CEPA eprint 1698 (HVF-113)

Understanding Computers and Cognition. Book Review of Understanding Computers and Cognition: A New Foundation of Design by Terry Winograd and Fernando Flores

Foerster H. von (1987) Understanding Computers and Cognition. Book Review of Understanding Computers and Cognition: A New Foundation of Design by Terry Winograd and Fernando Flores. Technological Forecasting and Social Change, An International Journal 32(3): 311–318. Available at http://cepa.info/1698
Table of Contents
Polemic
Construction
(i) Heidegger
ii. Austin
iii. Maturana
Tool
Book Reviews
Terry Winograd and Fernando Flores, Understanding Computers and Cognition: A New Foundation of Design, Norwood, NJ: Ablex Publishing Corporation, 1986, 207 pages, $24.95.
Terry Winograd’s and Fernando Flores’ Understanding Computers and Cognition has been described by earlier critics, some in a laudatory way, some in an objecting mood, as profound, esoteric, incomprehensible, innovative, brilliant, disturbing, controversial, important, provocative, disappointing, annoyingly flawed, interesting, and in many other ways. For this reviewer Understanding Computers and Cognition is a delight to read. It reads like an elegantly written detective story which, when one has started, pulls one through the book by a style of writing that keeps sentences bristling with life and urgency. And also like a good detective story, one’s search for the villain is kept alive throughout, for he is omnipresent but not recognized, though always visible to all, and when one thinks one has snatched him, he slips away.
As the story unfolds, the villain turns out to be Language, hiding behind the story’s hero: Language. How can this be understood? When attempting to understand language, language is its own best enemy, for its appearance contradicts its function. In its ap-pearance it seems to be denotative, monologing about things in the world: in its dialogical function it is connotative, appealing to concepts in the other’s mind. How can one get from appearance to function or from function to appearance, since one needs the one for the other? A catch 22, a vicious circle, a case of perverted – or better – inverted logic; a logic Aristotle would not dream of, except in his nightmares. One asks, can this still be called logic? And, of course, one asks why bother? Why being concerned at all?
The concern here arises from the problem of design in a sociocultural context. Both Winograd and Flores lived with, participated in, and contributed to the evolution of the concept of computing and its realization in technology, management, and utilization. Winograd was one of the first who contemplated access to the huge data bases in machines though natural language, and the structure of such a data base that it may be so accessed. Flores participated in perhaps the then largest socioeconomic management design (1973) for a whole nation of about nine million people; that was Chile under Salvador Allende.
Their collaboration began in Stanford, California, soon after Amnesty International secured Flores’ release (1976) from one of Pinochet’s prisons near Cape Horn. As they describe it in their Preface, they entered into a dialogue that began with a concern about the design of computer-based tools, which then grew into questions about what people can do with computers, and finally evolved into this book, that asks the question of what people do in a domain of coordinated linguistic action.
If we wish to know what this book is about, they tell us in the first sentence: “This is a book about the design of computer technology.” Fortunately, contextual depth to this apparently narrow program is given in the next sentences when consequences of technological development are seen: “The use of technology … leads to fundamental changes in what we do, and ultimately in what it is to be human. We encounter the deep questions of design when we recognize that in designing tools we are designing ways of being.” (p. xi)
Those who are conversant in commercialese or computerese may quickly translate the language used above into their own: the concern with language in connection with computing is taken care of by the term: Artificial Intelligence (AI), and the human orientation is translated into the term “user friendly.” Hence, Winograd’s and Flores’ program is apparently nothing but “user-friendly AI.”
It is precisely this or related interpretations of their goal which the authors wish to avoid by all means. Indeed, they make a major effort to distance themselves from AI, for, as they argue, AI grew from a conceptual soil, they call “the rationalistic tradition,” which inhibits understanding of exactly that which its offspring purports to address, namely intelligence, and inhibits moreover even the understanding of understanding itself.
