A Review of Neural Mapping and Consciousness Research on the
通用人工智能AGI测评DIKWP实验室
A Review of Neural Mapping and Consciousness Research on the Networked DIKWP * DIKWP Semantic Model
International Standardization Committee of Networked DIKWPfor Artificial Intelligence Evaluation(DIKWP-SC)
World Artificial Consciousness CIC(WAC)
World Conference on Artificial Consciousness(WCAC)
(Email: duanyucong@hotmail.com)
The networked DIKWP model is a networked framework that breaks down the cognitive process into five semantic levels (Data D, Information I, Knowledge K, Wisdom W, Purpose P), emphasizing multi-directional interaction and closed-loop feedback between the layers. Unlike the traditional linear DIKW pyramid model, this model implements inter-layer bidirectional mapping and feedback through a
5
×
5
transformation matrix, allowing output to react upon input, forming a self-contained cognitive loop. This semantic network model not only provides a theoretical basis for the interpretability of artificial intelligence and artificial consciousness but also brings new perspectives to the study of brain cognition and consciousness mechanisms, as well as the modeling of cognitive disorders. This article reviews the basic ideas of the DIKWP networked semantic model and its associated research with neural structures, consciousness theories, psychological disorders, etc., aiming to lay a foundation for the application of this model in neuroscience and cognitive science.
The DIKWP model is based on five levels: Data, Information, Knowledge, Wisdom, and Purpose.
The Data layer contains objective raw data entities, focusing on describing the "sameness" of things;
The Information layer expresses the semantic associations and context between data, emphasizing "difference" or relational properties. For example, "body temperature is elevated and blood pressure is low" is a structured description of multiple data points at the information layer.
The Knowledge layer forms knowledge rules or patterns through the structuring and generalization of information, representing a complete system of cognitive content;
The Wisdom layer makes dynamic decisions and value assessments based on knowledge, reflecting experience and insight.
Finally, the Purpose layer defines the system's subjective goals and direction, driving the interaction and feedback between all levels.
When the above five elements are tightly coupled through bidirectional feedback, the cognitive process is no longer a linear pipeline, but a highly integrated cyclic network.
In the context of neuroscience, the brain is seen as a multi-scale neural network system, and its functional networks also exhibit complex dynamic loop structures in different states. For example, recent studies have found that large-scale cortical network activity in the resting state shows periodic cyclic patterns on the scale of 300–1000 milliseconds, with each network state reappearing in a fixed sequence. Figure 1 shows a schematic of state cycling inferred by a Hidden Markov Model (HMM) from MEG experimental data, revealing that cortical network activity is essentially inherently periodic, which guarantees the periodic activation of cognitive functions. These findings coincide with the DIKWP model's view of cognition as a multi-layer network cycle, suggesting we can try to map semantic levels to neural structures such as the cerebral cortex and limbic system.
Figure 1: Schematic of Hidden Markov Model (HMM) state switching in large-scale cortical networks at resting state (adapted from). Research shows that the functional network activation of the human cerebral cortex exhibits periodic cyclic patterns, with each network state having a fixed position in the cycle. This orderly flow ensures the periodic emergence of basic cognitive functions.
The DIKWP model originally stemmed from the concept of semantic mathematics, aiming to directly construct semantic levels using a mathematical framework without pre-relying on natural language definitions. Under this framework, each layer of semantic entities can be abstracted as a mathematical structure, for example, representing data as equivalence classes in a feature space, information as a mapping that measures differences in data features, and knowledge as a semantic concept graph, thus allowing the five DIKWP semantic layers to naturally arise from internal relationships. Furthermore, the concept of the DIKWP information field treats the entire model as a holographic information space, encompassing all kinds of information from the biological level, such as genetic expression and neural firing, to the psychological level, such as memory, emotion, and conscious content. From a neuroscience perspective, this information field can be understood as the set of all meaningful symbols in the brain, where the data layer corresponds to sensory input, the information layer to perceptual extraction, the knowledge layer to stored neural representations, the wisdom layer to high-level contextual judgment, and the purpose layer to motivation and intention. This information field organizes and connects information at different levels, enabling the brain to extract useful patterns from massive signals and assign semantics. This is crucial for consciousness research, as conscious experience is essentially the subjective presentation of information content and meaning.
On this basis, the DIKWP energy field is defined as the energy distribution and flow triggered by the interaction of the information field. For example, the synchronous firing of cortical neurons generates a local electromagnetic field, and the synchronization of thousands of neurons forms a macroscopic global brain electromagnetic energy field. As stated by the CEMI field theory, the global electromagnetic field can integrate scattered neural activities into a unified conscious content. In the DIKWP model, the bidirectional interaction between semantic layers (DIKWP*DIKWP) can be seen as the propagation and feedback of energy within the system, while neural network dynamics (such as synchronous oscillations and synaptic coupling) in turn determine the spatio-temporal distribution of the energy field. Conversely, the state of the energy field (such as changes in EEG wave amplitude and phase) can regulate the activity patterns of neurons. This intertwining of the information field and the energy field indicates that the information transmission of each semantic layer in the cognitive process is accompanied by the flow and reorganization of energy, ultimately shaping the dynamic network state of the brain and affecting the formation of consciousness. For example, the spectral characteristics of brain electrical activity are significantly different in different states of consciousness (wakefulness, sleep, meditation, etc.), reflecting that the configuration of the energy field has changed the integration method of the information field, resulting in different subjective experiences.
