Möbius Strip Model for Augmenting Organizational Knowledge Creation Dynamics by Integrating Human and Artificial Knowledge: A New Driving Force for Business Sustainability

Abstract

The emergence of artificial knowledge created by the generative artificial intelligence applications challenges the theory developed by Ikujiro Nonaka and Hirotaka Takeuchi concerning the organizational knowledge creation dynamics by showing its limits. It is necessary to reimagine this theory within a hybrid framework that integrates both human knowledge and artificial knowledge, being aware of their specific features. Several researchers have already suggested how the SECI (socialization–externalization–combination–internalization) cycle developed by Ikujiro Nonaka and Hirotaka Takeuchi can be augmented by introducing artificial knowledge next to human knowledge in each stage of that cycle. However, tacit knowledge is embodied, and it cannot be processed directly by generative artificial intelligence. Therefore, their suggestions ignore the nature and specific features of tacit and explicit knowledge, leading to non-coherent models. The purpose of this paper is to propose a new model based on the Möbius strip metaphor that contains an open SECI cycle coupled with an open artificial knowledge cycle. Knowledge is flowing continuously along the strip, converging in time toward a strange attractor. The value of the new model is given by its novelty of introducing an artificial knowledge cycle and augmenting with it the SECI model centred on human knowledge. The resulting model is more complex and allows a continuous flow of knowledge. Therefore, the organizational knowledge creation dynamics is not represented by a time-evolving spiral, but by the phase space of a strange attractor. The proposed model can be conceived as a new driving force of business sustainability.

1. Introduction

The organizational knowledge creation theory is based mostly on the work initiated by Nonaka [1] and developed further by Nonaka and Takeuchi [2,3] and Nonaka, Toyama, and Hirata [4]. This theory introduces the dyad of tacit knowledge–explicit knowledge based on the iceberg metaphor [2] and the tacit dimension defined by Polanyi [5]. Tacit knowledge is personal knowledge that is created through direct experience [6] and expressed through body language. It is subjective and represents the emotional states of the human body generated by external stimuli. Also, it includes intuitions, insights, ideals, and values [2]. In the iceberg metaphor, tacit knowledge is represented by the hidden part of the iceberg that is beneath the water’s surface. Explicit knowledge represents the visible part of the iceberg. It is expressed using natural or symbolic languages. It is codified knowledge that allows communication between people. It is processed by the conscious zone of the human brain [2,3,7].

Understanding tacit knowledge is the cornerstone of understanding the complexity of the knowledge construct and the domain of knowledge management theory and practice [7,8,9]. For many researchers, tacit knowledge is seen as a potential that can be converted into explicit knowledge using a natural or symbolic language. In this cognitivist framework, tacit knowledge is seen as a weak form of explicit knowledge [10,11,12,13]. From a phenomenological perspective, tacit knowledge is like a root for explicit knowledge. It is a sine qua non condition for the existence of explicit knowledge. Tacit knowledge is created through action (e.g., swimming, biking, or operating a machine) and is processed mostly unconsciously. It resonates with our emotional states. “Tacit knowledge cannot be ‘captured’, or ‘converted’, but displayed—manifested—in what we do. New knowledge comes about not when the tacit is converted to explicit, but when tacit knowledge is re-punctuated (articulated) through dialogical interaction” [10] (p. 473).

According to Nonaka and Takeuchi [2], knowledge is an integration of tacit and explicit knowledge through learning and social interaction. From the epistemological perspective, human knowledge is a justified true belief [14,15,16]. Therefore, a certain belief an individual may have should be justified as being true. In philosophy, justification is performed conceptually using different logical methods. In organizational practice, justification does not focus on reflecting the truth anymore but on managerial “processes of determining if the newly created concepts are truly worthwhile for the organization and society” [2] (p. 86). Thus, the truth is replaced by the market’s needs, which decision-makers must know.

Going beyond the iceberg metaphor and the dyadic structure of knowledge, Bratianu introduced the energy metaphor and the theory of knowledge fields [17]. The theory was further developed by Bratianu and Bejinaru [18,19]. The theory of knowledge fields contains three fundamental ideas: (a) knowledge is a field that is non-substantial, non-uniform, and nonlinear; (b) there are three basic fields: rational knowledge, emotional knowledge, and spiritual knowledge; (c) knowledge from one field can transform into any other field (e.g., rational knowledge can transform into emotional knowledge or spiritual knowledge and vice versa). Rational knowledge is the result of rational learning, and it is expressed using a natural or symbolic language [16,20]. It is objective and justified as a true belief [2,14]. For many researchers, the concept of knowledge reduces to rational knowledge, which is an oversimplification.