An equivalent effort is put into the development of their own position, which is centered on the notion of language as a coordinating agent for actions among conversing human beings. And they are also aware that: readers with a background in science and technology may find it implausible that philosophical considerations have practical rel-evance for their work. Philosophy may be an amusing diversion, but it seems that the theories relevant to technical development are those of the hard sciences and engineering. We have found quite the opposite. Theories about the nature of biological existence, about language, and about the nature of human action have a profound influence on the shape of what we build and how we use it (p. xii).
The two types of effort, the one polemic, the other one constructive, allows the discussion of the work of Winograd and Flores along these two lines.
Polemic
In order to call unencumbered upon those philosophical foundations on which they wish to build their own epistemology and technology, the authors had to begin with a tabula rasa and, at the same time, to be aware that this tabula rasa is but one of the many items placed on a tabula prodigiosa, for they use language to undo some language.
Their first victim is the “rationalistic tradition,” the prevalent frame for establishing meaning form such fundamental concepts as world, cognition, language, truth, decision making, problem solving, etc. The point of departure for this tradition is “naive realism” or any one of its political cousins. Here an objective, observer-independent world “out there” is stipulated, of which the cognizing subject makes an internal representation, a symbolic map. In turn, language produces another map whose fidelity, i.e., “truth,” is to be checked against its correspondence with the features of the world.
In their devastating critique of this position, a critique that is spread out in bits and pieces all over the book, Winograd and Flores do not so much make use of the impossibility of checking the correspondence of any proposition with the world, for the world is here accessible only over maps, they address themselves to tacit assumptions within, and consequences of, this position.
First, they note that language arises here in a vacuum, without being spoken to anyone, and without any contex or background. But language is not a monologic, but a dialogic affair, where:
background is the space of possibilities that allow us to listen to both what is spoken and what is unspoken. Meaning is created by an active listening, in which the linguistic form triggers interpretation, rather than conveying information. The background is not a set of propositions, but is our basic orientation or ‘care’ for the world. The world is always already organized around fundamental human projects, and depends upon these projects for its being and organization (p. 57).
Second, they point out that since the world must be unequivocal, its (linguistic) map must likewise be so. From this, rationalistic tradition draws several conclusions. One is that utterances allow for one and only one interpretation; another one is that from un-ambiguous propositions all inferences can be derived in an algorithmic fashion: thought can be formalized.
The authors destroy the first of these conclusions by submitting case after case of multiple interpretations, one of which is this exchange:
A: Is there any water in the refrigerator?B: Yes.A: Where? I don’t see it.B: In the cells of the eggplant (p. 55).
Winograd and Flores continue:
A claims that B’s first response was a lie (or at best `misleading’), while B contends that it was literally true. Most semantic theories in the rationalistic tradition provide formal grounds to support B, but a theory of language as a human phenomenon needs to deal with the grounds for A’s complaint as welli.e., with the ‘infelicity’ of B’s reply.
The other conclusion about the formalization of thought processes was, of course, the impulse for undertaking a major computer design effort: artificial intelligence. This effort had two sides. One to simulate mental activity with appropriate computer architecture, the other one to learn from this effort more about how the mind works.
In a superb discussion of our understanding of the organization of the nervous system as a closed network of interacting neurons, Winograd and Flores show that not only does the notion of mapping have no place in this picture, but also that anthropomorphic metaphors applied to machines may, at best, be misleading, and at worst, be dangerous: “In uttering a sentence containing mental terms (`intelligent,”perceive,”learn’), we are adopting an orientation towards the thing referred to by the subject of the sentence as an autonomous agent. The issue is not whether it really is autonomous … Mather, in using mental terms we commit ourselves towards it as an autonomous agent” (p. 105).
They continue: “An essential part of being human is the ability to enter into com-mitments and to be responsible for the course of actions that they anticipate. A computer can never enter into a commitment, … and can never enter as a participant into the domain of human discourse” (p. 106).
And finally: “it must be stressed that we are engaging in a particularly dangerous form of blindness if we see the computer – rather than the people who program it – as doing the understanding” (p. 123).
