Defense of Integrative Pluralism in the Cognitive Sciences

Sushchin Mikhail Aleksandrovich ORCID: 0000-0002-8805-6716 PhD in Philosophy Senior Researcher; Federal State Budgetary Institution of Science Institute of Scientific Information on Social Sciences of the Russian Academy of Sciences 117418, Russia, Moscow, Nakhimovskiy, 51/21 sushchin@bk.ru Other publications by this author

Introduction

The confrontation between pluralistic and unificationist points of view seems to be one of the most important topics in discussions about the foundations of modern cognitive sciences. Back in 1973, Alan Newell, one of the founders of cognitive sciences, expressed dissatisfaction with the extreme fragmentation of cognitive research at that time in a well-known work ^[1]. Already at that time, many different cognitive phenomena were identified in cognitive psychology (such as perceptual illusions, perception of pieces on a chessboard, mental images, linear search on the screen, etc.). Individual studies were poorly related to each other, and there was no accumulation of knowledge from experiment to experiment. For a theoretical understanding of the results obtained, the so-called binary oppositions were involved: conscious and unconscious, sequential processing and parallel processing, nature and upbringing, etc. Newell wrote: "Let's assume that for the next thirty years we will continue in the same spirit as we are now. A hundred more phenomena will be discovered and investigated, give or take a few dozen. The resolution of forty more oppositions will be put forward and initiated. Will psychology mature then? Will it properly cover its subject – human behavior – which we all consider to be a characteristic of mature science?" ^{[1, p. 287-288]}. In an attempt to counter this fragmentation, Newell came up with the idea of creating a so-called unified theory of knowledge. This idea has inspired many to develop so-called unified cognitive architectures aimed at modeling the multitude of individual abilities studied by cognitive scientists under the auspices of a single theoretical framework.

Similar concerns about the fragmentation of cognitive research have been expressed recently. In the oft-cited work of Nunez et al. An institutional and bibliometric analysis of the development of cognitive sciences was carried out ^[2]. Based on it, the authors expressed the point of view that the project of a unified and coordinated interdisciplinary cognitive science launched by the cognitive revolution has not been realized. In their opinion, from the early days of the cognitive sciences, it was assumed that they would have a single subject, methods and integrated theories. Subsequently, early unificationist aspirations led only to the emergence of an eclectic group of practices without any common goal or integrated theory. According to them, institutionally, the development of cognitive disciplines has led to the opening of only a few specialized university faculties. At the same time, bibliometrically, the field of cognitive research turned out to be largely subordinated to cognitive psychology ^{[2, p. 782]}. (While cognitive disciplines usually also include linguistics, artificial intelligence, neuroscience, philosophy, and anthropology.)

Meanwhile, some authors express the point of view that unification in the cognitive sciences can follow a rather radical scenario of establishing the so-called grand unified theory. In particular, the leading proponents of such an influential field in cognitive sciences as predictive processing (predictive coding) have great unificationist ambitions of this kind. In a rather vague form from the point of view of the philosophy of science, statements about the ability of predictive processing to become a theory of grand unification were made by the main theorists of this field: Friston, Clark, and Hochvi. Thus, according to Clark, predictive models represent a guiding thread on the way to a unified theory of mind and action ^{[3, p. 180]}. This kind of excessive optimism could not but incur criticism from a number of authors. After reviewing studies of dopamine brain systems, Columbo and Wright, contrary to the aspirations of proponents of predictive processing, pointed out that a number of hypotheses are currently being used to explain their activity, none of which provides a complete explanation of the operation of these systems based on a single set of fundamental principles ^[4]. Following this, Milkovsky and Litvin came to a similar conclusion about the insolvency of the unificationist claims of the ideologists of predictive processing ^[5]. From the point of view of these authors, the problem lies in the insufficient degree of falsifiability of predictive models ^[6].

