斯坦福哲學百科全書詞條：意識（3）
原標題：斯坦福哲學百科全書詞條：意識（3）
Consciousness
9. Specific Theories of Consciousness
Although there are many general metaphysical/ontological theories of consciousness, the list of specific detailed theories about its nature is even longer and more diverse. No brief survey could be close to comprehensive, but seven main types of theories may help to indicate the basic range of options: higher-order theories, representational theories, interpretative narrative theories, cognitive theories, neural theories, quantum theories and nonphysical theories. The categories are not mutually exclusive; for example, many cognitive theories also propose a neural substrate for the relevant cognitive processes. Nonetheless grouping them in the seven classes provides a basic overview.
9.1 Higher-order theories
Higher-order (HO) theories analyze the notion of a conscious mental state in terms of reflexive meta-mental self-awareness. The core idea is that what makes a mental state M a conscious mental state is the fact that it is accompanied by a simultaneous and non-inferential higher-order (i.e., meta-mental) state whose content is that one is now in M. Having a conscious desire for some chocolate involves being in two mental states; one must have both a desire for some chocolate and also a higher-order state whose content is that one is now having just such a desire. Unconscious mental states are unconscious precisely in that we lack the relevant higher-order states about them. Their being unconscious consists in the fact that we are not reflexively and directly aware of being in them. (See the entry on higher-order theories of consciousness.)
Higher-order theories come in two main variants that differ concerning the psychological mode of the relevant conscious-making meta-mental states. Higher-order thought (HOT) theories take the required higher-order state to be an assertoric thought-like meta-state (Rosenthal 1986, 1993). Higher-order perception (HOP) theories take them to be more perception-like and associated with a kind of inner sense and intra-mental monitoring systems of some sort (Armstrong 1981, Lycan 1987, 1996).
Each has its relative strengths and problems. HOT theorists note that we have no organs of inner sense and claim that we experience no sensory qualities other than those presented to us by outer directed perception. HOP theorists on the other hand can argue that their view explains some of the additional conditions required by HO accounts as natural consequences of the perception-like nature of the relevant higher-order states. In particular the demands that the conscious-making meta-state be noninferential and simultaneous with its lower level mental object might be explained by the parallel conditions that typically apply to perception. We perceive what is happening now, and we do so in a way that involves no inferences, at least not any explicit personal-level inferences. Those conditions are no less necessary on the HOT view but are left unexplained by it, which might seem to give some explanatory advantage to the HOP model (Lycan 2004, Van Gulick 2000), though some HOT theorists argue otherwise (Carruthers 2000).
Whatever their respective merits, both HOP and HOT theories face some common challenges, including what might be called thegenerality problem. Having a thought or perception of a given item X—be it a rock, a pen or a potato—does not in general make X a conscious X. Seeing or thinking of the potato on the counter does not make it a conscious potato. Why then should having a thought or perception of a given desire or a memory make it a conscious desire or memory (Dretske 1995, Byrne 1997). Nor will it suffice to note that we do not apply the term 「conscious」 to rocks or pens that we perceive or think of, but only to mental states that we perceive or think of (Lycan 1997, Rosenthal 1997). That may be true, but what is needed is some account of why it is appropriate to do so.
The higher-order view is most obviously relevant to the meta-mental forms of consciousness, but some of its supporters take it to explain other types of consciousness as well, including the more subjective what it"s like and qualitative types. One common strategy is to analyze qualia as mental features that are capable of occurring unconsciously; for example they might be explained as properties of inner states whose structured similarity relations given rise to beliefs about objective similarities in the world (Shoemaker 1975, 1990). Though unconscious qualia can play that functional role, there need be nothing that it is like to be in a state that has them (Nelkin 1989, Rosenthal 1991, 1997). According to the HO theorist, what-it"s-likeness enters only when we become aware of that first-order state and its qualitative properties by having an appropriate meta-state directed at it.
