1 What Did Mary Know? Russellian Monism Without Intrinsics Jussi Jylkkä Note

Home , Global workspace theory, Knowledge argument, Pain (philosophy)

What did Mary know? Russellian Monism Without Intrinsics

Jussi Jylkkä Note: this is a non-peer reviewed preprint.

Abstract. The Mary thought experiment aims to demonstrate that science cannot capture what experiences feel like. Russellian Monism (RM) avoids this problem by claiming that phenomenality is an intrinsic (non-relational and non-dispositional) property of matter and beyond the scope of science, which is limited to describing extrinsic (relational and dispositional) properties. Against RM, I argue that metaphysical intrinsicality is not compatible with neuroscientific theories where experiences are considered as causal processes. Second, I argue that if intrinsic properties have causal power, they can also affect neuroscientific measuring devices and be scientifically modeled. Thus, intrinsic properties are not inscrutable, as RM holds. In the third part of the article, I sketch the outlines of RM without intrinsics. I propose that the core Kantian thesis of RM about limits of science can be maintained without postulating metaphysical intrinsics. I argue that metaphysical intrinsicality can be replaced with Weak Intrinsicality, meaning model-independence. Science is confined to observations and models, whereas an experience is the concrete, model- independent process that produces observations of its neural mechanisms. On this account, the epistemic gap is difference between a model and the modeled.

Keywords: Knowledge argument, Hard problem of consciousness, Russellian Monism, Kant, Transcendental idealism, Intrinsic properties, Neural correlates of consciousness, Constitutive mechanisms of consciousness

1. Mary and the limits of science Jackson's (1982) Knowledge Argument, also known as the Mary thought experiment, is one of the most influential arguments against physicalism, i.e., the view that experiences are physical phenomena. It aims to demonstrate that science cannot afford knowledge about certain aspects of experiences, namely what it feels like to undergo them (see also Feigl, 1958; Maxwell, 1965; Meehl, 1966; Nagel, 1979; Robinson, 1982). More generally, the argument aims to show that physicalism is false: because science doesn’t afford knowledge of what experiences feel like, that feel cannot be physical. If it were, science could describe it like it describes all other physical phenomena. Jackson (1982, p. 130) formulates the thought experiment as follows:

Mary is a brilliant scientist who is, for whatever reason, forced to investigate the world from a black and white room via a black and white television monitor. She specializes in the neurophysiology of vision and acquires, let us suppose, all the physical information there is to obtain about what goes on when we see ripe tomatoes, or the sky, and use terms like ‘red’, ‘blue’, and so on. She discovers, for example, just which wavelength combinations from the sky stimulate the retina, and exactly how this produces via the central nervous system the contraction of the vocal chords and expulsion of air from the lungs that results in the uttering of the sentence ‘The sky is blue’ […] What will happen when Mary is released from her black and white room or is given a color television monitor? Will she learn anything or not? It seems just obvious that she will learn something about the world and our visual experience of it. But then is it inescapable that her previous knowledge was incomplete. But she had all the physical information. Ergo there is more to have than that, and Physicalism is false.

Despite her complete scientific knowledge, Mary does not know what seeing red feels like: upon leaving the room, she learns the phenomenal or qualitative character of red experiences. There are a range of responses to the Mary argument, from purely epistemic or philosophy of language –based (e.g., the phenomenal concepts strategy; Stoljar, 2005), through naturalistic ones (e.g., the ability hypothesis; Nemirow, 1980), to heavily metaphysical ones (e.g., Russellian monism [Goff, 2017; Russell, 1927; Strawson, 2003] or property dualism [Chalmers, 1996]). Here I focus on Russellian Monism (RM), whose most central claim I take to be Kantian: science is in some sense limited. On the standard Russellian view, the limits of science are assumed to stem from what types of properties it deals with: Science can only model extrinsic (relational or causal-dispositional) properties of matter, but not their intrinsic categorical basis. Phenomenal properties, in turn, are assumed to form the intrinsic or categorical basis of causal dispositions, which, unlike the dispositions themselves, are beyond science (Alter & Nagasawa, 2012; Goff, 2017; Russell, 1927). These properties are also called quiddities (Chalmers, 1996) or inscrutables (Montero, 2015). I argue that we can accept the core Kantian claim of RM that science is limited without postulating metaphysical intrinsics. As an alternative to RM’s notion of intrinsicality, I propose Weak Intrinsicality which means model- or representation-independence: science has no access to reality beyond observations and models. On my approach, an experience is the concrete process that produces observations of its neural mechanisms. On the account I put forth, science can indeed model all aspects of experiences, including what they feel like, but only model. Thus, the epistemic gap between an experience and its scientific model doesn’t reflect an ontological difference, but rather only the difference between a model and the modeled. I will proceed as follows. The first part of the paper (§2) is negative and presents against RM’s notion of intrinsics two problems. First, I argue that the notion of intrinsics as non-relational and non- dispositional is incompatible with neuroscientific theories which consider the neural mechanisms of consciousness as relational processes. A plausible approach for the Russellian is to hold that the phenomenality that we are familiar with is the intrinsic nature of what neuroscience models as constitutive mechanisms of experiences (CMEs). However, given that the CMEs are relational processes, they cannot have intrinsic (non-relational) natures. This is closely related to the combination problem of RM (Chalmers, 2017; Goff, 2017), i.e., the problem of how the intrinsic natures of fundamental particles combine to form human experiences. My second argument against the Russellian notion of intrinsics relates to the claim that they are supposed to be scientifically inscrutable; something that science cannot describe (Montero, 2015; see also Goff, 2017; Strawson, 2006, 2019). I present a dilemma against RM: If intrinsics are causally efficacious,