In addition, the pluralistic attitude has received great support in the philosophy of science in general over the past decades. Influential pluralistic concepts have been developed by such well-known authors as Longino ^[7], Mitchell ^[8], Chang ^[9] and others. The pluralism of theories and explanations has also been supported in the works of authors who study cognitive sciences directly: Dale ^[10], Dietrich ^[11], Edelman ^[12], de Jong ^[13], etc. These and other authors continued the pluralistic tradition in the philosophy of science laid down by Mill, Popper, Lakatos and Feyerabend. In a series of recent papers, I proposed the idea of theoretical complexes in the cognitive sciences ^{[14, 15]}, entailing pluralism in a normative sense ^[16].

At the same time, as already mentioned, there are supporters of the idea of integrating a very heterogeneous and diverse landscape of cognitive research. Despite the critical assessment of the integrative potential of predictive models, Polish philosopher Marcin Milkowski advocates the idea of unification of cognitive sciences in a general sense. In his recent work, Milkovsky conducted a critical analysis of the idea of integrative pluralism in the cognitive sciences ^[17]. In his opinion, integrative pluralism plays only a temporary role in the development of scientific disciplines, whereas in the long term, the value of unification is more fundamental and significant, which should provide a deeper understanding of a certain subject area.

Thus, in this paper I will try to answer the main objections to the pluralistic attitude in the philosophy of cognitive sciences, taking as a starting point my own idea of theoretical complexes. In addition, the possibility of combining theoretical pluralism with moderate integrative projects will be shown. This kind of integrative pluralism, in a fundamental sense, turns out to be incompatible with radical unification projects like the idea of establishing a single cognitive monopolist theory.

The idea of theoretical complexes in cognitive sciences: key points

So, in a number of recent publications, I have proposed the idea of theoretical complexes (or theory complexes) in order to give a proper description of major theoretical areas in the cognitive sciences: classical cognitivism, connectionism, embodied cognition, predictive coding. More specifically, this perspective is aimed primarily at providing answers to the following questions: "what exactly allows us to talk about a particular group of cognitive theories and models as a certain theoretical direction?", "what factors serve as the basis for identifying these directions?", "as separate theories within these do the directions relate to each other?", "what are the main functional properties of these macro-theories?". (The last two questions can be formulated as follows: "what is the structural organization and functions of cognitive macro-theories?")

To answer these questions, I turn to the analysis of the main components, which, as traditionally considered in the philosophy of science, make up individual scientific theories (or mini-theories, to use Laudan terminology ^[18]). Focusing on the analysis of individual theories and models in the cognitive sciences, one can see that a typical cognitive theory or model consists of various kinds of statements, mathematical formulas, diagrams, illustrations, as well as algorithms or computer code in one of the specialized programming languages (in modern times, most often in Python, Matlab, C, C++).

One special type of statement, which I call general hypotheses about the nature of cognitive processes, is of key importance for describing the nature of major theoretical trends in the cognitive sciences. Sometimes implied only implicitly, such statements (or propositions) can be expressed in the form: "cognition or any of its important components is X", where X can be replaced by "calculation based on mental representations", "parallel distributed information processing", "actions of a corporeal agent in a certain environment" and etc . As concrete and perhaps the most well-known examples of this kind of general hypotheses, we can take the classical hypothesis about the combinatorial syntax and semantics of the structure of mental representations within the framework of the theory of the language of thought ^[19] or key assumptions about the organization and functioning of connectionist models (that is, a number of simple computational elements, the activation state of an individual element, the exit function for of each element, the pattern of communication between the elements, the rules for the propagation of activity in the network of simple computing elements, the activation rules for each element, as well as the rules for training the network to change the pattern of communication between elements based on experience ^[20]).