Critics of the HO view have disputed that account, and some have argued that the notion of unconscious qualia on which it relies is incoherent (Papineau 2002). Whether or not such proposed HO accounts of qualia are successful, it is important to note that most HO advocates take themselves to be offering a comprehensive theory of consciousness, or at least the core of such a general theory, rather than merely one limited to some special meta-mental forms of it.
Other variants of HO theory go beyond the standard HOT and HOP versions including some that analyze consciousness in terms of dispositional rather than occurrent higher-order thoughts (Carruthers 2000). Others appeal to implicit rather than explicit higher-order understanding and weaken or remove the standard assumption that the meta-state must be distinct and separate from its lower-order object (Gennaro 1995, Van Gulick 2000, 2004) with such views overlapping with so called reflexive theories discussed in the section. Other variants of HO theory continue to be offered, and debate between supporters and critics of the basic approach remains active. (See the recent papers in Gennaro 2004.)
9.2 Reflexive theories
Reflexive theories, like higher-order theories, imply a strong link between consciousness and self-awareness. They differ in that they locate the aspect of self-awareness directly within the conscious state itself rather than in a distinct meta-state directed at it. The idea that conscious states involve a double intentionality goes back at least to Brentano (1874) in the 19th century. The conscious state is intentionally directed at an object outside itself—such as a tree or chair in the case of a conscious perception—as well as intentionally directed at itself. One and the same state is both an outer-directed awareness and an awareness of itself. Several recent theories have claimed that such reflexive awareness is a central feature of conscious mental states. Some view themselves as variants of higher-order theory (Gennaro 2004, 2012) while others reject the higher-order category and describe their theories as presenting a 「same-order」 account of consciousness as self-awareness (Kriegel 2009). Yet others challenge the level distinction by analyzing the meta-intentional content as implicit in the phenomenal first-order content of conscious states, as in so called Higher-Order Global State models (HOGS) (Van Gulick 2004,2006). A sample of papers, some supporting and some attacking the reflexive view can be found in Krigel and Williford (2006).
9.3 Representationalist theories
Almost all theories of consciousness regard it as having representational features, but so called representationalist theories are defined by the stronger view that its representational features exhaust its mental features (Harman 1990, Tye 1995, 2000). According to the representationalist, conscious mental states have no mental properties other than their representational properties. Thus two conscious or experiential states that share all their representational properties will not differ in any mental respect.
The exact force of the claim depends on how one interprets the idea of being 「representationally the same」 for which there are many plausible alternative criteria. One could define it coarsely in terms of satisfaction or truth conditions, but understood in that way the representationalist thesis seems clearly false. There are too many ways in which states might share their satisfaction or truth conditions yet differ mentally, including those that concern their mode of conceptualizing or presenting those conditions.
At the opposite extreme, one could count two states as representationally distinct if they differed in any features that played a role in their representational function or operation. On such a liberal reading any differences in the bearers of content would count as representational differences even if they bore the same intentional or representational content; they might differ only in their means or mode of representation not their content.
Such a reading would of course increase the plausibility of the claim that a conscious state"s representational properties exhaust its mental properties but at the cost of significantly weakening or even trivializing the thesis. Thus the representationalist seems to need an interpretation of representational sameness that goes beyond mere satisfaction conditions and reflects all the intentional or contentful aspects of representation without being sensitive to mere differences in underlying non-contentful features of the processes at the realization level. Thus most representationalists provide conditions for conscious experience that include both a content condition plus some further causal role or format requirements (Tye 1995, Dretske 1995, Carruthers 2000). Other representationalists accept the existence of qualia but treat them as objective properties that external objects are represented as having, i.e., they treat them as represented properties rather than as properties of representations or mental states (Dretske 1995, Lycan 1996).
Representationalism can be understood as a qualified form of eliminativism insofar as it denies the existence of properties of a sort that conscious mental states are commonly thought to have—or at least seem to have—namely those that are mental but not representational. Qualia, at least if understood as intrinsic monadic properties of conscious states accessible to introspection, would seem to be the most obvious targets for such elimination. Indeed part of the motivation for representationalism is to show that one can accommodate all the facts about consciousness, perhaps within a physicalist framework, without needing to find room for qualia or any other apparently non-representational mental properties (Dennett 1990, Lycan 1996, Carruthers 2000).
