Contents

Preface

Acknowledgments

PART I - Theoretical Background

1 Introduction

1.1 The question of design

1.2 The role of tradition

1.3 Our path

2 The rationalistic tradition

2.1 The rationalistic orientation

2.2 Language, truth, and the world

2.3 Decision making and problem solving

2.4 Cognitive science

3 Understanding and Being

3.1 Hermeneutics

3.2 Understanding and ontology

3.3 An illustration of thrownness

3.4 Breaking down and readiness-to-hand

4 Cognition as a biological phenomenon

4.1 The closure of the nervous system

4.2 Autopoiesis, evolution, and learning

4.3 The cognitive domain

4.4 Consensual domains

4.5 The observer and description

4.6 Domains of explanation

5 Language, listening, and commitment

5.1 Listening in a background

5.2 Meaning, commitment, and speech acts

5.3 Objectivity and tradition

5.4 Recurrence and formalization

5.5 Breakdown, language, and existence

6 Towards a new orientation

6.1 Cognition and being in the world

6.2 Knowledge and representation

6.3 Pre-understanding and background

6.4 Language and action

6.5 Breakdown and the ontology of design

PART II – Computation, Thought, and Language

7 Computers and representation

7.1 Programming as representation

7.2 Levels of representation

7.3 Can computers do more than you tell them to do?

8 Computation and intelligence

8.1 Why do we ask?

8.2 Intelligence as rational problem solving

8.3 The phenomenon of blindness

8.4 What about learning and evolution?

8.5 Can pigs have wings?

9 Understanding language

9.1 Artificial intelligence and language understanding

9.2 The problem of background

9.3 Understanding as pattern recognition

9.4 What does it mean to understand?

10 Current directions in artificial intelligence

10.1 The forking of the paths

10.2 Expert systems

10.3 The fifth generation computer system

PART III – Design

11 Management and conversation

11.1 Management and decision making

11.2 Decision making and resolution

11.3 Organizations as networks of commitments

11.4 Decision support systems

11.5 Tools for conversation .......

12 Using computers: A direction for design

12.1 A background for computer design

12.2 A design example

12.3 Systematic domains

12.4 Technology and transformation

1.2 The role of tradition

One cannot approach questions like those raised in the previous section from a neutral or objective standpoint. Every questioning grows out of a tradition — a pre-understanding that opens the space of possible answers. We use the word ‘tradition’ here in a broad sense, without the connotation that it belongs to a cohesive social or cultural group, or that it consists of particular customs or practices. It is a more pervasive, fundamental phenomenon that might be called a ‘way of being.’ In trying to understand a tradition, the first thing we must become aware of is how it is concealed by its obviousness. It is not a set of rules or sayings, or something we will find catalogued in an encyclopedia. It is a way of understanding, a background, within which we interpret and act. We use the word 'tradition' because it emphasizes the historicity of our ways of thinking—the fact that we always exist within a pre-understanding determined by the history of our interactions with others who share the tradition.

When we encounter people who live in a substantially different tradition, we are struck by the impression that they have a strange and apparently arbitrary 'world view.' It takes a careful self-awareness to turn the same gaze on our own lives and 'unconceal' our own tradition —to bring into conscious observation that which invisibly gives shape to our thought.

In examining how people have thought about and talked about computers, we become aware of the pervasive effect of a powerful tradition that emphasizes ‘information,’ ‘representation,’ and ‘decision making.’ This tradition has been the basis for a great deal of technological progress and it has also led to many of the problems created by the use of computers. Even in discussions of what computers can and cannot do, the questions that are posed reflect a particular blindness about the nature of human thought and language—a blindness that can lead to a broad misunderstanding of the role that will be played by computers.

We have labelled this tradition the rationalistic tradition' because of its emphasis on particular styles of consciously rationalized thought and action. In calling it rationalistic' we are not equating it with 'rational.' We are not interested in a defense of irrationality or a mystic appeal to nonrational intuition. The rationalistic tradition is distinguished by its narrow focus on certain aspects of rationality, which (as we will show throughout the book) often leads to attitudes and activities that are not rational when viewed in a broader perspective. Our commitment is to developing a new ground for rationality—one that is as rigorous as the rationalistic tradition in its aspirations but that does not share the presuppositions behind it.

