative of non-linearity as a vaccine against bureaucratic and other pathologies, and suggest novel architectures for computational models of lawyering.
B. The curvature of socio-legal space
Legal theorists from Cardozo [1928] to Tribe [1989] have found lessons for law in some of the most paradoxical doctrines of modern physics: the general theory of relativity and quantum mechanics. Drawing upon these and related writings, I develop ways in which it seems that routines can profitably be thought of as (1) indentations in socio-legal space, producing gravitation-like effects on human conduct; (2) phenomena which change upon observation (or thematization).
C. Talk, groceries, and all that jazz
The creative and improvisational dimensions of many “routine” activities such as conversational interaction [Goffman 1983], grocery shopping [Lave et al., 1984], and jazz playing (by experienced performers) [Sudnow, 1978] reveal depths of complexity that should give pause to those who would try to capture such activities in computable models. Good lawyering is similarly creative and improvisational, even when it is deployed in “routine” negotiations, interviews, or arguments. Microphenomenological investigations of lawyering that are informed by robust theories of situated action seem an essential precondition to any truly effective models of such behavior in artificial systems.
V. Routine revisited
Some tentative conclusions and questions about routine not already ventured include:
1. The routineness of an activity is not in itself a ground for disparagement or praise. Other qualities must account for our positive and negative valuations of an activity. Routineness, rather, is a potent enhancer of these other qualities. Its relative unproblematicalness and unreflexivity serve to magnify other qualities.
2. Routine can be both emergent and artifactual. Systems can fall into patterned dynamics through the cumulative effect of myriad tiny phenomena, or be compelled to behave in regular ways through concerted intervention. Control can be a self-regulatory phenomenon, or an external imposition. While the constituents of a system and their configuration may place powerful constraints on its responsiveness to arbitrary external control, no system is ever immune from such control. (Nor, as just noted, is external control necessarily odious.)
3. Routineness is orthogonal to both complexity and formality. Among other things, this means that the mere fact that something is routine gives us little warrant one way or the other of the amenability of that activity to effective computation using conventional architectures.
4. Routineness is always a matter of degree and relative to an actor or community of actors. What is routine for an expert is not routine for a novice. What is routine for one society can be central and extraordinary for another. Rather than certain things being routine, and others not, some things are more or less routine than others. It depends on context.
5. Goffman [1983] perceptively notes that “[a] competency can be defined as the capacity to routinely accomplish a given complicated end.” Although humankind may be able to achieve greater mastery over its environment by progressively delegating certain routines to machines, query whether individual humans can gain (and retain) mastery of important skills without personal engagement in such routines. The pedagogical, ritual, coordinative, and value-reinforcing functions of routines may counsel against their automation.
* * *
Rooting out reductionistic or simplistic notions of routine and coming to terms with related concepts in other fields promises rewards for a cognitive science of legal action as well as for pragmatics and pedagogy. While the quick analogies developed in this paper border on the sophomoric, more refined versions of them may serve to stir up unexamined (or prematurely settled) notions about everyday life in the law.
What is left, then, of the concept of routine? A root commonsense meaning of the term that we want and should be able to retain is the unproblematic. Nature has found ways to make extraordinarily complex activities routine, while leaving us enough fascination for a lifetime in the simplest of phenomena. It may be the destiny of human artifice to achieve similar accomplishments, and the vocation of law both to build on and undergird those accomplishments.