Representationalism has been quite popular in recent years and had many defenders, but it remains highly controversial and intuitions clash about key cases and thought experiments (Block 1996). In particular the possibility of inverted qualia provides a crucial test case. To anti-representationalists, the mere logical possibility of inverted qualia shows that conscious states can differ in a significant mental respect while coinciding representationally. Representationalists in reply deny either the possibility of such inversion or its alleged import (Dretske 1995, Tye 2000).
Many other arguments have been made for and against representationalism, such as those concerning perceptions in different sense modalities of one and the same state of affairs—seeing and feeling the same cube—which might seem to involve mental differences distinct from how the relevant states represent the world to be (Peacocke 1983, Tye 2003). In each case, both sides can muster strong intuitions and argumentative ingenuity. Lively debate continues.
9.4 Narrative Interpretative Theories
Some theories of consciousness stress the interpretative nature of facts about consciousness. According to such views, what is or is not conscious is not always a determinate fact, or at least not so independent of a larger context of interpretative judgments. The most prominent philosophical example is the Multiple Drafts Model (MDM) of consciousness, advanced by Daniel Dennett (1991). It combines elements of both representationalism and higher-order theory but does so in a way that varies interestingly from the more standard versions of either providing a more interpretational and less strongly realist view of consciousness.
The MDM includes many distinct but interrelated features. Its name reflects the fact that at any given moment content fixations of many sorts are occurring throughout the brain. What makes some of these contents conscious is not that they occur in a privileged spatial or functional location—the so called 「Cartesian Theater」—nor in a special mode or format, all of which the MDM denies. Rather it a matter of what Dennett calls 「cerebral celebrity」, i.e., the degree to which a given content influences the future development of other contents throughout the brain, especially with regard to how those effects are manifest in the reports and behaviors that the person makes in response to various probes that might indicate her conscious state. One of the MDM"s key claims is that different probes (e. g., being asked different questions or being in different contexts that make differing behavioral demands) may elicit different answers about the person"s conscious state. Moreover, according to the MDM there may be no probe-independent fact of the matter about what the person"s conscious state really was. Hence the 「multiple」 of the Multiple Drafts Model.
The MDM is representationalist in that it analyzes consciousness in terms of content relations. It also denies the existence of qualia and thus rejects any attempt to distinguish conscious states from nonconscious states by their presence. It rejects as well the notion of the self as an inner observer, whether located in the Cartesian Theater or elsewhere. The MDM treats the self as an emergent or virtual aspect of the coherent roughly serially narrative that is constructed through the interactive play of contents in the system. Many of those contents are bound together at the intentional level as perceptions or fixations from a relatively unified and temporally extended point of view, i.e., they cohere in their contents as if they were the experiences of a ongoing self. But it is the order of dependence that is crucial to the MDM account. The relevant contents are not unified because they are all observed by a single self, but just the converse. It is because they are unified and coherent at the level of content that they count as the experiences of a single self, at least of a single virtual self.
It is in this respect that the MDM shares some elements with higher-order theories. The contents that compose the serial narrative are at least implicitly those of an ongoing if virtual self, and it is they that are most likely to be expressed in the reports the person makes of her conscious state in response to various probes. They thus involve a certain degree of reflexivity or self-awareness of the sort that is central to higher-order theories, but the higher-order aspect is more an implicit feature of the stream of contents rather than present in distinct explicit higher-order states of the sort found in standard HO theories.
Dennett"s MDM has been highly influential but has also drawn criticism, especially from those who find it insufficiently realist in its view of consciousness and at best incomplete in achieving its stated goal to fully explain it (Block 1994, Dretske 1994, Levine 1994). Many of its critics acknowledge the insight and value of the MDM, but deny that there are no real facts of consciousness other than those captured by it (Rosenthal 1994, Van Gulick 1994, Akins 1996).
