|
pg 5
Mental Activity and Computation
Marking off the areas of mental activity for which a compulational analysis is not appropriate
Argument for the Need for a Non-Computational Analysis of Sensory, Experiental and Existential Phenomena
Among cognitive scientists and researchers in artificial intelligence are some who still argue for what John Searle calls 'strong AI'. This is a strong thesis that claims that the workings of a computer constitute a model of the workings of a mind, and that a mind is just a biochemical computer. According to this thesis, brain 'wet ware' and computer hardware are the same type of thing, which can be instantiated in either a biochemical or silicon chip medium .
For these claims to be true, the brain's operations must parallel the structure of the symbolic logic system with which computers are programmed, and must have many of the same functional and operational properties as a logical system. For example, the brain must be primarily a computational device that calculates mathematical and symbolic functions, if strong AI is to maintain its claims. While I think that artificial intelligence research has brought impressive insights and progress to recent study about the mind, there are certain features of mental operation that I think can be better explained. Certainly, much of the operation of a mind is computational and does operate as a symbol manipulator, in the way that advocates of strong AI claim. But I shall argue in this book that there are many key features of thinking that can be better understood in another way.
At the outset, I might indicate the direction that my analysis will take by pointing out that I think that the proponents of strong AI are concentrating on the products of thinking; the propositionally structured mental representations that might be said to be the objects of a knower's knowledge. I will concentrate, instead, on the processes whereby knowledge is generated by a knower. Thinking is, after all, an activity.
Thinking is an activity.
I propose an analysis of thinking that considers its structure to be that of an interactive relationship between a knower and his or her world. Since Aristotle's time, at least two ways of coming to have knowledge have been acknowledged. Aristotle distinguished between these two methods of knowledge acquisition in the Posterior Analytics, when he discussed the types of 'pre-existent' knowledge that one had to have if one was to learn anything.
The pre-existent knowledge required is of two kinds. In some cases admission of the fact must be assumed, in others comprehension of the term used, and sometimes both assumptions are essential...as regards 'unit' we have to make the double assumption of the meaning of the word and the existence of the thing. The reason is that these several objects are not equally obvious to us. Recognition of a truth may in some cases contain as factors both previous knowledge and also knowledge acquired simultaneously with that recognition—knowledge, this latter, of the particulars actually falling under the universal and therein already virtually known.
In this passage Aristotle claims that gaining knowledge requires both a recognition of the 'meaning of the word', which he allies with 'falling under a universal' and the recognition of the 'existence of a thing', which he allies with 'knowledge...of the particulars.'
Following Aristotle's distinction, I will claim that the process by which one acquires knowledge of particular existent things is quite different from the process by which one acquires knowledge of concepts and universals. Propositions combine these processes into products, as Aristotle himself does when he declares that all basic knowledge is of the form 'x is y'.
In this formula, the 'x' represents some subject, and the 'y', some property or concept being attributed to the subject. I adopt Aristotle's formula for basic knowledge, and his distinction between the two types of knowledge in this book in order to adapt his insights to my analysis of the structure of thinking.
People, I claim, perform two types of thinking processes. One corresponds to the generation of the 'x' in the Aristotelian formula, so I will call it 'object positing' loosely adapting Quine's use of that concept. The other, corresponding to the 'y', I will call 'property attributing.'
The names imply only that different methods for mental processing are involved in thinking of the relationship between oneself and some understood item of knowledge as a relationship to an object, than are involved in thinking of the relationship between oneself and some understood item as a relationship to a property or set of properties. Nothing else about properties or objects is implied by the choice of names.
My reason for distinguishing between these processes in this way is to point out how each can be separately analyzed, even if they rarely operate independently of one another. The type of analysis that I intend to give of each will be the interpretive kind of psychological explanation advocated and defended by Robert Cummins in "The Nature of Psychological Explanation".
The object-positing capacity of mind is an identifying and recognizing capacity, that deals with particulars in thought. Cases of immediate perception or direct experience are the chief types of knower-known relationships in which we would find the object-positing capacity operating relatively independently. It is an empiricist's direct hold on experience as reality.
The property-attributing capacity of mind is a sorting, qualifying and quantifying capacity, which deals mainly with universals in thinking. Apprehension of second or third order relations might rate as the types of knower-known relationships in which we would find the property-attributing capacity operating relatively independently.
Is the Basic 'Stuff' of the Universe Properties or Objects ?