General hypotheses about the nature of cognitive processes have several key features. Firstly, they set for researchers a schematic vision of the organization of cognitive systems and cognition, as well as using this cognitive ontology – strategic methods for the study of cognitive processes. (For example, if cognition is understood as a calculation based on mental representations, and the cognitive system as a physical symbolic system, then cognitive sciences should rely primarily on computer modeling, whereas data from neuroscience cannot have a significant impact on their development.) Secondly, due to their abstract nature, propositions of this type cannot be considered sufficient to build a private cognitive model, and the Popper criterion of falsifiability should not be applied to them. On the contrary, they are in a complex relationship of interdependence with the more specific components of cognitive theories and models that were listed earlier. (In a general sense, such abstract propositions help to build concrete hypotheses and models, and those, in turn, in the long term help to clarify, make more concrete abstract general hypotheses about the nature of knowledge.)

Having discussed the key concept of general hypotheses about the nature of cognition, we can now answer the question of what exactly serves as the basis for distinguishing a certain group of cognitive theories as a theoretical direction in the cognitive sciences. In general terms, the answer is as follows: when talking about a certain cognitive macro theory (say, "predictive coding offers or does not offer a unified explanation of perception, action and cognition"), researchers in fact hardly mean that this macro theory is an example of solving Kuhn puzzles ^[21], scientifically-a research program in the sense of Lakatos ^[22], a research tradition in the sense of Laudanus ^[18] or a conceptual research framework according to Von Eckardt ^[23]. Instead, it seems to be implicitly assumed that this macro-theory is simply a group of separate cognitive theories and models that share a number of general provisions about the nature of cognitive processes (for example, the already mentioned hypothesis of cognition as a calculation based on mental representations, etc.).

It is important to note that there should be no contradictions between the key abstract hypotheses, since otherwise they will not be able to act in a coordinated manner as a basis for combining a group of theories and models into a single macro theory. For this reason, it would be correct to consider the projects of moderate and radical embodied cognition as separate theoretical complexes in the cognitive sciences – their adherents clearly disagree with each other on the issue of the participation of mental representations in cognition (adherents of radical embodied cognition argue that cognition does not occur on the basis of mental representations).

In addition, general hypotheses about the nature of cognition are crucial for determining the structural organization and basic functional properties of cognitive macro-theories. When I talk about structural organization, I mean the way the individual theories within the complex relate to each other. There are two main types of structural organization of cognitive macro-theories. In the first case, all models and theories related to this complex explicitly or implicitly share the same set of general hypotheses about the nature of cognitive processes. In the second case, the key provisions for a number of cognitive theories and models correlate with each other according to the type of Wittgenstein family similarities, having only partial intersections with each other. In my other works, I have pointed out that there are grounds to attribute classical cognitivism, connectionism, and predictive processing to the first type of complexes. At the same time, as I have tried to show elsewhere, the core structure of the conceptual provisions of the complex of moderate embodied cognition should be interpreted through the prism of the idea of family similarities ^[14].

General hypotheses about the nature of cognition play a similar role in determining the basic functional properties of cognitive macro-theories. It is the functional emergent properties that distinguish cognitive macro-theories from just a group or bundle of loosely interconnected individual theories. Now we can concretize our previous statements about the key role of general hypotheses in defining a schematic vision of the subject area. Thus, for proponents of a particular cognitive macro theory, its key provisions, embodied in its particular theories and models, act as a model in the creation of new and/or modification of existing individual theories and models of cognitive phenomena. Thus, they suggest (rather than completely define) an approximate range of the most relevant problems for research (for example, high-level cognitive phenomena, perceptual phenomena, motor control, etc.), as well as key concepts (for example, "distributed representation", "action", etc.) and the methodological style of research (for example, pure computer modeling without regard to neuroscience data for classical cognitivism, conducting experiments in the real world, creating mobile behavioral autonomous robots for various branches of embodied cognition, etc. – depending on the key premises of this particular direction). (It is important to note that some assumptions and methods (neuroscientific, statistical, software) of a general nature can be equally used by supporters of different theoretical complexes. This is, for example, the idea of mental representations in a general abstract sense, which is relied upon by proponents of all key cognitive complexes of theories, except for the movement of radical embodied cognition.)

Bringing everything together, the main cognitive macro-theories can be preliminarily defined as, in essence, groups of separate theories and models of cognitive processes, typified on the basis of abstract assumptions about the nature of cognitive processes common to them. Functionally, cognitive macro-theories are designed to ensure the further proliferation of individual theories and models of cognitive phenomena.