2 then they can also affect the senses and measuring devices of observers. Thus, intrinsics are not inscrutable as RM would hold. Again, if intrinsics are not causally efficacious, then they can’t be scientifically observed and modeled, but neither is there mental causation, given that phenomenal properties are (a subset of) intrinsic properties. The take-home message is that one cannot integrate intrinsics or phenomenal properties in causal processes and simultaneously hold that they are in some strong sense private: it they have causal power we can observe them through their effects. These problems stem from the metaphysically strong notion of intrinsics in RM. On my view, the postulation of metaphysically distinct class of properties (intrinsic and extrinsic) is not justified at least by the Mary case alone—it is an epistemic problem that requires an epistemic explanation. In the second part of the paper (§3), I argue that the core Kantian thesis of RM can be preserved without postulating metaphysical intrinsics. Whereas traditional RM resembles metaphysical interpretations of Kant (e.g., Langton, 2001), I propose an alternative account that is similar to epistemic interpretations (e.g., Allais, 2015; Allison, 2004). I suggest that we replace RM’s metaphysically strong notion of intrinsicality with the epistemic notion of Weak Intrinsicality. Whereas RM’s metaphysical intrinsicality means a thing considered independent from all its relational properties, an object’s weakly intrinsic nature is the way it is independent of our modes of representing it. As Allais (2015) argues, we can only know objects as they manifest to us in experience, not as they are independently of being related to us. The way objects appear to us is shaped by our conscious-cognitive system, types of measuring devices, and the theories we use to conceptualize our observations. We cannot abstract ourselves away from how objects appear to us, since the appearance is determined not only by the observed object, but also ourselves. In short, we can’t represent objects as they are independent of our representing them. On the account I propose, science is limited simply because it is based on observations and models of external phenomena; science cannot say anything of objects beyond scientific representations. Again, we can know our experiences independently of models and observations. I call the nature of an object in abstraction from how it is represented by an observer its “weakly intrinsic nature”. I argue that Weak Intrinsicality can do the same explanatory work as the traditional Russellian metaphysical notion of intrinsicality but is more parsimonious and compatible with neuroscience of consciousness. On my approach, science is limited simply because it is “shaped by the lens of human consciousness” (Hawking & Mlodinow, 2010) or based on “pointer-readings” (Eddington, 1929): Any scientific model is an abstraction in the mind of a scientist, based on observations caused by something “out there”, independent of our observations and models. I conclude by responding to possible criticisms that the notion of Weak Intrinsicality is either too weak to explain Mary’s epistemic situation, or too strong and incompatible with scientific realism.

1.1. Experiences and their constitutive mechanisms On my terminology, neuroscience of consciousness models the constitutive mechanisms (CM) of experiences, not just their neural correlates. According to RM, the phenomenal feel of E is part of the intrinsic nature of CME, so Russellians would arguably consider the terms “E” and “CME” as coextensional (at least in the actual world1). “Constitutive mechanism” means that empirically observable process that correlates perfectly with the experience, is fully isomorphic with the experience (so that every change in the experience E corresponds to a change in the CME), and explains how the experience (CME) is based on lower-level constituent processes (Jylkkä & Railo, 2019; Revonsuo, 2006). It is more common to speak of the neural correlates of consciousness (NCC) (Koch, Massimini, Boly, & Tononi, 2016), but here I wish to use the term “constitutive mechanism” instead to emphasize that neuroscience does more than discover correlations. Another advantage of the term “constitutive mechanism” is that it is uncommitted with respect to the internalism-externalism debate: it leaves open the possibility that the mechanism of an experience is not just internal to the brain, but instead could be, say, an interaction process between the brain, body, and the environment (e.g., a process of active inference; Friston, 2010; Hohwy, 2013). However, accepting the notion of constitutive mechanisms is not necessary for this paper, and the reader is free to replace it with “neural correlate” if they wish; that does not affect the argumentation.

1 The Russellian typically accepts the conceivability argument and holds that there is a possible world where CME and E are dissociated: in a zombie world, CME occurs but E does not (Chalmers, 1996). I do not accept the possibility of such worlds, because I consider the term “CME” to be a rigid designator for E, but that is the topic of another paper.

2. Two problems with metaphysical intrinsics According to RM, even an ideal scientific CME-description cannot capture the subjectively felt character of an experience. That phenomenal feel, in the Russellian scheme, is the intrinsic nature of what the term “CME” refers to. On typical interpretations of RM, intrinsic properties are, roughly, non- relational and non-dispositional, whereas extrinsic properties are relational and dispositional (Alter & Nagasawa, 2012; Seager, 2006). Science can only describe relational and dispositional properties, but not non-relational and non-dispositional properties. For instance, physics can describe mass, force, and acceleration in the equation F=ma, but this only tells us how these parameters are related to each other: how an object with mass m is disposed to behave when it is subjected to a force F. RM is typically coupled with the claim that relations and dispositions of an object A must be grounded in what A is like in itself, intrinsically or categorically. The categorical nature of A grounds how it is disposed to act, and what types of relations it can stand in (Goff, 2017; Hiddleston, 2019). RM thus implies that there is a gap in the scientific description of the world, and here phenomenal properties can be fitted in. The metaphysical notion of intrinsic properties has been criticized in several ways. For instance, Hiddleston (2019) argues that there is no sharp boundary between categorical and dispositional properties, as is required by RM—instead, most scientifically interesting properties lie in the middle ground. Similarly, Howell (2015) presents problems related to mental causation in RM: if categorical (phenomenal) properties are metaphysically distinct from causal powers or dispositions, then it appears that phenomenal properties only contingently possess their causal powers. In the same vein, Grasso (2019) argues that RM’s categoricalism is prone to fickle qualia -scenarios, due to its distinction of categorical (phenomenal) properties from causal powers. Kind (2015), in turn, argues that RM fails to break free from the dualism/physicalism debate, mainly because it postulates that phenomenal properties are ontologically distinct from scientifically observable extrinsic properties (see also Chalmers, 2019; Jylkkä & Railo 2019). I will not go into these problems here, but it is worth noting that they all pertain to the metaphysically strong notion of “intrinsic”. I wish to focus next on two less discussed problems, namely: 1) The problem of specifying what is the intrinsic nature of a neural process that is the CM of an experience, and 2) The question of whether phenomenal properties can be observed.

2.1. What is the intrinsic nature of a neural process? The Russellian holds that the phenomenal feel of an experience E is constituted by the intrinsic nature of its neural mechanism CME. For example, Strawson writes:

“I am happy to say, along with many other physicalists, that experience is ‘really just neurons firing’ […] But when I say these words I mean something completely different from what many physicalists have apparently meant by them. I certainly don’t mean that all characteristics of what is going on, in the case of experience, can be described by physics and neurophysiology or any nonrevolutionary extensions of them. That idea is crazy. It amounts to radical ‘eliminativism’ with respect to experience, and it is not a form of real physicalism at all.” (Strawson, 2006: 7; emphasis added)