Construction
Winograd’s and Flores’ construction rests on three pillars of philosophy built by the German existentialist Martin Heidegger (1889), the British phatologist John L. Austin (1911), and the Chilean biologist Humberto Maturana R. (1928). Not only the ideas of these three men of substantially different spheres of interest and competence, but also their highly idiosyncratic terminology have been used by the authors in an impressive tour de force of integration.
What are these philosophies? In brief, they are Heidegger’s recapture of the state (of innocence) of non-(self)refiecting daily life: to be now and here (“Dasein”); Austin’s pointing to the fascinating properties of utterances that do what they say (“performative utterances,” and Maturana’s insight about the specific organization of living things whose interactive components produce these very components (“autopoiesis”). Why this diversity of resources, and to what end are they called upon?
The Herculean task the authors have set for themselves is nothing less than to toss overboard the predominant attitude toward language as a cargo cult in which meaning, feelings, intents, etc., are “transmitted,” or ideas, thoughts, information, etc., are “ex-changed,” whereby information may be “processed,” “stored,” “retrieved,” “compressed,” “chopped,” etc., as if it were hamburger meat.
Instead, Winograd and Flores would like to create an understanding of language where “language does not convey information, [but] evokes an understanding or ‘listening,’ which is an interaction between what is said and the pre-understanding already present in the listener.”[Note 1]
While here each person is unique, each, however, is “sharing a background to varying extents with other people. Some amount is universally human; more is shared with members of the same culture; more yet with those in the same line of work; and still more with partners in frequent conversations.”[Note 1]
How then can Heidegger, Austin, and Maturana help in developing this position? This point the authors explain, not only in chapters devoted to the main ideas of each of these men, but again in bits and pieces throughout this volume, all together making up perhaps more than one third of the book.
(i) Heidegger
(Terms at-hand: breakdown, present-at-hand, ready-to-hand, thrownness, unready-to-hand). Unfortunately, this reviewer is allergic to Heidegger’s tortured language which is tough enough in German, and even more bloated in (sometimes badly translated) English. Nevertheless, Heidegger is up to something that is not easy to grasp. Among the broad spectrum of ideas Winograd and Flores use in their construction, only two points will be alluded to here. One is Heidegger’s invitation to us to perform the mental acrobatics in which we are to reflect upon the nonreflecting state of the daily doing, functioning and acting: just to be, here and now (“Dasein,” being-in-the-world), as if cast into this state (thrownness). For example, when putting on one’s shoes, riding a bicycle, driving a car, etc., one is not reflecting on ones actions: “it” does it. [Note how the literal and metaphorical meaning of manipulative actions – i.e., those one does with one’s hands (Heidegger’s X-at-hand) – like to grasp, comprehend, etc., connect action with mentation].
Why reflect on this state? Because only when interrupted: the shoelace breaks, the tire blows, etc., breakdown occurs and “objectness” emerges: To the person doing the hammering, the hammer as such does not exist. It is a part of the background of readinessat-hand that is taken for granted without explicit recognition or identification as an object … Its `hammerness’ emerges if it breaks or slips form grasp or mars the wood, … As observers, we may talk about the hammer and reflect on its properties, but for the person engaged in the thrownness of unhampert hammering, it does not exist as an entity (p. 36).
The other point is connected with the problem of establishing meaning and under-standing. As Gottlob Frege (1884) observed: “Never to ask for the meaning of a word in isolation, but only in the context of a proposition.”[Note 2] However, how is context created except through words? This is the problem of the hermeneutic circle. Winograd and Flores refer directly to Heidegger: “But if we see this circle as a vicious one and look out for ways of avoiding it, even if we just sense it as an inevitable imperfection, then the art of understanding has been misunderstood from the ground up” (p. 32).