Emotional knowledge is subjective and reflects the emotional states individuals may have as a result of their perceptions [21,22,23]. Emotional knowledge is tacit, and it cannot be expressed in words or symbols. It is expressed using non-verbal languages, and it can be shared only in conditions of proximity. Due to its capacity to integrate experience, emotional knowledge is valuable in knowledge sharing and decision-making, as demonstrated by Nonaka and Takeuchi [2,3] and Konno [9]. While rational knowledge is linear, as a result of using language, which is linear, emotional knowledge is nonlinear because emotions and feelings are nonlinear and characterized by intensity [22,24]. Spiritual knowledge contains values and principles that guide our decisions and behaviour [21,25,26,27]. Phronesis or practical wisdom guides managerial decision-making, especially when there is a high level of uncertainty and a long-term framework [28,29]. Wise companies are created by wise leaders who change efficiency with wisdom and develop a new paradigm for innovation and sustainable business [3].

In recent years, human knowledge has been challenged by the emergence of artificial knowledge as a product of generative artificial intelligence (GenAI) [30,31,32,33]. In his seminal book on the sciences of the artificial, Herbert explains how the natural world is mirrored by an artificial world that contains artefacts as products of the human mind [34]. Artificial knowledge is an emergent artefact of this artificial world, although artificial intelligence has been known for a long time. The paradox can be understood if one makes the difference between data, information, and knowledge, as well as the difference between the concept of Shannonian information used in computer science and the concept of information used in the knowledge management domain [25]. Human knowledge is a result of learning and unlearning processes [35], while artificial knowledge is a result of the machine learning processes [6,31,33].

Artificial knowledge is a direct product of GenAI that has exploded in recent years due to its capacity to generate texts, images, audio content, videos, and even new software upon request from users [31,36,37]. When a user asks ChatGPT or a similar chatbot a question, the programme generates text in natural language, creating the illusion of a human dialogue. However, the human-like artificial knowledge is not a justified true belief. Artificial knowledge is not about any truth, and it does not originate from a semantic framework. It is constructed by an algorithm based on syntactic rules, not semantic interpretations [30,33,38]. Computers do not think, and GenAI does not use the meanings associated with words but rather their probability distributions within a given set of input data, information, and human knowledge. The algorithm is searching for the word with the highest probability of following in a text sequence based on the pattern learnt during the training session. Therefore, artificial knowledge content is path-dependent, and it is created in accordance with the probability distribution specific to a given set of input data. It has no direct relationship with external reality and no convergence towards a certain truth.

Artificial knowledge is exclusively rational. Machines cannot have emotional or spiritual states leading to cognition. Although there are simulations of emotional intelligence that focus on words that describe emotions and feelings, the texts generated by algorithmic patterns are rational [31,33,38,39]. Users of ChatGPT or similar GenAI applications should be aware of the possibility of dealing with the program’s “hallucinations”. They are distortions of reality caused by biases in the training dataset or in the algorithms’ design. For instance, if ChatGPT is asked to provide some significant references for a certain topic, the answer may contain references that do not exist, references created by the algorithm that mimic the existing ones. Users should check each answer they get, so that the artificial knowledge remains free of errors generated by those hallucinations. Also, users should be aware that the algorithms are sensitive to the formulation of the input, and that sensitivity may change the answer. It looks like a human-like dialogue in which rational knowledge is associated with its emotional counterpart, but the algorithm’s behaviour is determined exclusively by rational rules.

Because artificial knowledge is so different from human knowledge, its integration within the hybrid knowledge management systems represents a challenge for which researchers suggest a series of solutions. However, these solutions ignore the specific nature of tacit knowledge that cannot be processed directly by GenAI. As a consequence, their models for augmenting the well-known SECI (socialization–externalization–combination–internalization) model prove to be non-workable. Thus, we identified a gap between the needs of the new hybrid human–machine knowledge management systems and the attempts made so far to provide augmentation models for the knowledge creation model developed by Nonaka and Takeuchi [2,3]. Therefore, we formulate the following research question:

RQ: How to integrate human knowledge with artificial knowledge to augment the organizational knowledge creation dynamics model developed by Nonaka and Takeuchi [2,3] within a company?