If one asks the metaphysical question, 'What exists?', or 'What things are there?', the natural response is 'objects' or 'matter'. Metaphysical questions most frequently yield universe inventories in response. Once one has compiled a list of objects or components of objects in one's universe inventory, analysis of the list can go several ways. One could be impressed with the unity of the list, and insist with Parmenides and Spinoza that there is really only one, very complex substance, here. Or one might be impressed with the diversity of the list, and defend some form of atomism to explain recurrent patterns across differing items in the list. One could claim that modern chemistry and physics have united the primitive, natural answers to the metaphysical question by offering a table of elements, together with an account of bonding principles for intermediate composites of elements, plus an account of the relationship of all material diversity to one thing, energy, in E = mc2. But, however one analyses one's object lists, the lists themselves will contain only particular things. Concepts, relations, and other platonic entities would seem to be ruled out by an enterprise aimed at compiling a universe inventory. Mental phenomena of any kind make very questionable objects.
If, on the other hand, one starts one's inquiry with questions about knowledge, such as 'What do we know?' or 'What can be known?', the natural answers to these questions seem to be 'properties' or 'universals'. Plato and Berkeley both start their investigations with epistemological questions, and both ultimately have trouble with particalar, material objects. For, once the properties and universals have been established as prior, objects become reducible to sets of properties. The third man argument exhibits the chief weakness of a property-oriented account of the nature of the world. Properties and relations are too variable to rate as the basic content of a recalcitrantly solid reality.
pg 9
....two types of mental process by which people generate for themselves mental interpretations of the way the world is; object positing, and property attributing. I'm claiming only that humans have two distinct methods by which they characterize their experience for themselves, and therefore, two quite differently organized types of experiences of mental data can be presented to the mind for thought.
The end products of the processes will be objects and properties. These, in turn, in combination, will yield facts, expressible as Fregian propositions. Once full-blown propositions are in place, computational processes may operate on them, as I see the situation. But this occars many steps beyond the basic operations of the mind that I am arguing are the gronnd-floor operations.
As I am presenting the situation, the dispute between Plato and Parmenides, or between Quine and Goodman, is a question of preference for method of thinking. Whether by nature or nurture, I don't wish to argue here. But, it seems to me rather apparent that some people are property-thinkers and others are object-thinkers, to a greater extent.
This analysis is not a proposal that there is no external world, or that there are no objects or properties. It is only a proposal that more productive results might be accomplished in philosophical inquiry by examining mental processes, first.
pg 12
To analyze thought properly, one must attend to two types of processes; the process in which one incorporates the passing show into thought, and the one in which one computationally structures the raw data incorporated by experience.
The language of thought, as a computational structuring system, will not do alone, because its substitutionality deprives it of substantive content; it lacks the raw data of experience, as it exists, unincorporated into a structured system like a language. And the passing show is inconclusively determinate without the organization imposed by the structuring system. This is not a claim that 'the passing show' is unknowable, but only that it does not arrive in conscious experience equipped with a program. So, only an analysis of the union of the two can 1) give an adequate account of stable objects, 2) give an adequate account of causation, and 3) give an adequate account of how the mind uses mental data.
In what follows, I will develop arguments for independent theories of the 'x' and the 'y' in thought, together with proposals for constructing those theories and suggestions for ways to construe the composite.
Syntax and Logical Structures are Platonic, but
Semantics belongs to Aristotle's 'x'
In some respects, my distinction between a process yielding knowledge by acquaintance and a process yielding knowledge by chains of reasoning or "meanings of terms" is not new in the history of philosophy. But recent philosophical work, especially in the analytic school and among cognitive scientists like Stalnaker, has concentrated on developing theories about the 'y' process in thought, at the expense of the 'x'. First, let me sketch the line between the two types of theory where I think it should go, using the terms of the contemporary debate on issues related to thought. Then I will argue tbat the contemporary platonism of Quine, Stalnaker and kindred thinkers is inadequate to deal with the data that I have segregated as being properly analyzed as x-type data.
As I am analyzing these processes, a theory about the y-process would give analyses of truth conditions for statements, entailment relations among propositions, concatenation rules for languages, and accounts of the computational properties of mental reasoning. Logical or necessary truths, kinds, whether natural or otherwise, de dicio truths and syntactical analyses would also be included in the theory about y-processes in thought. Possible-world semantics is also an analysis of how thoughts concatenate, not of what their content is.
A theory of the role of the'x' in thought, however, will be perceptually or sensorily based, dealing with singular items, and related to the existence or appearance of the data involved in the thinking process, rather than to generalizations about that data. De re truths, indexicals, names, and singular referring expressions will be analyzed under these auspices.
This theory will analyze the way that we use perceptual input to image objects, and how, conversely, our conceptualization about the data generates selectivity in perception. One of the most central features of this data will be its intentionality. Logically intensional and opaque expressions will belong here as well. Semantics is amenable to an analysis in truth-functional terms when construed linguistically, but is also notoriously infected with Gricean intentionality, which, I think must be analyzed in an Aristotelian way as some 'x', some 'this', playing the role of 'matter' for a 'substance-making' process in thought. Cognition has no content without experiential interaction.