Winograd’s and Flores’ job is to make Heidegger’s notions understood (the didactic problem), and to clarify how they will use these notions (the P.R. problem). They succeed superbly on the latter, which helps the reader understand the former. However, because they have to explain Heidegger’s notions in the first place, the definientia could have been used instead of staying with the definienda. In other words, could this section be done better? Since nobody has tried before, nobody knows.
Here is the authors’ summary on Heidegger:
In sum, Heidegger insists that it is meaningless to talk about the existence of objects and their properties in the absence of concernful activity, with its potential of breaking down. What really is is not defined by an objective omniscient observer, nor is it defined by an individual – the writer or computer designer – but rather by a space of potential for human concern and action (p. 37).
And here is Martin Buber’s comment on Heidegger
It is in togetherness that boundlessness and unconditionedness are experienced. Heidegger not only turns away from the relationship to a godly unconditionedness, but also from a relationship where the human discovers another than himself in unconditionedness, and so experiences this unconditionedness. Heidegger’s “Being” is a monological being. For a while the monologue may very well disguise itself artfully and appear as dialogue. It may well be that an unknown layer after another of the human self may answer to the inner address, so that the human makes new discoveries and thus can believe that he experiences a “calling,” and a “hearing”; but there comes the hour of the stark loneliness, when the silence of being becomes unsurmountable, and when the ontological categories do not want to be applied any more.[Note 3]
ii. Austin
(Terms at-hand: assertives, commissives, declarations, directives, expressives, il-locutionary force, illocutionary point.) In 1956, in the third Programme of the BBC, Austin delivered a speech entitled “Performative Utterances” in which he reported some fascinating linguistic forms: utterances that do what they say. In other words, they speak about themselves. For instance, after having stepped on somebody else’s toes one utters: “I apologize!,” one has indeed apologized. Similarly this is true with promises, requests, declarations, etc. The magic of these utterances becomes particularly apparent when, in the marriage ceremony of a young couple, the priest utters the incantation: “I declare you husband and wife.” And presto! they are now husband and wife. Moreover, Aristotle’s test for a proposition to make sense by being either true or else false does not apply to these utterances.
A further study of speech acts in general persuaded Austin to distinguish at least two kinds of act that are performed simultaneously when something is uttered: (1) the utterance itself, the very noises that are produced, which he called “locutionary act” (from Latin “loquor” to speak); (2) besides the performance of making the appropriate noises there is something else going on that is not the noise of the saying: the non-locutionary, or il-locutionary act, which may consist (e.g., in standing up while saying something) in making this utterance.
Austin himself developed a provisional catalogue of such illocutionary acts. This catalogue was expanded by others into a more complete list, containing, for example, assertives (committing the speaker to something as being the case), commissives (com-mitting the speaker to some future actions), directives, declaratives, and some more. Because of its significance in their own work, the notion of commitment assumes a central role for Winograd and Flores. For illustration they cite the German sociologist Juergen Habermas:
The essential presupposition for the success of an illocutionary act consists in the speaker’s entering into a specific engagement, so that the hearer can rely on him. An utterance can count as a promise, assertion, request, question, or avowal, if and only if the speaker makes an offer that he is ready to make good insofar as it is accepted by the hearer. The speaker must engage himself, that is, indicate that in certain situations he will draw certain consequences for action (p. 59).
The important point in their presentation of speech act theory is Winograd’s and Flores’ amplified attention to tying utterance to action, because it is this point that is the basis for their design of computer technology. Hence, a broader treatment in introducing terms like “illocutionary” that cannot be found even in major dictionaries (e.g., the New Oxford Dictionary) would be in their own interest. Their strategy is, so they tell us, to refrain from “definitions” in the lexical sense by encouraging the reader to find meaning through usage. But that requires closing the “hermeneutic circle.”
By calling upon Heidegger, the authors brought us down to earth or, in Heideggerese, down to being-in-the-world, alas, alone. By calling upon Austin, we learn that utterances are not for their own sake, but are components in conversations for action.