Neural Mapping Model Construction
Mapping the DIKWP model to brain structures can help find anatomical and functional correspondences for different semantic layers. It is generally believed that:
The Data layer corresponds to sensory input and raw neural signals, such as the raw awareness activity in sensory cortical areas;
The Information layer corresponds to primary processing and feature extraction, such as the interpretation of sensory data by visual and auditory cortices;
The Knowledge layer corresponds to long-term memory and neural network representations, such as neuron ensembles in the hippocampus and association memory;
The Wisdom layer corresponds to high-level cognitive processing and contextual assessment, such as the prefrontal cortex integrating context and experience to generate insights;
The Purpose layer represents high-level semantics such as motivation and goals, corresponding to behavioral plans driven by the limbic system and frontal lobes.
Specifically, the data layer can be regarded as the basic link for the brain to receive external stimuli (e.g., vital sign measurements, sensory input); the information layer forms pattern recognition and semantic associations related to these inputs (e.g., abnormal physiological indicators associated with fever and low blood pressure); the knowledge layer integrates clinical and physiological knowledge to form diagnostic conclusions (e.g., "possible septic shock"); the wisdom layer represents the diagnosis and treatment strategies formulated by doctors or systems based on experience; the purpose layer is the goal driving the treatment (e.g., maximizing the probability of recovery).
Under this mapping framework, there is not only a bottom-up hierarchical information flow between brain regions, but also parallel horizontal network connections, realizing the DIKWP*DIKWP multi-directional interaction. For example, one study pointed out that raw signals from the data layer (sensory input) are passed up layer by layer to the prefrontal cortex to generate purpose (bottom-up path); at the same time, high-level purpose will also influence the required information and perception downwards (top-down path). Two typical computational path schematics are as follows (pseudo-code):
Bottom-up Purpose Generation:
D
→
I
→
K
→
W
→
P
, each step processes and refines the input (e.g., generating a diagnostic purpose from body temperature and blood pressure data).
Top-down Purpose Generation:
W
→
K
→
D
→
P
, high-level decisions deduce the required knowledge and data in reverse, for example, inferring new detection information needed from the treatment goal.
Through the above mapping, researchers can conduct more systematic functional connectivity and dynamic analysis at the neural level. For example, the synchronization of neural oscillations (energy field characteristic) can be linked to the information broadcast process in the global workspace (information field characteristic); or the metabolic activity of specific brain regions (energy consumption) can be associated with cognitive load and purpose-driven behavior. Studies have shown that there is a close relationship between brain network synchronization and global accessibility in different conscious modes, which is consistent with the information integration predicted by the DIKWP framework. Through this mapping, the DIKWP model provides new ideas for understanding how structures such as the cerebral cortex and limbic system support semantic levels: the data and information layers are mostly related to primary sensory networks, the knowledge and wisdom layers are mostly related to memory and decision-making networks, and the purpose layer is related to motivation and emotion networks, while all layers interact closely through large-scale cortical-cortical and cortical-subcortical loops.
Consciousness Theory Verification
Placing the DIKWP model within the framework of consciousness research allows for comparison and mutual verification with mainstream consciousness theories.
Integrated Information Theory (IIT) holds that consciousness corresponds to the maximization of the system's information integration, and the physical substrate of consciousness must have the greatest intrinsic causal power. In the DIKWP framework, the integration emphasized by IIT corresponds to the degree of interweaving and integration of high-level semantics (wisdom, purpose) and low-level semantics (information, data) in the information field.
Global Neuronal Workspace Theory (GNWT) posits that information only becomes conscious content when it is widely broadcast to multiple modules of the brain. In the DIKWP model, this is similar to the broadcast of the information field: the content of the information layer and knowledge layer needs to be propagated to the prefrontal cortex and memory system through networked DIKWP interaction to be widely accessed and applied.
Predictive Coding Theory emphasizes that the brain acts as a prediction machine, forming perception through the interaction of top-down predictions and bottom-up error signals. The DIKWP model does not directly use the framework of predictive coding, but it also supports hierarchical interaction: the purpose layer can be regarded as a high-order prediction goal, and its feedback to the knowledge and information layers guides data collection; while the perceptual results fed back from the data layer form the verification of the purpose. This is highly consistent with the bottom-up error propagation and top-down prediction in predictive coding.