The approach proposed above to understanding the basic functions of cognitive macro-theories quite clearly entails theoretical pluralism in a normative sense. However, several questions immediately arise here. First, what is the purpose of this process of proliferation of scientific theories and models? Secondly, what are the advantages of the pluralism of theories in comparison with the projects of unification of modern cognitive sciences? Thirdly, is it possible to find a way to combine a pluralistic attitude with the unifying aspirations of some cognitive scientists and philosophers? In the next part of the article, I will try to give answers to these questions.

Theoretical complexes and theoretical pluralism: the answer to the main challenges

Perhaps one of the first and most natural questions in connection with the above scheme for understanding the basic cognitive macro-theories is as follows: why exactly does the cognitive sciences require further proliferation of individual theories and models? What are the advantages of a normative statement about the need for further proliferation? It is easy to see that cognitive sciences do not lack a variety of theories and models of cognitive processes. Such large macro-theories as connectionism, embodied cognition, predictive coding have contributed to the creation of a considerable number of individual theories and models. For example, connectionist models of interactive activation ^[24], the study of the past tense of English verbs ^[25], the simple recurrent Elman network ^[26], and others have gained considerable fame. In this regard, the situation of the multiplicity of theories, which Newell wrote about, has become even more pronounced. Accordingly, someone could, on the same grounds as Newell did at the time, ask why we need even more separate cognitive theories and models? To put forward and initiate the resolution of an additional forty theoretical oppositions? Thus, developing the question, critics could say that cognitive sciences do not require proliferation, but theoretical unification: the creation of a theory that could cover as many cognitive phenomena as possible under a single umbrella. In a similar vein, doubts have been expressed in the literature that proliferation can eventually lead to the creation of many useless theories ^[27] or that it can lead to a dispersion of efforts and a lack of understanding ^[28].

To meet these challenges, we need to supplement the previous normative statement that cognitive macro-theories are designed to ensure the further proliferation of individual theories and models of cognitive processes. It can be supplemented as follows: the multiplicity of theories and their collisions, mutual criticism, in the long term should lead to the fact that new theories will become better and better in relation to the so-called characteristics or advantages of a good theory. It is precisely this potential fruitfulness of competitive pluralism that was emphasized in the classical works of Lakatos ^[22] and Feyerabend ^[29].

In the literature on the philosophy of science, there is an established tradition of discussing the characteristics of a good theory, which are also called epistemic (or cognitive) values. In the classical works of Dugem ^[30], Popper ^[31], Lakatos ^[22], Kuhn ^[32] and others ^[33], a number of similar epistemic values were highlighted. These include (1) internal consistency ^{[30, 31, 22]}, (2) empirical adequacy ^[33], (3) simplicity ^{[31, 32]}, (4) breadth of coverage of the subject area ^[32], (5) formulation of new predictions of empirical facts (which, however, are still no empirical confirmation was found) ^[22], (6) predictive success (i.e. formulation of new predictions of empirical facts with their subsequent experimental confirmation) ^{[22, 33]}, (7) the ability to give unforeseen explanations to known facts ^[34], (8) the ability to successfully pass a series of experimental tests (which Popper in He called his time the degree of reinforcement, degree of corroboration ^{[31, 33]}), (9) the ability to combine disparate subject areas, provided, of course, that they are fundamentally interconnected. According to the philosopher of science Douglas, internal consistency and empirical adequacy are the minimum requirements for the acceptability of a scientific theory ^[35]. At the same time, values from the fifth to the ninth are considered more difficult to achieve (one can say ideal or desirable), which is why they are sometimes associated with the ability of theory to approximate the truth ^[33].