But who else is in this world? With whom does one converse? Where is life?
iii. Maturana
(Terms at hand: autopoiesis, closure, consensual domain, structural coupling.) What is life? The usual approach in answering this question is to generate a list of attributes like adaptation, biomolecular configurations, growth, homeostasis, metabolism, purpo-siveness, being reactive, reproductive, stimulable, etc., that will ultimately catch life in a semantic net. The remarkable thing about such lists is how lifeless they are. The reason for this is that they list the epiphenomena only, without touching the phenomenon proper.
A new strategy is necessary that calls for identifying an organization that, under a broad range of conditions, reconstitutes itself under its own operations. Maturana, with his student and co-worker Francisco Varela, solved this problem and, in 1973, published its solution.[Note 4] They called the process underlying this organization “autopoiesis” (from the Greek “autos” self, and “poiesis” the art of poetry, a making). The idea here is the identification of a specific form of interactive process acting between components of a system such that the outcome of these processes are the very components that, in the first place, gave rise to those interactions. Hence, this system is organizationally closed, and functionally (re-) generating itself recursively. The realizations of this organization are called “structures” which manifest themselves in myriads of forms from the amoeba to the cockroach to Homo habilis, etc., maintaining themselves by being thermodynamically open.
The importance of the solution to the problem “What is Life?” is not only what it says, but also the form in which the solution is presented, namely, not as a list of attributes, but as the identification of a process, specifically of recursive nature. This is crucial for Winograd’s and Flores’ effort to help us in understanding computers and cognition, because now they can treat “Cognition as a Biological Phenomenon,” which is indeed the title of their Chapter 4.
In Chapter 4 they touch upon some of the epistemological consequences of the notion of autopoiesis, e.g., the autonomy of a living organism and the maintenance of its own integrity, but they exploit another aspect of the underlying circularity, and that is the concept of the nervous system as an operationally closed system.
A logically consequential, but nevertheless stunning, feature of an organizationally closed system is the absence of inputs and outputs. This is because if any point in this vast network of neuronal interconnections (in the human central nervous system about 1016 synapses) is perturbed, the operations of the whole system are modified, hence “input” and “output” lose their meaning. To illustrate, Winograd and Flores cite Maturana:
Learning is not a process of accumulation of representations of the environment; it is a continuous process of transformation of behavior through continuous change in the capacity of the nervous system to synthesize it. Recall does not depend on the indefinite retention of an … invariant that represents an entity (an idea, image or symbol), but on the functional ability of the system to create, when certain recurrent conditions are given, a behavior that satisfies the recurrent demands (p. 45).
Perhaps the most important contribution to the authors’ plan to set out for new directions in the design of computer technology is Maturana’s approach to the origin of language. It is indeed the concept of autopoiesis that provides the necessary bridge between Heidegger’s loneliness and Austin’s illocutionary acts that need at least two – the speaker and the hearer – to make any sense at all. Winograd and Flores reproduce excerpts from Maturana’s own account that is given here more fully:
When two or more organisms interact recursively as structurally plastic systems, each becoming a medium for the realization of the autopoiesis of the other, the result is mutual ontogenetic structural coupling. From the point of view of the observer, it is apparent that the operational effectiveness that the structurally coupled organisms have for the realization of their autopoiesis is established during the history of their interactions. Furthermore, for an observer, the domain of interactions specified through such ontogenetic structural coupling appears as a network of sequences of mutually triggering interlocked conducts that is indistinguishable from what he or she would call a consensual domain. In fact, the various conducts or behaviors involved are both arbitrary and contextual. The behaviors are arbitrary because they can have any form as long they operate as triggering perturbations; they are contextual because their participation in the interlocked interactions of the domain is defined only with respect to the interactions that constitute the domain. Accordingly, I shall call the domain that results from ontogenic reciprocal structural coupling a consensual domain (p. 48).
Please note the appeal to ontogenesis in this approach to the phenomenon of language, that is “when the ontological categories do not want to be applied any more.” Also note in the first sentence the condition of mutuality for the realization of the autopoiesis of each.