We try to answer this question by creating a new model based on metaphorical thinking [17,19], the theory of knowledge creation dynamics [2,3], artificial knowledge generation [33], and the sciences of the artificial [34]. The novelty of this model consists of leaving the SECI cycle only for human knowledge for which it was created and constructing a separate cycle for artificial knowledge. Then, we integrate both cycles into a single complex cycle, analogous to a Möbius strip. That creates a new continuum of knowledge generation and amplification. The Möbius strip is a geometric topology with only one side and one boundary. It was discovered in 1858 by August Ferdinand Möbius and Johan Benedict Listing [https://en.wikipedia.org, accessed on 16 February 2026].

Organizational knowledge creation within the hybrid knowledge management systems that integrate GenAI has a significant role in developing a sustainable business. GenAI applications have the capacity to optimize business processes and provide fast answers to complex problems generated by the external economic turbulences. Integrating human knowledge with artificial knowledge is not only a theoretical and managerial problem but also a challenge for the wise leaders to create adequate conditions for developing the company’s resilience and sustainability [40,41,42,43].

The structure of the paper is as follows: after this Introduction, we present a critical literature review and the methodology, which are followed by the Results, Discussion, and Conclusions sections.

2. Literature Review

2.1. Human Knowledge Creation Dynamics

One of the most cited models for knowledge creation is the experiential learning cycle developed by Kolb [6]. The cycle contains four main stages: concrete experience, reflective observation, abstract conceptualization, and active experimentation. Concrete experience is the first stage of that cycle, where an individual learns through perception of a certain piece of his external environment. Their sensory system generates data and information that is processed mostly by the unconscious zone of their brain [21,22]. The result of this first stage is emotional knowledge. Reflective observation is the second stage of that cycle, where emotional knowledge transforms into rational knowledge using reflection. The third stage of the Kolb’s cycle is abstract conceptualization. It is a result of rational thinking and semantic analysis. New concepts and ideas are born, and they are integrated within the known paradigms. Active experimentation is the fourth stage, where the new concepts and ideas are tested in practice. “Learning, the creation of knowledge and meaning, occurs through internal reflection about the attributes of these experiences and ideas” [6] (p. 78). If we consider time as the third dimension of that model, then the learning cycle is transforming into an evolving three-dimensional spiral.

A significant step further is made by Nonaka [1], who imagined a cycle for organizational knowledge creation dynamics called SECI (socialization–externalization–combination–internationalization). The model has been developed further by Nonaka and Takeuchi [2,3] and explained from a strategic perspective by Nonaka, Toyama, and Hirata [4] and Konno [9]. The cycle evolves along two axes: epistemological and ontological. The epistemological axis shows the conversion of tacit knowledge into explicit knowledge and then into new tacit knowledge. The ontological axis shows the transformation of individual knowledge into the team’s knowledge and then into organizational knowledge. Later on, Nonaka and Takeuchi [3] added time as the third axis, transforming a two-dimensional spiral into a three-dimensional one.

Socialization is the first stage of the cycle, where the individual’s tacit knowledge is shared among other people, within a team or organizational context, where there is a space proximity and temporal simultaneity. This sharing process is quite special because it happens as a result of observation and imitation. There is no verbal explanation or online connection when, due to the natural or symbolic language used, tacit knowledge has already been transformed into explicit knowledge. A simple example of understanding tacit knowledge sharing is to see how children learn to swim by being in a pool and imitating the body motions of their instructor. Observation and imitations are the key processes involved in this socialization stage, successful in the Japanese communities due to the focus on team spirit in their education [2,3,4,9].

Externalization is an individual process through which tacit knowledge is transformed into explicit knowledge by the human brain. It is like a digitization process because tacit knowledge, expressed as emotional knowledge, is composed of analogue signals, while explicit knowledge, expressed as rational knowledge, is composed of words and ideas. Many researchers consider this a codification process because of its use of language. Knowledge codification represents a great achievement of the human mind because it is the key driving force of our communication. Language allows us to express an infinite number of thoughts using a finite mental dictionary and mental grammar [20,44].

Combination is a social process in which explicit knowledge shared by an individual is integrated within the collective knowledge of a team, within a dialogical space called Ba [1,2,3,4]. Ba is a Japanese word that can be translated to space. However, in Japanese philosophy, Ba has multiple meanings, from physical space to virtual space. “Ba is defined as a shared context in motion, emphasizing that knowledge is embedded within and changes according to the context. This concept underlines a relational perspective that goes beyond sharing explicit knowledge and encourages the sharing and transformation of tacit knowledge” [9] (p. 60).