From a more empirical perspective, the neuroscientist Michael Gazzaniga (2011) has introduced the idea of an 「interpreter module」 based in the left hemisphere that makes sense of our actions in any inferential way and constructs an ongoing narrative of our actions and experience. Though the theory is not intended as a complete theory of consciousness, it accords a major role to such interpretative narrative activity.
9.5 Cognitive Theories
A number theories of consciousness associate it with a distinct cognitive architecture or with a special pattern of activity with that structure.
Global Workspace. A major psychological example of the cognitive approach is the Global Workspace theory. As initially developed by Bernard Baars (1988)) global workspace theory describes consciousness in terms of a competition among processors and outputs for a limited capacity resource that 「broadcasts」 information for widespread access and use. Being available in that way to the global workspace makes information conscious at least in the access sense. It is available for report and the flexible control of behavior. Much like Dennett"s 「cerebral celebrity」, being broadcast in the workspace makes contents more accessible and influential with respect to other contents and other processors. At the same time the original content is strengthened by recurrent support back from the workspace and from other contents with which it coheres. The capacity limits on the workspace correspond to the limits typically placed on focal attention or working memory in many cognitive models.
The model has been further developed with proposed connections to particular neural and functional brain systems by Stanislas Dehaene and others (2000). Of special importance is the claim that consciousness in both the access and phenomenal sense occurs when and only when the relevant content enters the larger global network involving both primary sensory areas as well as many other areas including frontal and parietal areas associated with attention. Dehaene claims that conscious perception begins only with the 「ignition」 of that larger global network; activity in the primary sensory areas will not suffice no matter how intense or recurrent (though see the contrary view of Victor Lamme in section 9.7).
Attended Intermediate Representation. Another cognitive theory is Jesse Prinz"s (2012) Attended Intermediate level Representation theory (AIR). The theory is a neuro-cognitive hybrid account of conscious. According to AIR theory, a conscious perception must meet both cognitive and neural conditions. It must be a representation of a perceptually intermediate property which Prinz argues are the only properties of which we are aware in conscious experience—we experience only basic features of external objects such as colors, shapes, tones, and feels. According to Prinz, our awareness of higher level properties—such as being a pine tree or my car keys—is wholly a matter of judging and not of conscious experience. Hence the Intermediate Representational (IR) aspect of AIR. To be conscious such a represented content must also be Attended (the A aspect of AIR). Prinz proposes a particular neural substrate for each component. He identifies the intermediate level representations with gamma (40–80hz) vector activity in sensory cortex and the attentional component with synchronized oscillations that can incorporate that gamma vector activity.
9.6 Information Integration Theory
The integration of information from many sources is an important feature of consciousness and, as noted above (section 6.4), is often cited as one of its major functions. Content integration plays an important role in various theories especially global workspace theory (section 9.3). However, a proposal by the neuroscientist Giulio Tononi (2008) goes further in identifying consciousness with integrated information and asserting that information integration of the relevant sort is both necessary and sufficient for consciousness regardless of the substrate in which it is realized (which need not be neural or biological). According to Tononi"s Integrated Information Theory (IIT), consciousness is a purely information-theoretic property of systems. He proposes a mathematical measure φ that aims to measure not merely the information in the parts of a given system but also the information contained in the organization of the system over and above that in its parts. φ thus corresponds to the system"s degree of informational integration. Such a system can contain many overlapping complexes and the complex with the highest φ value will be conscious according to IIT.
According to IIT, consciousness varies in quantity and comes in many degrees which correspond to φ values. Thus even a simple system such a single photo diode will be conscious to some degree if it is not contained within a larger complex. In that sense, IIT implies a form of panpsychism that Tononi explicitly endorses. According to IIT, the quality of the relevant consciousness is determined by the totality of informational relations within the relevant integrated complex. Thus IIT aims to explain both the quantity and quality of phenomenal consciousness. Other neuroscientists, notably Christof Koch, have also endorsed the IIT approach (Koch 2012).