Platonic Cognition is Vacuous Without Aristotelian Perception
Quine rejected most of the data that I have placed in the theory of the x in thought as too unruly to be handled according to his principles, and hence, too unruly to countenance, period. In so distinguishing fields of study, I will argue that the principles of a theory of truth conditions are certainly not the ones that one ought to use to tackle the topics that I have characterized as x-type topics. But these topics must be tackled, even if new principles must be devised for the job. I consider Kripke's possible world semantics, as well as adaptations on Kripke's analysis by authors like Stalnaker attempts at reducing the data that I am distinguishing to something analyzable in terms of truth functions. But, I think that these authors have butchered the data to make it fit, and lost, not gained, intellectual clarity in the effort. I share Quine's lack of sympathy with the enterprise.
The effect of concentrating inquiry on the platonic entities, and despising or underrating the perceptual, ephemeral, 'x' side of my distinction, as Quine clearly does, is to discard as 'rationally irrelevant' all that passes in time. The platonic universe is, banally, eternal, as are eternal sentences and propositions. But, regrettably, we - poor shadows of our platonic selves - are mortal and transitory, as are most of the objects and events that we experience and discass. Although some fields of human endeavor, such as mathematics, may be able to get by without the 'x', one field that most certainly can't make do without the hedonistic devil is theories about causation. Banally, again, theories of causation are not about eternal truths, bot rather are about how transition takes place in time. I will argue in this book that the notion of a stable object depends on the notion of causation that one is using, and so, without a clear notion of causation, one's analysis of a stable object is bound to be badly skewed. Existence, I shall argue, belongs to stable objects.
Three Specific Areas of Mental Activity that are Butchered by Reducing x-Type Phenomena to y-type Phenomena:
Causation, Stable Mid-Sized Objects and Existence
Hume's critique of causation assigns causation to the domain of the transitory in order to argue against its alleged law-likeness. Remember, no causation is observed when the stone hits the glass because the event is a series of momentary 'ideas', each of which is a particular, unique 'x' in thought. In so analyzing the situation, Hume consciously and deliberately rejects the persistence through time of any stable object. In his discussion of personal identity, even the series of perceptual ideas that constitute his own identity turns out to be insufficient to yield even himself as a stable object over time.
Hume's problem with stable objects focuses particularly sharply on the conceptual interdependence of the roles of the 'x' and the 'y' in thought. Hume says a series of unconceptualized sensory 'hits' will not yield even oneself as a stable object. And, it is because no causation is observed that no stable object can be generated. Of course, without stable objects, any discussion of universal laws covering stable objects is going to be moot. Anyway that we construe this observation, it is bad news for philosophers like Kripke, who think that a mere, indiscriminate pointing is going to yield not only causation, but also a full blown natural kind, genetics, chemical elements, and all. If I am right, a pointing yields nothing but a completely unique sensory experience. From this, one cannot even presume that causation connects experiences or events, much less that the same objects are featured in various events.
From the point of view of the analysis that I am offering, Hume is right to look for causation and for the stability of objects in the realm of experience of particalars and the transitory, but by concentrating on products of mental operation, rather than processes used for mental operation, he cut his materials too thin. He is even indecisive about whether his'objects' are ideas, as mental percepts, or events, such as stones hitting windows. What he needed was the process by which an active mind makes a stone hitting a window its own thought.
Hume could then have said, as I do, tbat causation was experienced as the process by which a mind incorporates an experience into thought. Via this process, a mind identifies the mental idea with the experienced, physical event, attributing substantiality to the stable aspects of the experience, and causation to the transitory aspects of it.
pg 26
Summary
In this chapter, I have argued that there are non-computational mental processes that must be considered if one is to understand thinking. I have argued that, in addition to the computationally based platonic mental processes, there are also experientially based, object-positing processes. The object-positing processes deal with data of an unclassified sort which may be sensory or perceptual and is always intentional, private, singalar, immediate and indexical. The object-positing process produces knowledge by acquaintance or recognition, and produces the 'x' that occurs in propositions as a 'this something'.
By contrast, the computational property-attributing process is a quantifying and qualifying process, used by the mind to sort and classify data offered by the object-positing process for classification. The distinctive marks of this process are syntactic order, and use of concepts or categories. Property-attributing produces eternal, truth-functional propositions, and well-structured concepts. Computers, as syntactic engines, can mimic most of the mind's propertyattributing processes. But as John Searle has argued, in the case of human thinking, this is only part of the story.
The two processes are interactive, both with each other and with the thinker's environment. Replacing the current static view of knowledge, that represents knowledge of a proposition as a mind having apprehended an isomorphic structure of the world, with this, more dynamic view of thinking, will open new doors for research on the mind. This dynamic, interactive, operational view of mental processes provides an inherently more complex view of human thinking, that does a better job of accounting for the range and subtlety of thinking that humans actually do. On this view, an active human mind interacts constantly with a dynamic environment, both manipulating it through thought, and manipulated by it in sensation and experience.
|