Note: Classmate Jalinus was the first to make a PPT since the course began (though the PPT still wasn’t very good, with too many long quoted passages), and the notes he整理ed after class were also very认真, so I hereby commend him~ I was the one who took over for Chapter Five, and in the end I also appended the outline I had sorted out.
Discussion-class notes by Jalinus
November 28, 2017
Chapter One: The View from Scientific Realism
One of Churchland’s major aims in this book is to sketch a new blueprint for human understanding. If one cannot provide an account of human understanding, one cannot speak glibly of the success of theories, and one likewise cannot evaluate the relation between science and reality (on the three dimensions of metaphysics, semantics, and epistemology).
The traditional distinction between theoretical and non-theoretical concepts, and the priority accorded to non-theoretical concepts, is also arbitrary in his view. Our common-sense framework must accept new criticism and challenges, even at the level of perception. The classical empiricist conception of objectivity, as well as foundationalist justification in epistemology, are unsatisfactory. A more promising path is to develop our perceptual and conceptual frameworks.
Chapter Two: The Plasticity of Perception
Churchland first constructs a thought experiment to clarify that the semantics of observational predicates do not derive from sensation and perception themselves, but depend on networks of belief. Using temperature as an example, he imagines a race V that can grasp “temperature” through vision. Race V is characterized by sensitivity to electromagnetic radiation; vision senses temperature through “brightness”; they lack a color vocabulary and have no touch. Aside from this physiological difference, they are entirely the same as human beings, and their perceptual reports are also in fact accurate (for V and the normal human race H, the extension of “hot things” is the same, though the intension differs). If we think that the meaning of observational terms is given in sensation, then we will think that the “cold, warm, hot” spoken by race V actually mean “black, gray, white.” Insisting on the rule of异音 translation guided by sensation would make many of their beliefs—such as “fire is white”—false. The predicament brought by异音 translation reveals that the semantics of observational predicates are not given by perception. The principle of同音 translation will reveal that, if temperature is taken as a common external factor, then the difference between race V and us is only in the methods by which this factor is discovered; there is systematic agreement between the two. Churchland is more inclined to accept the latter conclusion. From this we can imagine richer and more effective sensory apparatuses to develop the network of beliefs that underlies human understanding.
He then proposes a distinction between the understanding of observational words and their non-inferential use, and thereby gives the distinction between objective intentionality and subjective intentionality. From this he derives three implications: what a person is physiologically capable of perceiving is a matter of the objective intentionality of his senses; what he thinks the objective world presents is a matter of his subjective intentionality; objective intentionality and subjective intentionality are independent of each other, and physiological setup as well as sensation are merely causal intermediaries, replaceable, and need not entail any correction of the human sense organs. With the help of new patterns of conceptual response, the domain of human perceptual consciousness can be massively revised and expanded.
Immediately afterward he imagines the thought experiment of the caloric-race, used to display cases of mistaken perceptual development, and contrasts it with the caloric theory of heat and the molecular theory of motion, so as to reflect on our present situation. We cannot prove that caloric does not exist, nor is it easy to make the caloric-race believe that their theory is groundless and systematically mistaken. Perhaps one can show them the relative weakness of the explanatory power of the caloric theory and its own incoherence, but the caloric-race could defuse this through new ad hoc hypotheses. Through this Churchland demonstrates the conceptual network as theory in itself and its cognitive inertia. On this basis, if one is to ask whether the world instantiates observational predicates, one must first ask whether the theory containing this predicate is true—that is, whether it is appropriate and richly explanatory. Our perception may well from the outset have misrepresented reality, and as for how we are to understand the world better, that requires the continual development of our perception. He will leave discussion of the evaluation of different theories to Chapter Three.
Through the above discussion, Churchland rejects the sharp distinction within the human set of beliefs between a set of basic beliefs and other beliefs. Perception cannot provide conceptually neutral information. But this does not mean that human understanding of the world is detached from reality; he believes there remains a systematic causal relation between the two. Modern science can provide a new picture of the world, but this does not mean it must be able to reduce our common-sense cognition. It can fully introduce large numbers of new explanatory items while eliminating old ones.
Chapter Three: The Plasticity of Understanding
A key advance in this chapter is that Churchland introduces the distinction between semantic significance and systematic significance, and thereby reinterprets, or rather dissolves, the traditional distinction between the analytic and the synthetic. The latter does not satisfactorily explain the nature of some propositions that lie in an intermediate zone, but this distinction still captures some substantive content—in Churchland’s view, issues such as the source of meaning and the status and relations of propositions within a particular belief network.
He proposes a semi-holistic theory of meaning, holding that the truth of a sentence is closely related to the meanings of the words it contains, and that the latter depend on the language framework in which they are embedded. Semantic significance is determined by the linguistic community, whereas systematic significance depends on the degree of tightness in the relations between a sentence and other sentences under the same language framework. Churchland gives examples, respectively, of the phlogiston theory and the truth-value evolution of statements like “whales are fish,” in order to show that the traditional distinction between the empirical and the non-empirical is inappropriate.