9.7 Neural Theories
Neural theories of consciousness come in many forms, though most in some way concern the so called 「neural correlates of consciousness」 or NCCs. Unless one is a dualist or other non-physicalist, more than mere correlation is required; at least some NCCs must be the essential substrates of consciousness. An explanatory neural theory needs to explain why or how the relevant correlations exist, and if the theory is committed to physicalism that will require showing how the underlying neural substrates could be identical with their neural correlates or at least realize them by satisfying the required roles or conditions (Metzinger 2000).
Such theories are diverse not only in the neural processes or properties to which they appeal but also in the aspects of consciousness they take as their respective explananda. Some are based on high-level systemic features of the brain, but others focus on more specific physiological or structural properties, with corresponding differences in their intended explanatory targets. Most in some way aim to connect with theories of consciousness at other levels of deion such as cognitive, representational or higher-order theories.
A sampling of recent neural theories might include models that appeal to global integrated fields (Kinsbourne), binding through synchronous oscillation (Singer 1999, Crick and Koch 1990), NMDA-mediated transient neural assemblies (Flohr 1995), thalamically modulated patterns of cortical activation (Llinas 2001), reentrant cortical loops (Edelman 1989), comparator mechanisms that engage in continuous action-prediction-assessment loops between frontal and midbrain areas (Gray 1995), left hemisphere based interpretative processes (Gazzaniga 1988), and emotive somatosensory hemostatic processes based in the frontal-limbic nexus (Damasio 1999) or in the periaqueductal gray (Panksepp 1998).
In each case the aim is to explain how organization and activity at the relevant neural level could underlie one or another major type or feature of consciousness. Global fields or transient synchronous assemblies could underlie the intentional unity of phenomenal consciousness. NMDA-based plasticity, specific thalamic projections into the cortex, or regular oscillatory waves could all contribute to the formation of short term but widespread neural patterns or regularities needed to knit integrated conscious experience out of the local activity in diverse specialized brain modules. Left hemisphere interpretative processes could provide a basis for narrative forms of conscious self-awareness. Thus it is possible for multiple distinct neural theories to all be true, with each contributing some partial understanding of the links between conscious mentality in its diverse forms and the active brain at its many levels of complex organization and structure.
One particular recent controversy has concerned the issue of whether global or merely local recurrent activity is sufficient for phenomenal consciousness. Supporters of the global neuronal workspace model (Dehaene 2000) have argued that consciousness of any sort can occur only when contents are activated with a large scale pattern of recurrent activity involving frontal and parietal areas as well as primary sensory areas of cortex. Others in particular the psychologist Victor Lamme (2006) and the philosopher Ned Block (2007) have argued that local recurrent activity between higher and lower areas within sensory cortex (e.g. with visual cortex) can suffice for phenomenal consciousness even in the absence of verbal reportability and other indicators of access consciousness.
9.8 Quantum theories
Other physical theories have gone beyond the neural and placed the natural locus of consciousness at a far more fundamental level, in particular at the micro-physical level of quantum phenomena. According to such theories, the nature and basis of consciousness can not be adequately understood within the framework of classical physics but must be sought within the alternative picture of physical reality provided by quantum mechanics. The proponents of the quantum consciousness approach regard the radically alternative and often counterintuitive nature of quantum physics as just what is needed to overcome the supposed explanatory obstacles that confront more standard attempts to bridge the psycho-physical gap.
Again there are a wide range of specific theories and models that have been proposed, appealing to a variety of quantum phenomena to explain a diversity of features of consciousness. It would be impossible to catalog them here or even explain in any substantial way the key features of quantum mechanics to which they appeal. However, a brief selective survey may provide a sense, however partial and obscure, of the options that have been proposed.