As a further help in appreciating the notion of consensual domain the authors give this additional citation: “The basic function of language as a system of orienting behavior is not the transmission of information or the description of an independent universe about which we can talk, but the creation of a consensual domain of behavior between lingu-istically interacting systems” (p. 50).
As colophon to this section on construction, Winograd’s and Flores’ own reflection is appropriate:
In this view, language … is no longer merely a reflective but rather a constructive medium. We create and give meaning to the world we live in and share with others. To put the point in a more radical form…. We design ourselves … in language (p. 78).
Tool
“Coordinating people is the business problem of management,” writes R.S. Richman in FORTUNE in an article about team work.[Note 5] This is precisely the problem to which Winograd and Flores have addressed themselves in the first place, for whose solution they have amassed their epistemological machine, to whose implementation they turn in the final chapter of their book.
They begin by analyzing a typical business situation reflected in conventional ter-minology like choice structure, decision making, optimization, etc. It is quite impressive to watch the authors’ conceptual bulldozer push all this out of the way by, for example, showing that in important case to solve a problem is to dissolve it: “resolution comes in the creation of a new alternative. The issue is not one of choosing but of generating” (p. 149). Ontogenesis again! It is the possibility of “the new,” or better, its necessity, that is embraced in this work.
Its implementation can be seen in three phases: (1) determination of the structure of conversations; (2) the role of illocutionary acts; (3) integration of these ideas into CO-ORDINATOR, a computer program.
1. Structure. The authors make a convincing argument that the type of actions that are being initiated, monitored, modified, etc., in an organization’s network of conversations require in almost all cases not more than that these networks include requests and promises to fulfill commitments, reports, declarations, etc. In other words, it turns out that to a large extent organizational needs can be fulfilled by: “the activation of certain special networks of recurrent conversations, where only certain details of the content of the conversation differ, not the general structure. These networks of recurrent conversations are the core of the organization” (p. 108).
2. Illocutionary acts: “There are surprisingly few basic conversational building blocks (such as requests/promise, offer/acceptance, and report/acknowlegement) that frequently occur in conversations for action. The development of a conversation requires selection among a certain finite set of possibilities that is defined by the opening directive and the subsequent responses. It is like a dance, giving some initiative to each partner in a specific sequence” (p. 159).
3. COORDINATOR. The manuscript of this book is the “Version of December 21, 1984,” at which time early versions of COORDINATOR were already operational. Some of its essential features such as keeping temporal relations, monitoring completion, examination of the network, automated application of recurrence, recurrence of propositional content, etc., are outlined in crisp and clear language.
The sympathetic reader is curious about COORDINATOR’s future fate. Fortunately, again FORTUNE has the answer: Bonnie Johnson, a corporative technical planning manager at Aetna Life says about the COORDINATOR: “It is the first effective group communications tool I have seen.”[Note 5]
As reported earlier, Winograd and Flores say in their Preface that “readers with a background in science and technology may find it implausible that philosophical consid-erations have practical relevance for their work. Philosophy may be an amusing diversion.” However, according to Bonnie Johnson, this amusing diversion turns out to be very successful indeed.
Endnotes
1
Winograd, T. and Flores, F.: Response to review of Understanding Computers and Cognition, Artificial Intelligence, An International Journal 31, 251, 1987.
2
Frege, G., The Foundations of Arithmetic (translated from the German by J.L. Austin.) Philosophical Library, New York, p. x’ 1950.
3
Buber, M., Das Problem des Menschen (translated from the German by HVF). Lambert Schneider, Heidelberg, 1961.
4
Maturana, H. R., and Varela, F. G., De Machinas y Seres Vivos. Editorial Universitaria, Santiago, 1973.
5
Richman, L. S., Software catches the team spirit, FORTUNE 105 12,125-136, 1987.
Found a mistake? Contact corrections/at/cepa.infoDownloaded from http://cepa.info/1698 on 2016-05-13 · Publication curated by Alexander Riegler