REFERENCES
Alfred C. Aman, Jr. “Studying Music, Learning Law: A Musical Perspective on Clinical Legal Education” 13 Cornell Law Forum №3, 8–12 (1987)
Scott Austin. “An Introduction to Genetic Algorithms,” AI Expert, March 1990, p. 49
Gary Bellow. “Turning Solutions into Problems: the Legal Aid Experience” NLADA Briefcase, August 1977, 106–122
Gary Bellow and Bea Moulton. The Lawyering Process (Foundation Press, Mineola, 1978)
Abraham S. Blumberg. “The Practice of Law as Confidence Game: Organizational Cooptation of a Profession,” 1 Law & Soc. Rev. 15 (1967)
George Boolos and Richard Jeffrey. Computability and Logic (1974)
Benjamin N. Cardozo. The Paradoxes of Legal Science (1928)
Robert Clairborne. The Roots of English (1989)
Dennis P. Donnelly, ed. The Computer Culture: A Symposium to Explore the Computer’s Impact on Society (1985)
Hubert A. Dreyfus and Stuart E. Dreyfus, What Computers Can’t Do (1986)
Charles Dyke. The Evolutionary Dynamics of Complex Systems (198_)
Lawrence Ferlinghetti. Routines (1964)
Marc Galanter, Frank S. Palen, and John M. Thomas. “The Crusading Judge: Judicial Activism in Trial Courts” 52 S. Cal. L. Rev. 699 (1979)
Clifford Geertz. “Common Sense as a Cultural System.” In Local Knowledge: Further Essays in Interpretative Anthropology 73 (1983)
James Gleick. Chaos: Making a New Science (1987)
Erving Goffman. Forms of Talk (1983)
Erving Goffman. The Presentation of Self in Everyday Life (1958)
Ary L. Goldberger, David R. Rigney, and Bruce J. West. “Chaos and Fractals in Human Physiology” Scientific American, February 1990, 43–49
Douglas R. Hofstadter. Gödel, Escher, Bach: An Eternal Golden Braid (1979)
Douglas R. Hofstadter. “On the Seeming Paradox of Mechanizing Creativity” In Metamagical Themas: Questing for the Essence of Mind and Pattern (1985), p. 526
Douglas R. Hofstadter, “Waking up from the Boolean Dream, or, Subcognition as Computation”, in Metamagical Themas, p. 631
John H. Holland, Keith J. Holyoak, Richard E. Nisbett, and Paul R. Thagard. Induction: Processes of Inference, Learning, and Discovery (1986)
M. Ethan Katsh. The Electronic Media and the Transformation of Law
Duncan Kennedy. “Freedom and Constraint in Adjudication: A Critical Phenomenology.” 36 J. Leg. Ed. 518 (1986)
Duncan Kennedy. “Form and Substance in Private Law Adjudication” 89 Harv. L. Rev. 1685 (1976)
George Lakoff and Mark Johnson. Metaphors We Live By (1980)
Jeanne Lave, Michael Murtaugh, and Olivia de la Rocha. “The Dialectic of Arithmetic in Grocery Shopping.” In Everyday Cognition ([Rogoff and Lave, 1984, infra])
Philip Leith. Formalism in AI and Law (1990)
Marcia Lipetz. Routine Justice (1984)
David Luban. “Legal Modernism” 82 Mich. L. Rev. 1656 (1986)
Marvin Minsky. “Why People Think Computers Can’t.” In The Computer Culture, supra.
Minsky, Naftaly H. and Rozenshtein, David. “System = Program + Users + Law.” In Proceedings of the First International Conference on AI and Law, Boston, 1987, p. 170.