Strawson can be interpreted as saying that phenomenality is part of the intrinsic nature of specific neuronal processes, whose extrinsic nature science describes. But how could a neuronal process have an intrinsic, i.e., non-relational nature, given that processes are relational entities? Most neuroscientific accounts of the CMEs hold that experiences are interaction processes. For instance, the Global Neuronal Workspace theory (Dehaene & Changeux, 2011) says that a red experience is interaction between the visual sensory cortex and the widely distributed thalamo-cortical global neuronal workspace. In this process the sensory information becomes globally accessible and is integrated into an undivided experience. Neither the visual cortex process nor the global workspace process alone is conscious, but instead their interaction is. To take another example, Lamme’s (2018) local recurrent processing theory holds that a red experience is recurrent activation between two neuronal coalitions. The question for the Russellian is now this: How could an interaction process have an intrinsic, non-relational nature? To illustrate, let’s focus on a toy example of two objects A and B standing in a relationship R (e.g., causal interaction). It is possible to conceive of A and B having some nature independently of R, but could the causal interaction process R(A,B) itself have some intrinsic nature in the Russellian sense? Relevant here are the notions of relative vs. absolute intrinsic properties. Alter (2019) defines Relative Intrinsicality (RI) as follows:

(RI) P is a relatively intrinsic property of X just in case P is an intrinsic property of X and P is grounded in extrinsic properties of either X or parts of X.

A neural process can have relatively intrinsic properties, such as frequency, which the process can possess irrespective of its relations to other things. However, relative intrinsicality is not what RM takes to be crucial for consciousness, but instead Absolute Intrinsicality:

(AI) P is an absolutely intrinsic property of X just in case P is an intrinsic property of X and P is not grounded in extrinsic properties of either X or parts of X (Alter 2019).

How could a causal interaction process have an absolutely intrinsic nature that is not grounded in the extrinsic properties of its parts? By definition, the interaction process R(A,B) requires its constituents A and B to interact in a certain way. However, that A interacts with B is an extrinsic property of A, and that B interacts with A is an extrinsic property of B. Thus, (AI) implies that the absolutely intrinsic nature of R(A,B) cannot be constituted by such properties of A and B. By definition (AI) it appears logically impossible for a process to have an absolutely intrinsic nature. And if phenomenality is equated with absolutely intrinsic nature, it follows that a neural process cannot have phenomenal nature. Maybe the Russellian would admit that a process cannot literally have an absolutely intrinsic nature, and would rather say (loosely) that the phenomenality associated with a process (e.g., the phenomenal feel of red associated with the process CMred) is constituted by the intrinsic natures of the fundamental constituents of the process being combined in a certain way. For example, Goff (2016) suggests that two intrinsic natures (or microexperiences in Goff’s panpsychist version of RM) can be combined through a “phenomenal bonding” relation. Without going into the details, he admits the possibility of “there being some state of affairs of certain subjects of experience being related in some specific way which necessitates the existence of some distinct subject of experience” (emphasis added). For the present purposes, we may take Goff to mean by “distinct subject of experience” a macroscopic intrinsic nature. The problem with strategies like this is that whenever the Russellian explains the intrinsic nature of a macroscopic object x through relying on a relation that holds between its parts, s/he is no longer speaking of the absolutely intrinsic nature of x. For example, suppose that the intrinsic (or phenomenal) nature of a macroscopic object O is constituted by how its constituents A and B are related to each other. This implies that the intrinsic nature of O is constituted by its constituent A having the extrinsic property of being related to B in a certain way and by B having the extrinsic property of being related to A in a certain way. However, by definition (AI), such relations cannot constitute the absolutely intrinsic nature of O. In short, the notion of absolutely intrinsic nature appears to preclude that a macroscopic intrinsic nature could in any way be constituted by how its constituents are related to each other. But how else could intrinsics combine than being related to each other? To explain how a relational brain process could have an intrinsic, phenomenal essence, the Russellian arguably must solve the combination problem first. If phenomenality is the intrinsic nature of the CME, it must be constituted by the intrinsic natures of the ultimate constituents of CME, whatever they are. It would be attractive to consult empirical science how this happens, but the Russellian can’t do that, since they hold that science merely models extrinsic properties, not phenomenality which is intrinsic. For the Russellian, intrinsics are beyond the scope of science and science cannot explain how they are combined. This leaves the job to philosophers alone, and at least thus far, they have failed (Chalmers, 2017). Why not simply hold that phenomenal properties are those that science models as CMEs? Strawson (2006: 27) flirts with this option, claiming that “we can take it that human or sea snail experientiality emerges from experientiality that is not of the human or sea snail type, just as the shape-size-mass-charge- etc. phenomenon of liquidity emerges from shape-size-mass-charge-etc. phenomena that do not involve liquidity.” That is, there might be a parallelism between intrinsic phenomenal properties and the extrinsic properties that science models. But how should this be explicated, given that on standard RM the intrinsic properties are distinct from the extrinsic properties that science models? This is a metaphysical question that I will not go into here. Instead, my goal is to show how RM can be formulated purely epistemically without going into metaphysics. I don’t assume the above to be a knock-down argument against traditional RM. Rather, it is intended to illustrate the problems that dualism between intrinsic and extrinsic properties leads to. The main problem of RM is that it is a metaphysically weighty solution to an epistemic problem. If we can

5 explain where the epistemic gap stems from without going into metaphysics, we should do so already for parsimony’s sake.