The task we have undertaken in this book is to challenge the rationalistic tradition, introducing an alternative orientation that can lead to asking new questions. In developing this new orientation, we were led to a critique of the current mythology of artificial intelligence and its related cognitive theories, drawing conclusions that contradict the naive optimism apparent in the quotations at the beginning of the chapter. Our ultimate goal, however, is not a debunking but a redirection. The alternative we pose is not a position in a debate about whether or not computers will be intelligent, but an attempt to create a new understanding of how to design computer tools suited to human use and human purposes.

8.5 Can pigs have wings?

Readers well trained in the analytic tradition will by this point have concluded that our argument that computers cannot be intelligent has several 'logical holes' in it:

In light of these points, aren't we being illogical or inconsistent in our assertion that computers cannot be intelligent?

In order to respond to this challenge, we need to recall the theory of language developed in Chapter 5. Sentences in a human language cannot be treated as statements of fact about an objective world, but are actions in a space of commitments. If this applies to the question "Is there any water in the refrigerator?" it must apply at least as strongly to "Can computers be intelligent?"

If we assume that the person asking the question is serious, there is an underlying background of purposes and understanding (the 'horizon' as Gadamer calls it) into which the question fits. If a questioner were to ask "Can pigs have wings?" a respondent within the analytic tradition might have difficulty answering, because although the idea is outrageously farfetched, current work in genetic engineering does leave open the logical possibility of creating a beast with the desired characteristics. Admittedly, there might be some refuge in challenging the asker as to whether such a monstrosity would still properly be called a pig, thereby invalidating the question. But if the question were asked seriously, neither the logical possibility nor the precise meaning of "pig" would be the issue at hand. The questioner would be asking for some reason in some background of understanding and purpose, and the appropriate answer (just like the appropriate answer to "Is there water in the refrigerator?") would have to be relevant to that background.

The background for serious questions about computer intelligence is the development of computer systems and their use in human contexts. What then is the basis for deciding whether it is appropriate to describe computers as potentially intelligent? In applying a predicate to an entity, one is implicitly committed to the belief that the entity is the kind of thing to which the predicate properly applies. 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—the question of free will has been debated for centuries and work in artificial intelligence has provided no new solutions. Rather, in using mental terms we commit ourselves to an orientation towards it as an autonomous agent.

There are many reasons why one can feel uncomfortable with the tendency to adopt the same orientation towards people (whom we take as autonomous beings) and towards machines. It is not a matter of being right or wrong, accurate or inaccurate, but rather of a pre-understanding that guides our discourse and our actions.

In attributing intelligence to machines, one is doing more than just taking what Dennett, in “Mechanism and responsibility" (1973, p. 246), calls the 'intentional stance.' He argues that in taking an intentional stance towards computers, all one is claiming is that "on occasion, a purely physical system can be so complex, and yet so organized, that we find it convenient, explanatory, pragmatically necessary for prediction, to treat it as if it has beliefs and desires and was rational " But treating a system as though it were rational (in the formalized sense of rationality) is very different from treating it as though it had beliefs and desires, and this is a significant confusion.

We treat other people not as merely 'rational beings' but as 'responsible beings.' An essential part of being human is the ability to enter into commitments and to be responsible for the courses of action that they anticipate. A computer can never enter into a commitment (although it can be a medium in which the commitments of its designers are conveyed), and can never enter as a participant into the domain of human discourse. Our earlier chapters point out the centrality of commitment for those aspects of intelligent behavior that initially seem based on more objective ideals of rationality. Even the ability to utter a 'true statement' emerges from the potential for commitment, and the absence of this potential gives computers a wholly different kind of being.

We do not treat the question of whether computers can be intelligent as a pure stance, with one or another choice to be taken for the sake of argument. We exist within a discourse, which both prefigures and is constituted by our utterances. The meaning of any question or statement lies in its role within this discourse. Our answer to the question of whether machines can be intelligent must be understood in the context of the questions raised by the other chapters, and in the orientation that these questions provide for action.