In Churchland’s view, what allows two speakers to understand each other and communicate fluently and effectively is the systematic similarity between the sets of sentences they each accept, and the corresponding similarity in the substantive inferences they tend to draw from them. Their having the same understanding of term T merely requires that those “preferred” subsets within the respective sentence sets containing T have the same systematic significance in their respective sentence sets; in other words, it means that there is a translation between the terms T in the two speakers’ respective sentence sets that preserves the same understanding, and this translation preserves the systematic significance of the corresponding sentence sets. In his view, the goal of translation is to discover within the target language an intensional structure isomorphic to the translator’s own linguistic structure. Competing with this translation theory is the view proposed by Quine, who holds that translation is a mapping that preserves dispositions to verbal behavior, matching observational sentences across different languages so that speakers have the same dispositions to assent to or dissent from the corresponding sentences. Churchland uses the earlier thought experiment to show that observability and translation are actually unrelated, and that the former is a redundant requirement. He then compares his own translation theory with Davidson’s translation theory, noting their similarities and differences.
A greater challenge comes from Putnam, who argues that there is a substantive distinction between the meaning of a given term and the speaker’s background set of beliefs; the meaning of a term is not fixed or determined primarily by its intension, but by some indexical formula that points toward the external world. This explanation comes from the well-known Twin Earth thought experiment, in which the intension of “water” and “watter” is the same, but their extensions differ; therefore extension is an independent vector component of meaning, as important as intension. What Churchland wants to refute, however, is that we can accept that the meanings of natural-kind terms contain some important indexical component, but need not think that it is independent of intension, because the expression “having the same properties” used in the indexical formula is in fact nothing more than a reiteration of the universal features in its intension, and like other components of intension it too cannot avoid the possibility of being reconfigured at any time by new discoveries and understandings.
The final important issue in Chapter Three is the issue of incommensurability. This issue concerns the mutual understanding and evaluation of different theories. Traditionally, incommensurable theories are those that are logically unrelated, compatible with one another, yet systematically differ in the meanings of terms within each theory, and thus are difficult to compare directly. Through the theory of meaning and theory of understanding proposed earlier, Churchland likewise turns incommensurability into a matter of degree. In his view, even if one imagines the extreme case in which two theories are entirely incommensurable, they can still be internally evaluated on the basis of inductive coherence, explanatory completeness, informational richness, and so on; in such a case, making a choice is still possible. Understanding does not require semantic identity, and the same is true in the reduction relations between different theories. What Churchland seeks to refute is the traditional view that a new theory must defend its own superiority by reducing the old theory and inheriting all of its empirical evidence. He believes the main fallacy here lies in mistakenly thinking that a reduction map, like a translation map, also requires synonymic matching. Churchland then gives a principle of correspondence for ideal reduction, for example: terms in T0 map to a subset of expressions in Tn, and are used well as singular terms; the main principles in T0 (those with semantic significance or systematic significance) are mapped to universal sentences in Tn. One need not require Sn to have the same importance; one only needs them to be consequences of the main principles of Tn; an ordered substitution is achieved; Tn and T0 are coordinated with the larger background of Tn in the same way, and accomplish all the same predictive and explanatory functions that T0 accomplishes; Tn inherits the confirmability of T0, and so on. But it must be added that what is derived from the principles of Tn is not T0, but its equiform image; the matching of correspondence rules need not be interpreted as an assertion of identity, nor even as substantive equivalence; correspondence rules can be regarded as ordered pairs of expressions; the image of T0 in Tn may not be a complete or very faithful image of T0. The following situation may arise: certain important principles of T0 map to false sentences in Tn. For example, people often think that special relativity reduced classical mechanics, and that the laws of the latter can be derived from the former; but in fact what special relativity derives is only an image of classical mechanics within the former, and this image may differ from classical mechanics in the meaning of basic categories and concepts. Concepts such as mass and length are unary predicates in classical mechanics, whereas in special relativity they are binary. Thus, rather than saying that “T0 has found an appropriate image in Tn,” the stricter formulation is: “some theory Tn’ very similar to T0 has found an appropriate image in Tn.” As an image of T0 expressed in the vocabulary of Tn, Sn may not be a pure consequence of the basic principles of Tn; this is only the case when we include some limiting condition or counterfactual assumption in the premises. More generally, reduction between theories is only a matter of degree; a true theory can reduce a false theory, a reduction is essentially not a reiteration and defense of the theory being reduced, and a theory can even reduce an incommensurable competing theory. From this Churchland draws the overall conclusion of Chapter Three: that a proper epistemology must from the very beginning acknowledge the radical plasticity of human understanding, even the possibility of its total formal transformation.
Chapter Four: Self-Conception and the Mind-Body Problem
On the basis of the first three chapters, Churchland turns to discussions in philosophy of mind, in order to show how his earlier theory can reshape some important traditional and contemporary debates, including the mind-body problem, the problem of other minds, and issues of self-conception. He believes that self-conception does not have any essentially privileged status; it too is a product of conceptual development. Therefore, judging others’ mental states from one’s own condition is an unreliable induction, leaving room for skepticism.