Thus, only unrestrained, unrestrained pluralism, which has no other purpose than the proliferation of theories and models, can lead to a dispersion of efforts and a lack of understanding. It seems that adherence to a constructive kind of pluralism, focusing on the improvement of theories and models through proliferation and mutual criticism, should ultimately lead to progress and improvement of these theories in relation to the mentioned epistemic values. In addition, answering the other part of the objection, it is impossible to say in advance whether a theory that seems useless will remain so forever: any theory can become useful sooner or later. As Feyerabend noted, there is no such idea, however ancient and absurd, that would not be able to improve our knowledge ^{[29, p. 33]}.

The same addition to our normative requirement, suggesting that the multiplicity of theories and mutual criticism should ultimately lead to an improvement in the explanatory, predictive and other characteristics of cognitive theories and models, should help us draw a clear line between pluralism and relativism ^[36]. So, another common objection is that the requirement of multiple theories, pluralism, leads to relativism. Generally speaking, the problem of the relation of pluralism to relativism is very extensive and complex, multidimensional. Now it is possible to answer this challenge only in the most general terms. Taking as a basis the argumentation of the philosopher of science Chang, it can be noted that pluralism and relativism focus on different things ^[9]. Relativism can be defined as the doctrine that all alternative points of view should be treated as equal: there is no universal truth, everyone has their own truth. Whereas pluralism emphasizes that the multiplicity of theories is a normal and even favorable situation for science, since it has to deal with extremely complex phenomena and since the competition between alternatives can lead to their improvement.

According to relativists, all points of view in cognitive research are equally true: classical cognitivism, collectionism, predictive coding, etc. The thesis of relativism is logically self-contradictory: this thesis is a conjunction of an antecedent in the form of a denial of universal truths and a consequent in the form of a statement of universal truth, namely that everyone has their own truth (in sum, this gives A and non-A at the same time). Meanwhile, the basic pluralistic concepts in no way imply that all points of view are equally true. Moreover, competitive pluralism explicitly supports the idea that theories can be compared with each other, as a result of which they can be accepted or rejected.

However, the appeal to the idea of epistemic values can be used to formulate a deeper objection to integrative pluralism. Such an attempt was recently made by Milkovsky. Based on his analysis of the classification of epistemic values proposed by the aforementioned philosopher of science Douglas ^[35], Milkovsky expressed the thesis that the doctrine of integrative pluralism is unstable by nature ^[17].

First of all, Milkovsky distinguishes between theoretical unification and theoretical integration. Unification, which, as Milkovsky notes, was carried out within the framework of Newtonian mechanics, seeks to develop general and simple theories by explaining a variety of phenomena within a single theoretical perspective. Whereas integration, carried out, for example, in protein folding studies, tends to combine many theories, without necessarily reducing all explanations to a single theory. (In the same protein folding studies, theories from physics, chemistry, and biology were combined to achieve a better understanding of the functional structure of the protein.)

The essence of the dispute here boils down to which epistemic values should be given priority. Accordingly, Milkovsky's argument is constructed as follows. Starting from the Douglas classification, Milkovsky argues that priority should be given to the value of explanatory depth rather than systematicity. Explanatory depth is what unificationism is designed to promote, whereas systematicity should be ensured by integrative pluralism. As Milkovsky notes, "deep theories provide a more detailed understanding of phenomena, shedding light on both their common connections and their unique characteristics. This dual nature of deep theories – the ability to formulate both comprehensive principles and particular (causal) conclusions – is what distinguishes them from superficial, overly simplified unifications" ^[17]. Unlike a simple conjunction of unrelated statements (say, superficially combined Kepler's laws and Boyle's law), a genuine explanation, Milkovsky is convinced, gives a deep understanding of phenomena, embracing them from a single perspective (say, deducing Kepler's laws from Newton's laws). For this reason, in the long run, he believes, preference should be given to theories offering deeper explanations.