Internalization is an individual process consisting of the conversion of explicit knowledge into new tacit knowledge that may induce changes in the individual’s behaviour. Although it is the last stage in the cycle, it is not the final one because SECI is designed as an open cycle evolving in time [3]. Therefore, internalization will challenge the initiation of a new socialization stage at a different level on the knowledge three-dimensional spiral.

Due to its simplicity and intuitiveness, the SECI model was accepted by almost all researchers and managers, creating a new paradigm for understanding organizational knowledge creation dynamics. We have to stress the fact that this model is designed for human knowledge and contains processes which are specific to human cognition and understanding. Figure 1 presents a simplified illustration of the SECI model, where S represents socialization, E represents externalization, C represents combination, and I represents internalization. The special spiral suggests the dynamics of organizational knowledge creation.

2.2. Augmenting the SECI Model with Artificial Knowledge

The emergence of artificial knowledge and the design of hybrid knowledge management systems composed of people and computers requires a new paradigm capable of explaining the creation of organizational knowledge by integrating human knowledge with artificial knowledge. However, the integration process is not straightforward because the two types of knowledge are fundamentally different from a creation perspective [2,3,31,33]. Also, tacit knowledge is embodied and cannot be connected directly to artificial knowledge [10,11,12,13].

Harfouche et al. [45] developed the recursive theory of knowledge augmentation to explain how human knowledge and artificial knowledge can be integrated within a hybrid knowledge management system. They start with the SECI model [2,3] and introduce artificial knowledge in their equation as a result of human-in-the-loop informed AI process. “Designing and implementing a human-in-loop Informed Artificial Intelligence (IAI) can be a strong operationalization of the Human-Centric AI approach adapted to knowledge augmentation challenges” [45] (p. 4).

The authors expanded the 2 × 2 matrix of the SECI model generated by tacit and explicit knowledge into a 3 × 3 matrix generated by tacit knowledge, explicit knowledge, and artificial knowledge, creating the Knowledge Augmented Model (KAM). The augmented matrix contains five new processes due to transformations involving artificial knowledge: alimentation, aggregation, amplification, reflection, and fusion. Alimentation is the conversion of tacit knowledge into artificial knowledge. It follows in the extended matrix after socialization and externalization. However, tacit knowledge is an embodied knowledge, and it cannot be transformed directly into artificial knowledge. In practice, alimentation cannot exist. Aggregation is the conversion of explicit knowledge into artificial knowledge. In the extended matrix, it follows after internalization and combination. Such a process can be found in practice because GenAI applications are trained on both human and synthetic data. Amplification is the conversion of artificial knowledge into explicit knowledge. Here, we may find a tautology because artificial knowledge is explicit knowledge being expressed in a natural language. Reflection is the conversion of artificial knowledge into tacit knowledge. It is a kind of augmented internalization process. In the extended matrix, reflection is positioned on the vertical dimension after socialization and internalization. Fusion is the conversion of artificial knowledge into artificial knowledge, although the name suggests an integration process.

The extended matrix can be interpreted on the ontological dimension as a process of scaling up individual knowledge to the organizational knowledge, like in the SECI model. The augmentation of organizational knowledge by including artificial knowledge is an interesting idea, but the extended matrix contains processes that are not feasible because of the specificity of tacit knowledge [2,3,10,11,12,13].

Böhm and Durst [46] adopt another perspective of organizational knowledge creation by constructing a new knowledge cycle called GRAI (Generative, Receptive AI). The authors start by acknowledging the need for augmenting the SECI model, taking each process designed by Nonaka and Takeuchi [2,3] for the human mind and expanding it with the direct participation of GenAI. For instance, they suggest enhancing socialization from SECI by including AI platforms. However, when people work on AI platforms, they exchange explicit knowledge codified for those platforms and not tacit knowledge that remains as embodied knowledge. As mentioned before, playing with tacit knowledge without understanding its specificity [2,3,10,11,12,13] is dangerous. Tacit knowledge is expressed through emotions and feelings (i.e., non-verbal communication) and cannot be exchanged on AI platforms designed for codified knowledge. The same error appears in augmenting the process of externalization from the SECI model. “GenAI is expected to enhance the process of articulating tacit knowledge into explicit concepts, models or metaphors which can then be shared and communicated within organization” [46] (p. 5). Externalization is an individual process that happens in our brains, where there is no GenAI. Therefore, the whole GRAI model, consisting of the basic SECI stages modified to incorporate GenAI, ignores the specificity of tacit knowledge and the initial meanings given by Nonaka [1] to each process in the SECI model for knowledge creation dynamics.