Ernest Nagel and James R. Newman. Gödel’s Proof (1958)
John Nivala. “Our Nature in Balance: An Essay on Eighteenth-Century Landscape Gardening and Twentieth-Century Lawyering” 38 J. Legal Educ. 305 (1988)
Roger Penrose. The Emperor’s New Mind: Concerning Computers, Minds, and the Laws of Physics (1989)
Ilya Prigogine and I. Stengers. Order out of Chaos (1984)
George N. Reeke, Jr., and Gerald Edelman, “Real Brains and Artificial Intelligence,” Dædulus, Winter 1988, 143–173
Barbara Rogoff and Jean Lave, eds. Everyday Cognition: Its Development in Social Context (1984)
Pierre Schlag. “Missing Pieces: A Cognitive Approach to Law” Texas L. Rev. ( )
Marcel J. Schoppers. “In Defense of Reaction Plans as Caches” AI Magazine, Winter 1989, 51
Richard K. Sherwin. “Dialects and Dominance: A Study of Rhetorical Fields in the Law of Confession” 136 U. Pa. L. Rev. 729 (1988)
Daniel L. Stein. “Spin Glasses,” Scientific American July 1989 52–59
David Sudnow. “Normal Crimes: Sociological Features of the Penal Code in a Public Defender Office” Social Problems 255 (1964)
David Sudnow. Ways of the Hand: The Improvisation of Organized Conduct (Cambridge, 1978)
Laurence H. Tribe. “The Curvature of Constitutional Space: What Lawyers Can Learn from Modern Physics” 103 Harv. L. Rev. 1 (1989)
Charles Walter and Michael Parks. “Representing Legal Concepts as Cognitive Variables in Computational Models.” (Paper presented at 4th International Congress on Informatics and Juridical Regulation, Rome, May 1988.)
T. Winograd and F. Flores. Understanding Computers and Cognition (1986)
Steven L. Winter, “Transcendental Nonsense, Metaphoric Reasoning, and the Cognitive Stakes for Law,” 137 U.Pa.L.Rev. 1105 (1989)
Paul C. Wohlmuth. “Nested Realities and Human Consciousness: The Paradoxical Expression of Evolutionary Process” 25 World Futures 199–235 (1988)
[1] George N. Reeke, Jr., and Gerald Edelman, “Real Brains and Artificial Intelligence,” Dædulus, Winter 1988, 143
[2] Steven L. Winter, “Transcendental Nonsense, Metaphoric Reasoning, and the Cognitive Stakes for Law,” 137 U.Pa.L.Rev. 1105, n.150 (1989)
[3] See, e.g., [Minsky and Rozenshtein, 1987].
[4] One intermediate form is research about legal planning. See, e.g., Schlobohm and McCarty 1989. Here, what is being planned is the client’s situation. Some of the representation schemes for legal analysis — e.g., several described in ICAIL 87 — are rich enough to encompass the action dimension. Could McCarty’s LLD [McCarty 1989] be turned back on itself and its action language applied to lawyering process? Kowalski and Sergot have also developed a “logic-based calculus of events.” Gardner [1989] takes up temporality in another sense — the development of legal doctrine over time.
[5] This section is largely based on browsings through [Clairborne, 1989.]
[6] It is interested to note that words of circularity, curvature, and cyclicalness — including apparently wheel — come from a different family of roots, ones which incidentally also produce culture. (See entry for KWEL in [Clairborne 1989].) And one could have some fun drawing connections with the problems of circularity in legal and other reasoning: loops whose escaping arguably requires the exercise of non-mechanizable discretion. Donald H. Berman, “Cutting Legal Loops,” in Proceedings of the Second International Conference on AI and Law (1989), 251–258.
[7] Wittgenstein offers an intriguing example in his discussion about propositions that “stand fast” for him, but which he never explicitly learned: “I can discover them subsequently like the axis around which a body rotates. This axis is not fixed in the sense that anything holds it fast, but the movement around it determines its immobility.” ON CERTAINTY, p. 152 (1969). (I owe finding this passage to Pierre Schlag, who quotes it in [Schlag 1989], n. 56.)
[8] It is interesting that the chaotic nature of heart beat fluctuations appears attributable to the buffeting of two opposite stimulations from the sympathetic and parasympathetic branches of the autonomic nervous system — reminiscent both of Charles Walter’s “disystem” [Walter and Parks, 1988] and of adversarialism in law. It is also notable that the chaotic aspects of heartbeat seem contrary to classical control theory models of homeostasis as a response to fluctuations in the outside environment. Mainstream control and feedback theory may exhibit similar shortcomings in accounting for routine in social behavior.
[9] Such a geometry of mentality may account generally for our distinctive human ability to “metabolize recursiveness” [Wohlmuth, 1988] and bear some relation to the quantum theory of cognition developed in [Penrose, 1989].