2.2. Phenomenal properties can be observed It is common to hold that it’s impossible to observe phenomenal properties; it’s intuitively plausible that I cannot observe what your pain feels like. The standard Russellian story is that science cannot observe, describe, or model phenomenal properties because they are (absolutely) intrinsic, whereas science is merely concerned with extrinsic (or relatively intrinsic; see above) properties. However, from a purely naturalistic perspective, scientific observation is based on causation: If something has causal power, it can be observed and modeled by science. So, do intrinsic or phenomenal properties have causal power? The typical Russellian account is that intrinsics are the categorical bases of causal dispositions. This quickly leads to metaphysical problems. For example, is the relationship between categoricals and dispositions necessary or contingent? If it were necessary, then intrinsics could not do their explanatory work—e.g., zombies would not be possible, since dispositional properties of pain would be necessarily tied to phenomenal pain. Again, if the relationship were contingent, then it appears that phenomenal properties would have their causal powers not due to the way they are in themselves, but due to a lucky coincidence, namely how they happen to be related to dispositions in the actual world (for a more detailed argument of this kind, see Howell, 2015). In short, if phenomenal properties are dissociated from causal dispositions, they threaten to become epiphenomenal. These are complex metaphysical questions. I will not go into them here, but rather present a simple dilemma. Suppose that either phenomenal properties have their causal powers due to how they are in themselves (they are identical with dispositions, or their connection with dispositions is necessary), or their causal powers are not due to how they are in themselves (they do no causal work by themselves, but instead only through being somehow related to dispositional properties). The dilemma now is this: If phenomenal properties are in themselves causally efficacious, then they can also affect measuring devices and be observed. Alternatively, if phenomenal properties are not in themselves causally efficacious, then they cannot be observed and scientifically described, but neither is there mental causation: The phenomenal feel of pain can’t move the pointers of measuring devices but neither can it move your finger away from the fire. It’s impossible to both integrate phenomenal properties in causality and consider them as private—if they are causally efficacious, they can be observed through their effects. If the Russellian wants to keep phenomenal properties private and outside the scope of science, they also need to deny them of causal role. If the Russellian says that phenomenal properties cause things—the painfulness of pain makes me take a pain killer—then it also follows that phenomenal properties can be observed: they can causally affect measuring devices. The Russellian can’t both integrate phenomenal properties in causality and claim that they can’t be observed and described by science, since scientific experimentation and observation are a subset of causal processes. The Russellian might reply that there is indeed a sense in which experiences are observable and can be described by science. What is, then, left out from the scientific picture? The typical answer is that science merely models the extrinsic properties of experiences, but not the intrinsic properties, which include phenomenal properties—what experiences subjectively feel like. However, if that subjective feel has causal power in the first place, then it must also be able to affect scientific measuring devices and be observed. Again, if the phenomenal property has no causal power to affect measuring devices, then why would it have causal power to do anything else? If the Russellian insists that there is still something about the phenomenal property that cannot be observed and described merely through causal interaction, then the burden of proof is on them to show what that aspect is. If something is inaccessible to third-person science and inexplicable in physical terms, as intrinsics are supposed to be, then how could they play any causal role? One simply cannot hold that phenomenal properties are completely private and simultaneously say that they cause things: anything completely private will not show its effects to the outside. Strawson (2008) resolves the dilemma about how intrinsic or categorical properties are related to causal dispositions by holding that the two are “in reality” identical. He argues that “there is no real distinction between an object’s categorical properties and its dispositional properties or power properties” (p. 275). On his view, we can conceptually distinguish between an object and its properties, but such a distinction does not hold “outside of thought” (ibid., p. 271). He commits to the following two theses (S for Strawson):

(Sa) C[CategoricalO ≠ DispositionalO] (Sb) R[CategoricalO = DispositionalO]

(“C” stands for “conceptually”; “R” stands for “in reality”). (Sa) says that the categorical and dispositional properties of an object O can be conceptually distinguished—e.g., Strawson might say that we can logically conceive of a certain categorical property grounding some other disposition than in the actual world. However, (Sb) claims that in reality the categorical and dispositional properties are identical. We could take him to mean that at least in the actual world categorical and dispositional properties are identical. Now, assuming that we can (in principle) observe anything that has causal power, we can also observe categorical (phenomenal) properties, which for Strawson are in reality identical with causal dispositions. In more detail (S for Strawson; R for “in reality”):

(S1) The phenomenal quality of E = the categorical nature of E (Russellian premise) (S2) E = CME (identity assumption) (S3) The phenomenal quality of E = the categorical nature of CME (from S1 & S2) (S4) The categorical nature of CME =R the causal dispositions of CME (Sb) (S5) The phenomenal quality of CME =R the causal dispositions of CME (S1 & S4) (S6) Science can model the causal dispositions of CME (this is what science does) Therefore, (S7) Science can model (in reality) the phenomenal quality of CME (S5 & S6)

It thus appears that, on Strawson’s account, science can in fact model the qualitative character of an experience E through describing it as “CME”. This, however, is a conclusion that Strawson cannot accept, given that he endorses the Russellian doctrine that experiences are inscrutable. He rejects “physicSalism”, the claim that “the nature or essence of all concrete reality can in principle be fully captured in the terms of physics” (Strawson, 2019). What Strawson has in mind here is the phenomenal character of experiences. If the term “capture” here is equated with “describe”, “characterize”, or “model”, then the conclusion (S7) above amounts to accepting physicSalism about phenomenal properties: science can in fact fully describe all aspects of experiences. I have argued that the Russellian metaphysically strong notion of intrinsics leads to problems. More generally, it is a metaphysically heavy solution to the epistemic gap and resembles property dualism. Next, I will argue that we can preserve the core Kantian thesis of RM that science is somehow limited without postulating metaphysically distinct classes of properties. My intention is not to present a full-blown theory, but instead to show that RM without intrinsics is possible.

3. Two readings of “intrinsic nature” In this section I put forth an epistemic reading of “intrinsic nature” as an alternative to the traditional, metaphysical notion of “intrinsic” that RM relies on. On the typical reading, the absolutely intrinsic nature of x is taken to be the nature of x itself, independently of all its relational properties or the relational properties of its constituents. However, Strawson also talks about how science relies on observations or appearances: “The neo-Kantian points out that all we ever directly know of [the physical] is an appearance, both in everyday life and also and equally when we do physics, and that we have no good reason to think that the appearance reveals the nature of the physical as it is in itself, except (no doubt) in certain general structural respects.” (Strawson, 2019, p. 16, emphasis added). Similarly, Eddington talks of pointer readings, or that which appears to the researcher: “the atom is, like everything else in physics, a schedule of pointer readings [on measuring devices]. The schedule is, we agree, attached to some unknown background” (Eddington, 1929, p. 259). Eddington is not making any metaphysical claims here, but rather points to our epistemic situation: we must rely on observations in modeling the world. The core Kantian claim of RM is that we are in some sense ignorant of the nature of the external world. Traditional RM holds that we can’t (empirically) know about intrinsic properties—that is, there exists a class of properties we are ignorant of. This resembles metaphysical interpretations of Kant, such as that of Rae Langton’s (2001), with the exception that Langton’s intrinsic natures are completely causally inert and detached from extrinsic properties. In contrast, there are epistemic interpretations of Kant, such as that of Henry Allison’s (2004), which holds that the in-itself character of an object is simply its nature considered in abstraction from our modes of knowing it. Importantly, this does not imply that there would one class of properties that we can know and another class that we can’t know—it’s the same set of