2.3. GenAI and Sustainable Business Development

Augmenting the knowledge creation dynamics cycle using GenAI leads to sustainable business development, as reported by many researchers [40,41,42,43,47]. The augmented knowledge contributes to enhancing operational processes through better optimization of resource allocation, supply chain logistics, and energy use. In manufacturing contexts, AI-enabled predictive analytics reduces machine downtime and improves energy efficiency. The integration of human knowledge with artificial knowledge enables rapid design iteration and analysis of environmental impact, thereby expanding innovation in sustainable products and services.

The augmented organizational knowledge creation dynamics impact managerial decision-making and enable better scenario design for emergent strategies [48]. Therefore, it enhances the strategic capability of the organization, which will be reflected in the business sustainability. The new hybrid knowledge management systems can be considered as powerful dynamic capabilities [49] of the organization within its strategic framework, focusing on improving resilience and sustainable development by decreasing the level of uncertainty and speeding up the decision-making [9,47,50].

Integrating human knowledge with artificial knowledge in new complex knowledge management systems leads to more efficient and responsible management of environmental, economic, and social resources [43,47]. The question is how to design a new knowledge creation dynamics model to understand the integration of human and artificial knowledge while recognizing that tacit knowledge cannot be processed directly by GenAI. The purpose of this paper is to offer a creative solution to this complex problem using the metaphor of the Möbius strip and that of a strange attractor from fractal geometry [51,52].

Business sustainability can be better supported through the implementation of hybrid knowledge management systems, as these systems respond more adequately to the current nature of organizational knowledge, which increasingly lies at the intersection of human intelligence and artificial intelligence [53,54,55]. While Nonaka and Takeuchi’s SECI model [2,3] convincingly explained the dynamics of organizational knowledge creation based on the interaction between tacit and explicit knowledge, the emergence of artificial knowledge generated by artificial intelligence applications requires moving beyond the exclusively human-centred framework of this theory. In this new context, sustainable performance no longer depends solely on the organization’s ability to socialize, externalize, combine, and internalize human knowledge but also on its ability to coherently integrate two distinct cognitive elements: human knowledge, which is contextual, experiential, and tacit in its roots, and artificial knowledge, which is generative, scalable, and recombinable.

The central argument is that sustainable business needs cognitive systems capable of managing complexity, uncertainty, and the accelerated pace of change. These requirements can no longer be optimally met either by human intelligence alone, limited in its processing and scaling capacity, or by artificial intelligence used autonomously, which cannot directly access tacit knowledge embedded in practices, experiences, and contextual judgments [56,57,58]. For this reason, we argue that hybrid knowledge management systems have superior potential: they enable articulation between human interpretation and the analytical power of artificial intelligence, between contextual meaning and processing speed and between organizational learning and AI-assisted knowledge generation.

Within this logic, business sustainability is better supported when the dynamics of knowledge creation are reconceptualized not as an exclusively human linear spiral but as an open, continuous, and convergent process between the human knowledge cycle and the artificial knowledge cycle. Such a hybrid model is better suited to contemporary business environment because it enables an organization to transform data into information and knowledge more efficiently, as well as knowledge into decisions for suitable adaptation. Therefore, integrating human intelligence with artificial intelligence within knowledge management systems is not merely a technological innovation but an increasingly important condition for strengthening long-term organizational resilience, innovation, and sustainability [53,54,55,56,57,58].

3. Methodology

The present research is based on the theory of knowledge fields and knowledge dynamics [18,19], the theory of knowledge creation dynamics [1,2,3,4], metaphorical thinking [17,51], artificial knowledge generation [33], and design science [59,60,61]. The first two theories have already been explained. Metaphorical thinking reveals that our minds use analogies and metaphors to frame new concepts, ideas, models, and theories. Metaphors are structured analogies composed of a source domain, a target domain, and a mapping function. In the source domain, we place a known concept, while in the target domain, we place a less-known one. The mapping function serves to transfer key attributes of the concept from the source domain to the target domain, thereby enriching its semantics. For instance, in the iceberg metaphor used by Nonaka and Takeuchi [2] to explain the main features of tacit and explicit knowledge, the mapping function assigns visibility to explicit knowledge and invisibility to tacit knowledge. In the energy metaphor [17], the mapping function conveys the following ideas: energy is a field, energy manifests in different forms, and one form of energy can be transformed into another.