At the same time, integrative pluralism is based on the need to combine many theoretical perspectives, a kind of mosaic of theories (patchwork). Thus, if we accept the premise that the value of depth of understanding is a priority, then, Milkovsky believes, for integrative pluralism this leads to tension, a situation of instability. The way out, in his opinion, is to recognize the temporary role of integrative pluralism, considering it as a kind of ladder, the ascent of which should lead to the acquisition of a truly unified deep theory of cognitive processes. In support of his argument, Milkovsky refers to the mentioned research in the field of creating cognitive architectures. If at first, following Newell's article, researchers remained optimistic about the ability of cognitive architectures to combine cognitive science, then later integrative models of cognition came to the fore. This change in research attitudes was due not to philosophical considerations, but to the complexity of building cognitive architectures. Thus, new architectures should work at least as well as their predecessors, striving to cover a huge number of cognitive phenomena, despite the fact that the criteria for evaluating their work are mostly informal in nature. Despite all this, according to Milkovsky, integrative pluralism should be considered as a temporary position in the field of creating cognitive architectures. Ultimately, he believes, proponents of the idea of creating cognitive architectures strive to build a unified theory, which runs counter to the installation of integrative pluralism.

Milkovsky's argument faces certain problems. The main difficulty here lies in his attempt to give priority to cognitive values of explanatory depth and unification without reference to empirical research.

The problematic nature of this style of argumentation was noted in the well-known joint work of philosophers Kellert, Longino and Waters. They give the following description of the key theses of the doctrine of monism:

"1. The ultimate goal of science is to establish a unified, complete and comprehensive theory of the natural world ... based on a single set of fundamental principles;

2. The nature of the world is such that it can, at least in the possibility, be fully described or explained by such a theory" ^{[7, p. x]}.

In response, Kellert, Longino and Waters point out that the validity of the second thesis can only be confirmed through empirical research. In other words, the question of whether the second thesis of the doctrine of monism is correct is open, empirical. According to these authors, this leads to the recognition that the multiplicity of theories represents the normal state of scientific research.

Based on the same considerations, it should be noted that the question of the possibility of explaining cognition from the point of view of a unified theory is empirical in nature. Accordingly, the value of unification cannot be imposed on cognitive research only by referring to successful physical theories of the past (for example, by referring to Newtonian mechanics, as Milkovsky does). Despite the fact that there is currently strong evidence in favor of a pluralistic approach to explaining cognition and the brain, it is still too early to make a verdict on which strategy (pluralistic or unificationist) will be more promising for cognitive research. Only empirical studies of the processes of cognition can provide an answer to this question ^[16].

These considerations lead us to the second important argument in favor of a pluralistic attitude in the cognitive sciences. This argument is connected with a huge amount of evidence in favor of the extreme complexity of both cognitive processes and their material substrate, that is, the brain and the central nervous system. So, in addition to the multiplicity of cognitive processes themselves (and related distinctions and classifications, as noted by the same Newell), their substrate, the brain, is studied at many levels, starting from molecules, synapses, neurons and further up the hierarchy up to the level of the entire nervous system. In addition, the picture already outlined is further complicated by the fact that, as proponents of embodied cognition like to emphasize, the mind has a certain type of physicality for successful actions in its environment. In the case of human cognition, this environment also includes intellectual artifacts and interaction with other people, which was especially explored in the classical works of L.S. Vygotsky.

Taking into account all these widely known facts of the incredible complexity of cognition and the brain, many authors express doubt that it will ever be possible to explain cognition in all its manifestations on the basis of only one theory, one set of principles ^{[4, 10, 11]}. For example, Dale suggests considering the following thought experiment. Let's assume that a certain person comes up to us on the street and says the following: "I've discovered everything about the Mississippi River, I have a complete theory of the Mississippi." It is quite natural to react to this by telling a stranger that what we call the Mississippi River is an extremely complex object, it is studied by many disciplines (including geology, ecology, microbiology, etc.): what aspect of this natural phenomenon is his theory intended to explain? The very idea of the Mississippi theory might seem strange to many. In the same way, if someone came up to us and said, "I have discovered everything about cognition, I have a complete theory of cognition," it would be fair to ask in response: "cognition, like the Mississippi River, is an incredibly complex set of phenomena - which aspect of cognition is this is the theory intended to explain?" ^{[10, p. 172]}.