7 properties that we know through representing them, but whose representation-independent nature we can’t possibly know. Lucy Allais (2015) argues that “the way the object is presented in consciousness is something more than the object alone, as it is outside of this relation” (p. 113; emphasis added). In the case of observing experiences, the “object” is an experience of one person, which is presented in the consciousness of another person. The way the observed experience appears to the observer is partly determined by the observer, such as their perceptual apparatus (determined by the biological layout of our species), technical equipment (e.g., fMRI vs. EEG), and conceptual repertoire (e.g., whether the observations are understood in terms of the integrated information theory or the global workspace theory). Allais puts forth the notion of “essentially manifest qualities”, i.e., “qualities which objects have as they are presented to subjects but which they do not have as they are apart from their appearing, as they are in themselves” (Allais, 2015: 117). It could be argued that the CME-quality of an experience E is essentially manifest in this sense: the experience has it only in relation to the observer. I will not, however, go deeper into Kant exegetics here—it’s beyond the scope of this paper. Rather, my intention is to show what kind of epistemic limitedness I am after. It is one thing to say that science is limited because it relies on observations and models and another that we can only observe and model the extrinsic properties of matter. There are two readings what “intrinsic” could mean. The classical Russellian reading refers to an object’s nature independently of all relations; we may call this Strong Intrinsicality:

(SI) Strongly intrinsic nature of x =def. the nature of x independently of how it is related to anything else

Accordingly, Strong Extrinsicality can be defined as follows:

(SE) Strongly extrinsic nature of x =def. how x is related to anything else

Weak Intrinsicality, in turn, can be defined thusly:

(WI) Weakly intrinsic nature of x =def. the nature of x independently of how S models x2

And Weak Extrinsicality as follows:

(WE) Weakly extrinsic nature of x =def. how S models x

Here “modeling x by S” means any act of representing x by S: observing, describing, or simply referring to x under some mode of representation.3 The weakly extrinsic nature of an object is the way it appears to me in observation or representation, in contrast to how it is in itself. The weak and strong notions of intrinsicality are superficially similar and do similar theoretical work (as I will argue), but on the metaphysical level they are in sharp contrast. (SI) implies that an object has some completely non- relational or categorical nature, whereas (WI) makes a superficially trivial point: objects (or processes) exist and have some nature independently of how they are related to observers and their models (observations, descriptions, characterizations—any mental representation of the object).

2 It is noteworthy that (WI) is similar to IIT’s notion of intrinsicality, where the claim about the observer- independence of experiences is explicitly made (Tononi, Boly, Massimini, & Koch, 2016). Similarly, predictive coding literature speaks of “hidden causes” in the world, which are inaccessible to an organism modeling its surroundings. The organism only has access to the effects of the external environment, based on which it creates a generative model of the world (Hohwy, 2013). The same applies to scientists constructing models based on the effects of hidden causes. 3 Proponents of direct reference could argue for the possibility of referring to x directly (e.g., ostensively), but this radical approach falls prey to qua-problems (Devitt & Sterelny, 1999). For example, pointing to a sample of water and saying “that is water” fails to uniquely fix the reference of “water” to H2O, because many natural kinds are instantiated in the sample (hydrogen, oxygen, bacteria, chlorine, etc.). For this reason, fixing the extension of a term always requires a mode of representation, sense, or intension that is internal to the speaker (Jylkkä, 2008).

How should (WI) and (WE) be understood? To begin with, it makes no sense to speak of the observer- or modeler-independent nature of x without specifying who we mean by the “observer” or “modeler”. If by those terms I mean myself, I cannot speak of WI-nature at all: the moment I create a concept for the modeler-independent nature of x, I immediately create a relation between x and my model of x as “x” or “model-independent”. The difference between WI and WE can only be understood from a “metamodeler’s” perspective, such that the metamodeler is distinct from the modeler. Consider metamodeler S2 who can observe the relationship between a modeler S1 and the modeled object x. From S2’s perspective, when S1 models x, S1 and x stand in the following relation (M for modeling):

(M) R(MS1,x)

Clearly, subject S1 cannot know the model-independent nature of x through modeling x. However, the metamodeler S2 can model the relationship between S1 and x (as in [M]) and can judge that x exists independently of being modeled by S1. Thus, from S2’s perspective, S1 can model the modeler-S1- independent nature of x. Similarly, and in more detail, Allais (2015) argues that we can perceive things themselves instead of our representations of them. Note that S2 cannot speak of modeler-S2-independent nature; to do that, we should postulate yet another observer S3, who models the metamodeling process as R(MS2,(R(MS1,x))). This reflects the regress that follows when trying to refer to model-independent x simpliciter.4 In short, it is logically impossible to know the model-independent nature of x through the act of modeling it. In practice, this means that I cannot know what the world is like independent of my consciousness, because any knowing takes place in my consciousness. The world outside my consciousness simply doesn’t exist for me, just like colors don’t exist for a person with complete congenital achromatopsia, or like the other side of the world doesn’t exist for a person with hemispatial neglect. Just like the person with achromatopsia cannot imagine colors, I cannot imagine the me- independent nature of the world. Again, when I do represent the world, I have no access to world independent of my representations of it, since my representing is constituted by my brain or mind together with the world—as Allais (2015) notes, representation is always more than the represented alone. This can account for why science is limited in its access to the nature of experiences: Science is limited to modeling reality based on observation, whereas experiences are part of the concrete reality beyond models, producing observations. Whereas traditional RM finds home for phenomenality in intrinsic properties, I suggest the home of phenomenality is among things in themselves, in concrete reality beyond scientific models. Whereas traditional RM considers the epistemic gap to reflect the duality between intrinsic and extrinsic properties, I take it to reflect the distinctness between concrete objects and our abstract models of them. This is best illustrated by focusing on how experiences are modeled.

3.1. Modeling an experience Let us apply the notions (WI) and (WE) to neuroscientific modeling of experience. When a neuroscientist Patricia models as M (e.g., CMred) an experience E (red experience) in a research participant Paul, this happens in the following relation:

(ME) R(MPatricia,EPaul)

Weak Intrinsicality (WI) implies that Patricia cannot know the model-independent nature of Paul’s experience, or the way Paul’s experience is outside Patricia’s consciousness. Patricia can represent Paul’s experience in many ways; through the model M, through lay psychological inference from Paul’s behavior, through reports, and so on, but Patricia cannot know the nature of Paul’s experience independent of and beyond her models of it5.

4 This is similar to Tarski's (1944) demonstration that we cannot define the truth conditions of a sentence in a language without postulating a metalanguage where the truth conditions are defined. E.g., in our metalanguage now, we can say that the sentence “snow is white” in the object language is true if and only if snow is white. Likewise, we cannot say what our terms refer to without using non-quoted terms, as in: the word “snow” refers to snow. 5 I leave open the possibility that empathy could afford knowing another person’s experience non-representationally. For example, the external signs of pain in S1 would trigger pain in S2, enabling S2 to “co-live” S1’s pain instead of merely representing it. It could be argued that empathy is necessary and sufficient for seeing other creatures as minded (see Stueber, 2019). From a purely objective perspective everyone else is a zombie.