The design science developed in the domains of information systems, engineering, and management. Its goal is to create and evaluate artefacts that offer solutions to identified problems [59,60,61]. Those artefacts can be new concepts, models, or theories for the conceptual framework or software applications in the domain of information systems. We need design science because the principles and laws of the natural sciences cannot be applied to the sciences of the artificial [34]. According to Hevner et al., “Design science addresses research through the building and evaluation of artefacts designed to meet the identified business need. The goal of behavioral science research is truth. The goal of design science is utility” [59] (pp. 79–80). For the present research, the identified problem is the integration of human knowledge with artificial knowledge within the knowledge management systems using generative AI (GenAI) applications. The difficulty of solving this problem comes from the fact that human knowledge can be expressed as tacit and explicit, while artificial knowledge can be expressed only in the explicit form.

Based on the critical literature review, we realized that augmenting the knowledge creation cycle is not about introducing GenAI in each stage of the SECI model but in creating a separate cycle for artificial knowledge and then finding how to couple this new cycle with the SECI cycle to create a new hybrid entity. The two knowledge cycles should be comparable and compatible, meaning that the new cycle should mirror the SECI model in structure. In the first phase, we will define the constructs for the new cycle and in the second phase, the constructs will be linked to form a cycle capable of being connected with the SECI knowledge cycle. Also, we will perform an analytical evaluation of the new artefact [59].

We use ChatGPT-5.3 to generate the illustration of a three-dimensional spiral on which we placed the four stages of the SECI cycle (see Figure 1), for generating the illustration of the Möbius strip, on which we placed the stages of the SECI and CASH cycles (see Figure 2), and to generate the image of a strange attractor (see Figure 3). We reviewed the edited outputs of each illustration and completed them with the elements of the knowledge cycles.

4. Results and Discussion

Artificial knowledge creation is a rather complex process that contains a series of operations that are hard to describe [30,31,33]. While human knowledge is a product of a learning process, artificial knowledge is a product of a machine learning process. Computers analyze a large database and, using their algorithms, try to identify relationships and structures that can be used to construct patterns. In supervised learning, algorithms receive datasets as inputs and associated outputs. The task is to learn the association and define a function that maps the inputs to the outputs. For instance, the input can be a series of dog images, and the algorithm should learn to recognize dogs. It is a recognition pattern. In unsupervised learning, the algorithm learns patterns from the input without having any well-defined outputs [33]. GenAI is characterized by deep learning based on many layers of neural networks, which resemble the activity of brain neurons.

According to Hevner et al., “The design process is a sequence of expert activities that produces an innovative product (i.e., the design artefact)” [59] (p. 78). The artefacts can be constructs, models, methods, and instantiations. For the present research, we aim to design a new knowledge creation model capable of integrating a human knowledge component and an artificial knowledge component, connected together to allow a continuous and iterative flow of knowledge.

In the first phase, the focus is on defining the constructs. Because the SECI model is a part of the new model, we already have the constructs of socialization, externalization, combination, and internalization presented in the literature review section. Combination is the only construct that does not contain tacit knowledge. That makes it a good choice for the connection with the artificial knowledge cycle. Therefore, combination is the first construct for the new cycle. Its input is represented by individual explicit knowledge expressed by employees within a certain dynamic context called Ba by Nonaka and Takeuchi [2,3]. The output of the combination is represented by human knowledge selected to enter the artificial knowledge cycle through the interaction between people and machines. The next construct we define is artificialization. It constitutes a process of selecting data and curating it to feed the algorithms of GenAI. Data can be collected from human knowledge bases, training sets, and messages in prompts. Also, it can be added by using methods like retrieval-augmented generation (RAG) [31,33]. The main role of artificialization is to transform the input data into machine language data, which is totally different from human language. The next construct we define is synthetization. It represents the core construct of the new cycle because it contains AI algorithms capable of being trained to process huge datasets and generate patterns for prediction [30,31,32,33]. The name of the construct shows that it is capable of combining different elements into a whole new entity. The input for synthetization is the output of artificialization. The output is the answer for the user, but in machine language. To create a human-like dialogue, it is necessary for this machine’s answer to be transformed into human language. This transformation is performed by a construct we call humanization. The output of humanization is expressed as artificial knowledge, being a product of artificial intelligence, in the same way in which human knowledge is a product of human intelligence. Therefore, the constructs for the new knowledge cycle are the following: combination, artificialization, synthetization, and humanization (CASH). Now we establish the links between them to create coherence in data processing.