Finally, there is experimental evidence of the fundamental splitting of human cognition into two different areas: conscious and unconscious information processing. There are good reasons to believe that each of these areas has a number of unique characteristics and properties. In particular, as it was shown in experiments by the French neuroscientist Deane and his colleagues, sequential conscious information processing has an integral, inherent role of connecting and directing individual steps in chains of multistep reasoning ^[37] (for example, mathematical reasoning: formulating a task and each of the subtasks, delegating subtasks to unconscious subsystems for calculation on each individual step, storing intermediate results, saving the final result). (Note, however, in this regard, that the question of the possibility or impossibility of explaining the conscious and unconscious on the basis of a single set of principles should be solved only in the course of future empirical research.)

So, the arguments we have considered from the progress and complexity of cognitive phenomena (coupled with considerations of a historical and epistemic nature ^[16]) currently speak against the need to establish a unified monopolist theory in the cognitive sciences according to the type of paradigm according to Kuhn ^[21]. At the same time, they in no way exclude integrative initiatives that are more modest in their intentions, compatible with a pluralistic attitude, and do not require the establishment of a single monopolist theory. In this sense, the space of possibilities for integrative pluralism is extremely large. As an example, a recently proposed project for the integrative design of experiments can be cited ^[38]. The project was presented to overcome the difficulty noted by Newell of integrating experimental results obtained in the course of individual studies. Its creators assume that the accumulation and unification of observations will be carried out by creating a so-called experimental space, which includes both experimental studies carried out on this particular problem and possible future experiments. Testing theories by sampling points from a given experimental space with their subsequent updating is, according to the authors of the project, the key to integrating experimental observations. Once again, a wide variety of initiatives can be compatible with integrative pluralism, provided that they do not encroach on subordinating the entire field of research to a single theoretical perspective.

Conclusion

So, in this article we have considered some important aspects of the discussion of proponents of pluralism and unificationism in modern cognitive sciences. This discussion is important both in terms of discussing current methodological practices and in the light of considering the prospects of cognitive sciences. As an example of a pluralistic approach in the philosophy of cognitive sciences, I took my own idea of theoretical complexes. After briefly reviewing its key points, I moved on to answering important challenges from adherents of unificationism in cognitive research. Thus, the issues of the need for further proliferation of cognitive theories and models, its advantages, and the differences between pluralism and relativism were consistently considered. Later, special attention was paid to a very clever objection recently formulated by the philosopher of science Milkovsky. According to Milkovsky, tension and instability are associated with the doctrine of integrative pluralism, based on the recognition of the priority of the epistemic value of the depth of explanation. As I have tried to show, tension does not really arise if empirical research is left to decide whether a pluralism of theories or a single unified theory is required to explain cognition. The value of unification should not be recommended to cognitive sciences on the basis of theoretical considerations alone, for example, by referring to successful unified scientific theories of the past (Newtonian mechanics, Maxwellian electrodynamics). Even taking into account convincing evidence in favor of the fundamental splitting of human cognition into conscious and unconscious processing, the question of the uniformity or heterogeneity of cognitive theories should be solved only taking into account the nature of the cognitive processes themselves (their heterogeneity or undiscovered fundamental unity so far), revealed in empirical research.

The current state of affairs in cognitive research is characterized by the multiplicity and heterogeneity of theories and models of cognitive processes. To the multiplicity of computational models of cognition mentioned by Newell, many connectionist, embodied, predictive and other theories and models of cognitive processes were added. According to a small interesting study conducted by Gentner, the multiplicity of theories and models does not run counter to the initial attitudes of a number of founders of modern cognitive sciences ^[39]. The cognitive scientists interviewed in this study, Donald Norman and Alan Collins, admitted that they did not believe in the early days of cognitive science that the development of this interdisciplinary field of research should lead to the establishment of a unified theory. It can be concluded that in this regard, things are going well in modern cognitive sciences and that, quite likely, this state of affairs does not need to change in the future.