The critical question now is: Can Paul himself know the weakly intrinsic or model-independent nature of his experiences? In a trivial sense this is impossible, if Paul’s experience itself is considered as a model—say, his red experience might be taken to represent red objects. In this sense, Paul cannot know his red experience independent of models since the red experience is itself a model, and the experience cannot be known independent of itself. This is hardly interesting. More important is whether Paul can know his experience independent of any other models than the experience itself. It strongly appears that the answer is yes: Paul can know his experience through experiencing it, without modeling it in any way. To know an experience, it suffices to undergo it. This is what Russell meant by saying that there is a “sense of ‘knowing’ in which, when you have an experience, there is no difference between the experience and knowing that you have it” (Russell, 1959: 105). We are all familiar with flow experiences where we are simply present in a moment without reflecting ourselves or our experiences in any way. This, I take it, is the strongest sense of knowing one’s experiences: being present. Any kind of modeling, even introspective, is always abstraction and separate from the experience itself. Brute experience and reflection on it are different processes also on the brain level: Metacognition requires global ignition and relies heavily on the prefrontal cortices (e.g., Chein & Schneider, 2012), whereas non-reflective, phenomenal experiencing can take place relatively locally on sensory cortices (Lamme, 2018). It is notoriously difficult to experimentally tease apart these two varieties of experiencing, brute first-order vs. reflective (Block, 1995, 2019). However, for the present purposes it suffices to acknowledge that brute non-reflective experiencing exists, and that we are aware of it, even if it is difficult to empirically capture the corresponding neural processes. It is also important to notice that we cannot help but be immediately aware of the current contents of our consciousness, be they first-order or reflective. Even during intensive metacognitive reflection, one is immediately aware of the ongoing act or process of reflection. I take this to be self-evident. If someone disagrees and holds that reflection on an experience is necessary to know it, it is difficult to convince them otherwise. An argument could be made that if reflection was necessary to know an experience, then one could never know one’s present experience. Suppose that at moment t1 one experiences the taste of coffee, and at the following moment t2 starts to reflect on it and thus becomes aware of the taste. Now, is the subject aware of reflecting on the taste at t2? To be aware of the reflection, the subject needs to take the act of reflection as the object of a yet higher-order cognitive process at t3. However, at t3 the subject could not be said to be aware of their metareflection, for it would require metametareflection at t4, and so on, ad infinitum. There are surely possible responses to this argument, but I will not go into it deeper. I take it as self-evident that we are aware of our experiences without reflecting on them (see also Jylkkä, 2016). To sum, subject S1 knows the weakly intrinsic or model-independent nature of their experiences simply through undergoing them. In contrast, another subject S2 cannot know the nature of S1’s experiences without modeling or representing them somehow—through scientific models as in the case of Patricia, through lay psychological models, or any other mode of representing (but see fn. 5). Thus, S2 cannot know the weakly intrinsic nature of S1’s experiences, but instead can only know their weakly extrinsic nature.

3.2. Mary revisited It is clear now how the notions of WI and WE can be applied to the case of Mary. During her black-and-white period, Mary knew the WE-nature of red experiences, namely what types of observations they produce through whatever type of measuring devices Mary uses, and how they can be modeled as “CMred”. Given that these descriptions are based on how red experiences affect measuring devices and observers, Mary did not learn through them the WI-nature of red experiences: the nature of that something that is “moving the pointers” and is modeled. When Mary experiences red for the first time, she learns the weakly intrinsic nature of red experiences, i.e., their nature independently of how they affect measuring devices and are modeled by neuroscientists. In fact, she learns the nature of phenomenal redness independent of how it is modeled by anyone in any way. Mary’s knowing the weakly intrinsic nature of a red experience is constituted simply by its happening in her.6 Crucially, when Mary learns the

6 Mary can then say that red is this experience, referring to the happening of the experience. Some versions of the phenomenal concepts strategy claim that the epistemic gap is reduced to the difference between such indexical concepts and natural kind concepts which are not in this way indexical (e.g., Ismael, 1999; Loar, 1990; Papineau, 2002). However, on my view the gap is deeper than conceptual: it pertains to the difference between a concrete entity itself and abstract models of it. Whereas the phenomenal concepts strategy emphasizes the immediacy of

WI-nature of red, there is no novel property that she acquires knowledge of; her previous neuroscientific knowledge was already about phenomenal redness. The new knowledge, however, is not about anything, it’s simply the happening of a red experience (the process that she previously had merely modeled as CMred). It could be objected that I am merely passing the explanatory buck: I claim that the epistemic gap is the difference between the model and the modeled, but I don’t show why the two are epistemically discrepant. The short answer is that representation of x is always more than x alone, since the former is constituted by how x interacts with and is processed in the one who represents (Allais, 2015). I cannot provide a full-fledged answer here. My main goal is to show how Mary’s epistemic situation can be explained with the notions (WI) and (WE), without going into metaphysics. More generally, my aim is to shift the focus from strongly intrinsic vs. strongly extrinsic properties to the epistemic question of how we represent things vs. how things are beyond our representing them. When formulated in this way, the question is not how two types of properties—qualitative vs. structural-relational—are related to each other, but instead, how things are represented vs. how they are in themselves. If experiences can be non- redundantly observed, and if they appear in observation as CMEs, then we must conclude that somehow what is felt as qualitative by the experiencer is observed as a structural-relational process by the neuroscientist. Allais’ notion of essentially manifest quality could be applicable here: an experience E has the CME-nature only when it is suitably related to an observer with the relevant type of technical, conceptual, and cognitive machinery (Allais 2015: 101 ff.).