The sequence of the stages of human knowledge creation, followed by the stages of artificial knowledge creation, is the following: socialization–externalization–combination–artificialization–synthetization–humanization–internalization. From the stage of combination, human knowledge can go directly to internalization like in the original SECI model, or it can continue to be the source of data for the CASH cycle. The Möbius strip model is an integrated and complex model designed for hybrid knowledge management systems (see Figure 2). It can be considered as an artefact capable of solving the problem of integrating human and artificial knowledge within a hybrid (i.e., man–machine) knowledge management system.

The new topology of knowledge flow is no longer a three-dimensional spiral like in the SECI model [2,3]. It can be viewed as a complex, iterative process that converges over time toward a strange attractor, like the weather or Lorenz’s strange attractor. “The path of the attractor is one long, continuous track that theoretically never intersects itself and in time would visit every part of its domain, filling all the available space” [52] (p. 7). The analogy with fluid dynamics is very suggestive and offers a perspective for organizational knowledge creation dynamics other that Nonaka and Takeuchi’s model [2,3]. Figure 3 is an illustration of a strange attractor. As can be seen from Figure 3, there is a significant difference between the knowledge spiral designed by Nonaka and Takeuchi [2,3] (see Figure 1) and the strange attractor phase space. Spiral knowledge creation is an oriented process, developing from one stage to another within the SECI model, while in the Möbius strip model, the iterative interaction can go forward and backwards until a satisfactory knowledge solution is obtained. The attractor knowledge dynamics is more complex and allows a better performance than the knowledge spiral.

The analogy with a strange attractor helps researchers and managers distinguish between the SECI evolving spiral over time and the iterative process of knowledge creation and integration within a hybrid knowledge management system. The iterative process between the human agent and the AI agent is driven by the motivation of obtaining a satisfactory and unambiguous formulation for the requested solution. For instance, in our search for design science research, we had several iterations going from the general question “What is design science research?” to “How to use design science research for problem-solving?”, “What are the stages of developing an artefact?”, “What are the evaluation methods for a new conceptual model?”, and “How to perform an analytical evaluation of an artefact?”. We may say that all these iterations ultimately converged on the search target using ChatGPT.

The problem now is to evaluate the Möbius strip model (MSM). At this stage of our research, we use ex ante evaluation based on the analytical validation [59,62].

Step 1: Clearly define the model structure. The MSM structure is composed of seven constructs aligned alongside a Möbius strip surface. For the human knowledge component, there are four constructs defined: socialization, externalization, combination, and internalization. They are well-known from the SECI model [2,3]. For the artificial knowledge side, there are four components: combination, artificialization, synthetization, and humanization. Combination is a common construct for both components. These constructs are linked with causal relationships, such that the input for a construct is the output of the previous construct. Boundaries of the MSM are given by the organizational context and the complexity of its dynamics. Figure 2 illustrates the architecture of the whole model.

Step 2: Check logical consistency. The MSM has internal consistency because it eliminates the conflicts between tacit knowledge and artificial knowledge found in the KAM [45] and GRAI [46] models. Also, there are no circular dependencies. The links between constructs and the Möbius strip geometry allow a continuous flow of data and knowledge. Also, it allows knowledge iterations and interaction between humans and machines.

Step 3: Assess completeness. The whole model has a complete functional structure and an integrated architecture. There is no critical construct missing. However, this MSM can be refined and extended as needed. We demonstrated that the previous KAM and GRAI models contain constructs that mix tacit knowledge with artificial knowledge, leading to internal conflicts and practical difficulties in their implementations.

Step 4: Evaluate parsimony. The new model is rather simple and balanced because the artificial cycle has been designed in similarity with the SECI model.

Step 5: Perform ontological analysis. MSM is designed for hybrid human–machine knowledge management systems. The SECI component represents the human component, while the CASH component represents the machine component. The combination process represents the interaction between people and computers.

Step 6: Theoretical grounding. The SECI component is based on the theory of knowledge creation developed by Nonaka and Takeuchi [2,3] and the CASH component is based on GenAI theories [33]. The whole model design is based on metaphorical thinking [51].