In his view, the basic error of behaviorism lies in denying the empirical nature of the “psychological/physiological” general laws, while also assuming that there exists a one-to-one correspondence between mental states and external behavior. Churchland believes that internal states and external behavior are linked through infinitely complex mediating relations. As for some issues of self-conception, he believes that there is no necessary relation between a mental state P and one’s judgment of this state P’; for example, an infant, even without a conceptual framework, can still respond to stimuli, and that conceptual framework is necessary for his judgment of his own state. Sensation and judgment are two different things, at most accidentally linked together. Therefore, those self-judgments thought to be incorrigible are unjustified; there are vast numbers of intermediate mental states, and the relation between the state itself and the judgment of that state is varied and complex.
As for the mind-body problem, Churchland is well known for his eliminativist materialist stance. He believes that dualism faces an improper problem of theoretical reduction, and cannot provide any conceptual resources that compete with materialism; it is merely negation on the basis of materialism. Churchland believes that an appropriate theory of the central nervous system will replace traditional theories, and that the former can provide us with more resources for prediction and explanation through its deeper understanding of neurophysiological structure and activity.
Chapter Five: Sentence Epistemology and the Natural Sciences of Cognitive Machines
This chapter is rather difficult, and was lectured by Professor Hu.
The following is Teacher Hu’s PPT:
17.Normative epistemology (vs. descriptive/dynamic/explanatory)
- Epistemological principles → (normative, guiding) → the evolution of the world-picture
- Children are not explicit; adults rarely use it (real people, explanatory)
- By what principles does an ideal person guide their cognitive behavior? (normative)
- Normative epistemology (the idealized rational agent) is separated from the problem of dynamics and the problem of explanation (130-131)
- Mathematical probability calculus is used in an attempt to clarify… but it does not at all imply that… normal people in fact secretly use this calculus… (131) The context of discovery is separated from the context of justification
- The premises assumed by normative epistemology (such as the commonsense P-theory concerning introspective consciousness) are questionable (not only a priori, but also developmental, Chapter 4)
- a priori epistemology → naturalized epistemology
18.Orthodox epistemology = sentence kinematics
- The task of normative epistemology: the cognitive machine continuously improves its cognitive state
- Sentence kinematics: (ISA, ideal sentence automaton)
- Cognitive state = a set of a series of sentences or propositions.
- The process of improvement = inputting new observation sentences or hypothesis sentences
- The result of improvement = the old set of sentences is refreshed into a new set of sentences
- Epistemology = the study of the relation R between input sentences and sentence sets
- The premise of ISA: conceiving the cognitive subject (the self) as a set of a series of sentences (such as x believes p)
- The author opposes sentence epistemology
19.Infants as a counterexample
- The cognitive development of infants cannot be represented by ISA (infants simply do not have propositional attitudes)
- There is no essential boundary between adults and infants; it is continuous development
- Infants’ cognitive activity does not require language; language is acquired after birth
- → the participation of language in adults’ cognitive activity is nonessential, acquired, and derivative
20.The role of language
- The specificity of human language is only one kind of ability; other higher animals that do not possess language ability also have cognitive activity. Intelligent beings that may exist in the universe probably would not have an information-exchange medium similar to human language either.
- The structure of language plays a relatively unimportant role in cognitive activity (145)
- Language = a high-level abstraction of the state of neurons in the brain (the primitive informational state) = top-level information exchange
- The abstraction from primitive information to sentence information requires postnatal learning. Because the mode of abstraction is similar among similar groups of people, effective communication can be carried out through language. But this is a derivative, plastic, rather than primitive, fixed cognitive pattern. Non-linguistic information-processing systems are more fundamental
The plasticity of language (language evolution 148)
- Observational predicates and theoretical predicates, analytic judgments and synthetic judgments… are all continuous, evolvable; there is no sentence that can stand on a fixed foundation.
- This does not mean sliding into nihilism or anarchism (Feyerabend)
▫ Cognitive development is not arbitrary; causal relations can be grasped by the natural sciences (what kind of growth determines what kind of cognition)
21.Naturalization of epistemology
Natural items (natural items, natural things?) are all “information banks” that record changes in the outside world, differing only in degree: from a stone to an animal to a human, all are carriers of information.
What distinguishes a cognitive machine from a stone: “learning”
Sensory input → motor output (first-order function)
Strengthening one input-output in a specific direction (second-order function) = learning
First order: rain → wet ground ≈ eating → satiety (information recording)
Second order: a stone cannot improve itself from being rained on; animals can learn to forage (information development)
▫ Using information to obtain more information, “reinvestment,” compound interest
▫ Causal relations rather than sentence relations, an “adaptation” process rather than a process of logical derivation
▫ Constantly adjusting the feedback mechanism to adapt to external changes (take fitting the tidal curve as an example)
▫ This constantly fitting process can be bodily memory rather than sentence records; sentence records are only one example of a feedback mechanism.
▫ The human brain producing propositions in response to external stimuli, and flowers, plants, and trees producing tree rings or nectar in response to sunlight, are the same process.
Translated from the Chinese original with AI assistance. The original text is authoritative.
Leave a Reply