3.3. Russellian problems revisited Let us now briefly return to the two problems I presented against classical RM. The first is how to conceive of neural processes having strongly intrinsic natures. When microscopic intrinsics are combined into macroscopic ones, it is arguable that the microscopic intrinsics must somehow be related to each other. However, such relations are extrinsic, and cannot constitute phenomenal properties, which are intrinsic and non-relational. The notion of weak intrinsicality escapes this problem since a relational process can have an observer-independent nature without having a strongly intrinsic nature. Since the CME-model is literally and non-redundantly a model of the experience, explaining how the CME is constituted amounts to explaining how the experience is constituted. This option is not viable for the traditional Russellian, who considers strongly intrinsic properties to be beyond the scope of science; they are a different class of properties than the strongly extrinsic ones that science models. I have argued that a scientific CME-model of an experience E is literally and non-redundantly a model of the experience itself, including what it feels like—nothing is left out. I grant that this is counterintuitive. It appears that the model fails to capture phenomenality, and, in a sense, this is true, if by phenomenality we mean something that is model-independent simpliciter. However, it would be a mistake to expect science to do anything besides modeling. On the account I have made, science can in fact model all aspects of phenomenality, but is limited to modeling—scientists can never step beyond their observations to what causes them, or to conceive of the represented independent of their representations. Experiences can be modeled as well as anything can ever be modeled, and science can explain how experiences are constituted by explaining how CMEs are constituted. A sense of mystery prevails, however, since in the case of experience we are aware of the limits of scientific models, unlike in the case of any other natural processes. My own experience is the only process whose weakly intrinsic nature I can know; the weakly intrinsic nature of anything else is a black box for me (although empathy might be an exception; see fn. 5). From a purely subjective perspective experiences appear to come out of nowhere, although an external observer can show how they are formed through explaining how CMEs are formed. The second problem is that either phenomenal properties are observable, or they don’t have causal power. The classical Russellian cannot render phenomenal properties inscrutable and private and yet consider them as causally efficacious, because if a property has causal power, then it can be observed and modeled through its effects. The burden of proof is then on the Russellian to show what is left unexplained by science and why. On a version of RM that relies on (WI) and (WE), it could be said that inscrutability is relative: phenomenal properties can indeed be observed and modeled, but only observed and modeled—science has no access to the observer- and modeler-independent reality. Experience is the concrete process in the world, the thing in itself that causes observations of its CMEs, and which observers can merely represent in their own minds. This shifts the explanatory target: whereas traditional phenomenal concepts, I (additionally) emphasize the mediacy and limitedness of scientific knowledge, as RM does (but without postulating metaphysical intrinsics).

RM aims to explain the relationship between phenomenal (intrinsic) and non-phenomenal (extrinsic) properties, I challenge us to explain why qualitative, phenomenal things in themselves appear as non- phenomenal or structural-relational when they are observed.

3.4. Is Weak Intrinsicality too weak or too strong? Finally, I wish to address two possible objections to the notion of Weak Intrinsicality, namely that it is either too weak, failing to explain Mary’s epistemic situation, or too strong and incompatible with scientific realism. The first objector could say that it is trivial that a model is not the modeled. I agree that the distinction is trivial on the superficial level, but not at all trivial when we acknowledge its implications. The model-modeled difference is the reason for our epistemic situation: being organisms with boundaries, we cannot know an external object without modeling it in some way. The notion of Weak Intrinsicality implies that the model-independent nature of the modeled is categorically beyond the grasp of the modeler. It is logically impossible for a subject to know the nature of an external object independently of how it relates to her; we can only know the world through our “lens of consciousness”. This epistemic tragedy is not resolved by future science: even the perfect model is still just a model in the conscious- cognitive system of the modeler. In sum, the model-modeled distinction is all but trivial. Again, even if it were considered trivial, this would require extra argumentation to count as an objection. If the distinction does its theoretical work—e.g., explain Mary’s epistemic situation—then it should be counted in its favor if it’s trivial (instead of highly controversial). Another objection is that the model-modeled distinction is too strong and leads to phenomenalism or instrumentalism about science: it implies that there is no resemblance or similarity between a model and the modeled, so no scientific model can be considered as true. This is simply not the case. Scientific realism is often defined as the claim that scientific theories can be structurally isomorphic with what they are about (Psillos, 1999: 144; Worrall, 1989). Russellian monism is likewise typically coupled with structuralism about science (Alter, 2019), but it additionally commits itself to the existence of quiddities which ground dispositions and relations. Whether such properties are postulated or not is a question independent of scientific structuralism: from the thesis that science can only model structure it does not follow that something besides structure exists. General metaphysical arguments can be made in defense of quiddities, but I have argued that the Mary case doesn’t motivate postulating them: Mary’s epistemic situation can be explained without going into metaphysics. Thus, an epistemic version of RM that relies on (WI) instead of (SI) is more parsimonious and compatible with a wider range of theoretical positions than traditional Russellianism. Moreover, it respects the intuition that one cannot draw metaphysical conclusions from epistemic premises.

4. What is inscrutable? The Kantian core of RM, which I take to be independent of the categoricalism-dispositionalism debate, is that science is somehow limited. However, whereas traditional forms of RM hold that there are certain kind of properties (i.e., strongly intrinsic) that cannot be modeled by science, I have argued for an alternative account where science is not limited to modeling a certain class of properties, but instead to modeling simpliciter. When we scientifically model something, we cannot eliminate ourselves from the modeling process. It is not possible for us to step outside our point of view and see what the nature of the world is independently of us—when we think, we cannot abstract away ourselves as thinkers. However, at least in the case of experiences and their scientific models, we can agree with Eddington that the models and pointer readings are “attached to a background of consciousness” (Eddington, 1929, p. 259). This, however, does not imply that science could not describe and model consciousness, it implies exactly the opposite. The fact that consciousness can move pointers in measuring devices implies that it can be scientifically modeled—but just modeled.

References Allais, L. (2015). Manifest Reality: Kant’s Idealism and his Realism. New York: Oxford UP. Allison, H. E. (2004). Kant’s Transcendental Idealism: An Interpretation and Defense. London: Yale UP. Alter, Torin. (2019). Russellian Monism. In Stanford Encyclopedia of Philosophy. Retrieved from https://plato.stanford.edu/entries/russellian-monism/