To see how the Möbius strip model (MSM) compares with the other models developed for knowledge management systems, we present the main features of each model, as well as its limitations, in

Table 1. A comparison of the knowledge creation models discussed.

Models	Authors	Main Features	Validation
SECI	Nonaka, I. & Takeuchi, H. [1,2,3]	Designed for human knowledge and human knowledge management systems. It generates a knowledge spiral.	Analytical and practice validation
GRAI	Böhm, K. & Durst, S. [46]	Designed for human and artificial knowledge, as well as for hybrid knowledge management systems. It keeps the structure of SECI and expands each stage with artificial knowledge. It ignores the specific nature of tacit knowledge.	No validation
KAM	Harfouche, A., Quito, B., Saba, M. & Saba, P.B. [45]	Designed for human and artificial knowledge, as well as for hybrid knowledge management systems. It extends the 2 × 2 SECI matrix into a 3 × 3 matrix by integrating artificial knowledge. It ignores the specific nature of tacit knowledge.	No validation
MSM	Bratianu, C., Bejinaru, R. & Banciu, D.	Designed for human and artificial knowledge, as well as for hybrid knowledge management systems. It contains the SECI model for human knowledge and the CASH model for artificial knowledge. Both cycles are connected through the combination process. It allows a continuous flow of knowledge alongside a Möbius strip.	Analytical validation

.

The Möbius strip model is designed for knowledge-intensive organizations with hybrid knowledge management systems. The model can be implemented in any organizational context where people use GenAI applications and where there is a need to integrate efficiently human knowledge creation with artificial knowledge generation.

5. Conclusions

The emergence of artificial knowledge as a product of GenAI applications raises many challenges to knowledge managers. They should be able to understand the specificity of this new type of knowledge and to find adequate ways to integrate it with human knowledge. In the literature, several models have been proposed to augment the well-known human-centric SECI model, like GRAI and KAM. However, those proposals ignore the fact that tacit knowledge is embodied and cannot be directly linked to any GenAI application. Therefore, the augmentation process should focus only on rational, explicit knowledge.

The present conceptual paper takes a different path. Instead of augmenting each stage of the SECI cycle with artificial knowledge, we create a new cycle dedicated to artificial knowledge and then weld it to the SECI cycle. The new cycle for artificial knowledge is composed of the following stages: combination, artificialization, synthetization, and humanization. Each stage comprises a series of operations that transform the input data sequentially into artificial knowledge. Users should be aware that artificial knowledge is generated differently from human knowledge and can sometimes be wrong due to algorithmic hallucinations. Also, artificial knowledge lacks a spiritual dimension and can raise ethical issues in decision-making if managers ignore ethical values and principles.

Based on a critical literature review, we imagined a metaphorical augmented model aligned with a Möbius strip, enabling knowledge to flow continuously in a recursive, iterative manner. The welding point is the combination stage because it is the only stage containing only explicit knowledge. Within the new SECI-CASH cycle, knowledge creation cannot be represented by an evolving spiral in time. It is a more complex process that can be illustrated by a strange attractor, whose phase space resembles the Möbius strip.

The originality of this paper is based on the following arguments: (a) it focuses from the beginning on the hybrid structure of the new knowledge management systems that incorporate GenAI applications; (b) it keeps the SECI model for human knowledge as it was created for and develops a new cycle for artificial knowledge capable of interacting with the SECI cycle; (c) it uses the Möbius strip metaphor to show the continuity and iterative flow of data and knowledge; and (d) it changes the knowledge dynamics spiral of the SECI model into a phase space of a strange attractor, showing the complexity of knowledge integration. The new model has the capability of integrating human and artificial knowledge and creating a continuous and coherent flow of knowledge alongside a Möbius strip surface.

The new knowledge creation model designed for hybrid knowledge management systems is a driving force for business sustainability due to its dynamic capability to process both human and artificial knowledge in an integrated manner.

6. Limitations and Directions for Future Research

The main limitation of the present paper is that of being a conceptual paper aiming at developing a new complex knowledge creation model adapted to the new hybrid structure of knowledge management systems. The second limitation is its focus on the internal knowledge processes of a company, without clear connecting possibilities with the external business environment.

The directions for future research focus on overcoming the above limitations. It is necessary to perform an empirical validation of the model within a given company. This empirical research may uncover some other limitations of the MSM. Also, it is important to consider the interconnections among companies.