The physicist Roger Penrose (1989, 1994) and the anesthesiologist Stuart Hameroff (1998) have championed a model according to which consciousness arises through quantum effects occurring within subcellular structures internal to neurons known as microtubules. The model posits so called 「objective collapses」 which involve the quantum system moving from a superposition of multiple possible states to a single definite state, but without the intervention of an observer or measurement as in most quantum mechanical models. According to the Penrose and Hameroff, the environment internal to the microtubules is especially suitable for such objective collapses, and the resulting self-collapses produce a coherent flow regulating neuronal activity and making non-algorithmic mental processes possible.
The psychiatrist Ian Marshall has offered a model that aims to explain the coherent unity of consciousness by appeal to the production within the brain of a physical state akin to that of a Bose-Einstein condensate. The latter is a quantum phenomenon in which a collection of atoms acts as a single coherent entity and the distinction between discrete atoms is lost. While brain states are not literally examples of Bose-Einstein condensates, reasons have been offered to show why brains are likely to give rise to states that are capable of exhibiting a similar coherence (Marshall and Zohar 1990).
A basis for consciousness has also been sought in the holistic nature of quantum mechanics and the phenomenon of entanglement, according to which particles that have interacted continue to have their natures depend upon each other even after their separation. Unsurprisingly these models have been targeted especially at explaining the coherence of consciousness, but they have also been invoked as a more general challenge to the atomistic conception of traditional physics according to which the properties of wholes are to be explained by appeal to the properties of their parts plus their mode of combination, a method of explanation that might be regarded as unsuccessful to date in explaining consciousness (Silberstein 1998, 2001).
Others have taken quantum mechanics to indicate that consciousness is an absolutely fundamental property of physical reality, one that needs to be brought in at the very most basic level (Stapp 1993). They have appealed especially to the role of the observer in the collapse of the wave function, i.e., the collapse of quantum reality from a superposition of possible states to a single definite state when a measurement is made. Such models may or may not embrace a form of quasi-idealism, in which the very existence of physical reality depends upon its being consciously observed.
There are many other quantum models of consciousness to be found in the literature—some advocating a radically revisionist metaphysics and others not—but these four provide a reasonable, though partial, sample of the alternatives.
9.9 Non-physical theories
Most specific theories of consciousness—whether cognitive, neural or quantum mechanical—aim to explain or model consciousness as a natural feature of the physical world. However, those who reject a physicalist ontology of consciousness must find ways of modeling it as a nonphysical aspect of reality. Thus those who adopt a dualist or anti-physicalist metaphysical view must in the end provide specific models of consciousness different from the five types above. Both substance dualists and property dualists must develop the details of their theories in ways that articulate the specific natures of the relevant non-physical features of reality with which they equate consciousness or to which they appeal in order to explain it.
A variety of such models have been proposed including the following. David Chalmers (1996) has offered an admittedly speculative version of panpsychism which appeals to the notion of information not only to explain psycho-physical invariances between phenomenal and physically realized information spaces but also to possibly explain the ontology of the physical as itself derived from the informational (a version of 「it from bit」 theory). In a somewhat similar vein, Gregg Rosenberg has (2004) proposed an account of consciousness that simultaneously addresses the ultimate categorical basis of causal relations. In both the causal case and the conscious case, Rosenberg argues the relational-functional facts must ultimately depend upon a categorical non-relational base, and he offers a model according to which causal relations and qualitative phenomenal facts both depend upon the same base. Also, as noted just above (section 9.8), some quantum theories treat consciousness as a fundamental feature of reality (Stapp 1993), and insofar as they do so, they might be plausibly classified as non-physical theories as well.
10. Conclusion
A comprehensive understanding of consciousness will likely require theories of many types. One might usefully and without contradiction accept a diversity of models that each in their own way aim respectively to explain the physical, neural, cognitive, functional, representational and higher-order aspects of consciousness. There is unlikely to be any single theoretical perspective that suffices for explaining all the features of consciousness that we wish to understand. Thus a synthetic and pluralistic approach may provide the best road to future progress.
Bibliography
TAG:哲學園 |