Alter, Torin, & Nagasawa. (2012). What is Russellian Monism. Journal of Consciousness Studies, 19(9–10). Block, N. (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences. https://doi.org/10.1017/S0140525X00038188 Block, N. (2019). What Is Wrong with the No-Report Paradigm and How to Fix It. Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2019.10.001 Chalmers, D. (1996). The Conscious Mind: in Search of a Fundamental Theory. Oxford: Oxford UP. Chalmers, D. (2017). The combination problem for panpsychism. In Panpsychism. Contemporary Perspectives. Oxford: Oxford UP. Chalmers, D. (2019). Idealism and the mind-body problem. In W. Seager (Ed.), The Routledge Handbook of Panpsychism (pp. 353–373). New York: Routledge. Chein, J. M., & Schneider, W. (2012). The Brain’s Learning and Control Architecture. Current Directions in Psychological Science, 21(2), 78–84. https://doi.org/10.1177/0963721411434977 Dehaene, S., & Changeux, J. P. (2011). Experimental and theoretical approaches to conscious processing. Neuron, 70(2), 200–227. https://doi.org/10.1016/j.neuron.2011.03.018 Devitt, M., & Sterelny, K. (1999). Language and reality. An introduction to the philosophy of language. Oxford: Blackwell. Eddington, A. (1929). The Nature of the Physical World. Cambridge: Cambridge UP. Feigl, H. (1958). The Mental and the Physical. In H. Feigl, M. Scriven, & G. Maxwell (Eds.), Concepts, Theories, and the Mind-Body Problem. Minneapolis: University of Minneapolis Press. Friston, K. (2010, February 13). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience. Nature Publishing Group. https://doi.org/10.1038/nrn2787 Goff, P. (2016). The phenomenal bonding solution to the combination problem. In G. Bruntrup & L. Jaskolla (Eds.), Panpsychism: Contemporary Perspectives. Oxford: Oxford UP. Goff, P. (2017). Consciousness and Fundamental Reality. Oxford: Oxford UP. Grasso, M. (2019). IIT vs. Russellian Monism: A metaphysical showdown on the content of experience. Journal of Consciousness Studies, 26(1–2), 48–75. Hawking, S., & Mlodinow, L. (2010). The Grand Design. New York: Bantam Books. Hiddleston, E. (2019). Dispositional and categorical properties, and Russellian Monism. Philosophical Studies, 176(1), 65–92. https://doi.org/10.1007/s11098-017-1006-2 Hohwy, J. (2013). The Predictive Mind. Oxford: Oxford UP. Howell, R. (2015). The Russellian Monist’s Problems with Mental Causation. The Philosophical Quarterly, 65(258), 22–39. https://doi.org/10.1093/pq/pqu058 Ismael, J. (1999). Science and the phenomenal. Philosophy of Science, 66(3), 351–369. Jackson, F. (1982). Epiphenomenal qualia. The Philosophical Quarterly, 32(127), 127–136. Jylkkä, J. (2008). Theories of natural kind term reference and empirical psychology. Philosophical Studies, 139(2), 153–169. https://doi.org/10.1007/s11098-007-9107-y Jylkkä, J. (2016). Mind as an intrinsic property of matter. Philosophical Investigations, 39(1), 15–37. https://doi.org/10.1111/phin.12100 Jylkkä, J., & Railo, H. (2019). Consciousness as a concrete physical phenomenon. Consciousness and Cognition, 74, 102779. https://doi.org/10.1016/J.CONCOG.2019.102779

Kind, A. (2015). Pessimism about Russellian monism. In Torin Alter & Y. Nagasawa (Eds.), Consciousness in the Physical World: Perspectives on Russellian Monism (pp. 401–421). Oxford: Oxford UP. Koch, C., Massimini, M., Boly, M., & Tononi, G. (2016). Neural correlates of consciousness: Progress and problems. Nature Reviews Neuroscience. https://doi.org/10.1038/nrn.2016.22 Lamme, V. A. F. (2018, September 19). Challenges for theories of consciousness: Seeing or knowing, the missing ingredient and how to deal with panpsychism. Philosophical Transactions of the Royal Society B: Biological Sciences. Royal Society Publishing. https://doi.org/10.1098/rstb.2017.0344 Langton, R. (2001). Kantian Humility: Our Ignorance of Things in Themselves. Oxford: Oxford UP. Loar, B. (1990). Phenomenal states. Philosophical Perspectives, 4, 81–108. Maxwell, N. (1965). Physics and Common Sense. British Journal for the Philosophy of Science, 16, 295–311. Meehl, P. E. (1966). The Compleat Autocerebroscopist: A Thought-Experiment on Professor Feigl’s Mind-Body Identity Thesis. In P. K. Feyerabend & G. Maxwell (Eds.), Mind, Matter, and Method: Essays in Philosophy and Science in Honor of Herbert Feigl. Minneapolis: University of Minneapolis Press. Montero, B. (2015). Russellian physicalism. In T. Alter & Y. Nagasawa (Eds.), Consciousness in the Physical World (pp. 209–223). Oxford: Oxford UP. Nagel, T. (1979). What it is like to be a bat? In Mortal Questions (pp. 165–180). Cambridge: Cambridge UP. Nemirow, L. (1980). Review of Thomas Nagel, Mortal Questions. Philosophical Review, 89, 473–477. Papineau, D. (2002). Thinking about Consciousness. Oxford: Oxford UP. Psillos, S. (1999). Scientific Realism: How science tracks truth. London: Routledge. Revonsuo, A. (2006). Inner Presence: Consciousness as a Biological Phenomenon. Massachusetts: MIT Press. Robinson, H. (1982). Matter and Sense: A Critique of Contemporary Materialism. Cambridge: Cambridge UP. Russell, B. (1927). The Analysis of Matter. London: Kegan Paul. Russell, B. (1959). My Philosophical Development. Guernsey: Guernsey Press. Seager, W. (2006). The “intrinsic nature” argument for panpsychism. Journal of Consciousness Studies, 13(10– 11), 129–145. Stoljar, D. (2005). Physicalism and phenomenal concepts. Mind and Language, 20(5), 469–494. https://doi.org/10.1111/j.0268-1064.2005.00296.x Strawson, G. (2003). Real materialism. In M. Antony, Louise & N. Hornstein (Eds.), Chomsky and His Critics (pp. 49–88). Malden, MA: Blackwell. Strawson, G. (2006). Realistic Monism: Why Physicalism Entails Panpsychism. Journal of Consciousness Studies, 13(10–11), 53–74. Strawson, G. (2008). The identity of the categorical and the dispositional. Analysis, 68(4). Strawson, G. (2019). The Mary-Go-Round. In The Knowledge Argument (pp. 118–140). Cambridge: Cambridge UP. Stueber, K. (2019). Empathy. In Stanford Encyclopedia of Philosophy. Retrieved from https://plato.stanford.edu/entries/empathy/ Tarski, A. (1944). The semantic conception of truth and the foundations of semantics. Philosophy and Phenomenological Research, 4(3), 341–376. Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: From consciousness to its physical substrate. Nature Reviews Neuroscience. Nature Publishing Group. https://doi.org/10.1038/nrn.2016.44

Worrall, J. (1989). Structural realism: The best of both worlds? Dialectica, 43, 99–124.