Some research points out that the DIKWP information field-energy field framework can serve as a meta-framework to connect different consciousness theories: the information field corresponds to the integration of IIT and the information broadcast of GNWT, while the energy field corresponds to the physical integration emphasized by field theories (such as electromagnetic field theory). For example, under the DIKWP framework, the brain can be seen as a complex system, where the data layer includes sensory signals and neural firing, the information layer corresponds to perceptual processing and feature extraction, the knowledge layer includes memory networks and pattern representations, the wisdom layer involves the prefrontal cortex integrating experience (e.g., making decisions in specific contexts), and the purpose layer represents the motivation driving behavior (e.g., goal-oriented behavior of the limbic system and frontal lobes). Each layer not only performs bottom-up information processing but also engages in horizontal coupling through the network structure (DIKWP*DIKWP), breaking the traditional linear cognitive flow. Based on this framework, researchers can link the "broadcast-accessibility" proposed by GNW with the synchronization of neural oscillations, and the "information integration" proposed by IIT with the unity of the global electromagnetic field, providing a basis for integrating multiple theories. Studies have also begun to test these predictions, for example, coupling brainwave synchronization with working memory broadcast, and correlating the metabolic activity of specific regions with cognitive load and goal-oriented behavior. These integrated views are consistent with existing experimental evidence, providing support for the DIKWP model as an integrated framework for consciousness.
Disease Scenario Modeling
The DIKWP model also has potential in the cognitive modeling of mental disorders and neurodegenerative diseases.
Taking schizophrenia as an example, the DIKWP architecture can be used to describe various aspects of the patient's cognitive process: the data layer includes clinical observations (symptomatic behavior), patient reports (subjective experience), biological indicators (brain imaging, genetic factors), and environmental factors; the information layer processes this raw data into symptom clusters (positive, negative, cognitive symptoms, etc.) and diagnostic indicators; the knowledge layer involves the current understanding of pathogenic mechanisms (dopamine hypothesis, neurodevelopmental model) and treatment plans (drugs, psychotherapy); the wisdom layer focuses on individualized decision-making and ethics, such as formulating treatment plans based on experience; the purpose layer sets overall goals, such as symptom relief, functional recovery, and quality of life improvement. During diagnosis and treatment, multiple DIKWP transformation paths (e.g.,
D
→
I
→
K
→
W
→
P
) can be constructed to simulate the judgment process from data to purpose, and high-level goals can also reversely deduce the required information and data (
P
→
K
→
D
→
I
path). For example, in a case analysis of a schizophrenia patient, the high-level goal "symptom minimization" can be reflected in the knowledge layer's medication plan by adjusting the treatment strategy (wisdom layer), and then implemented in the biological data and patient reports (data layer) that need to be obtained, forming a closed-loop feedback. This modeling helps to achieve personalized diagnosis, for example, by updating the cognitive model based on the patient's specific data and continuously optimizing the treatment plan through the DIKWP cycle.
In depression, studies have found abnormal connectivity in the default mode network (DMN): depressed patients often show static hyperconnectivity within the DMN and between the DMN and other networks, related to ruminative thinking; at the same time, dynamic functional connectivity decreases, and the connection stability between key regions (such as the prefrontal cortex and cingulate) weakens. In the DIKWP framework, this may manifest as excessive spontaneous activation of the information layer (negative self-thinking) that cannot be effectively regulated by the wisdom layer, leading to an imbalance in the cognitive loop.
Similarly, Alzheimer's disease (AD) patients show severely weakened effective connectivity between the hippocampus and the cortex, reflecting the functional decline of the knowledge layer (memory network). In AD, hippocampal damage breaks the connection between the knowledge layer and the information layer, which in turn affects the wisdom layer's support for goals (such as memory recall), and thus can be explained in the DIKWP model as a break in the closed-loop information flow. Studies have shown that functional connectivity between the hippocampus and multiple regions such as the prefrontal cortex, cingulate gyrus, and parietal lobe is significantly reduced in AD patients.
In summary, the DIKWP model can provide a conceptual architecture for the above disease scenarios: through the analysis of the collapse of data-information-knowledge levels and the feedback loop, it helps to reveal the mechanisms of cognitive impairment and guide the design of diagnostic and therapeutic indicators.
Interdisciplinary Fusion Analysis
The research of the DIKWP model spans multiple fields.
In computational neuroscience, Graph Neural Networks (GNNs) and brain network mapping techniques can be used to achieve DIKWP neural mapping. For example, recent studies have shown that neural activity can be predicted by building models using only neuron connection structures, thereby understanding neural circuit function. Similarly, we can regard the brain's connectome as the skeleton of the DIKWP information field, and use deep learning to optimize unknown parameters, thereby simulating the dynamic transmission between different semantic layers.
In the field of psycholinguistics, multi-semantic path modeling focuses on the network representation of language and meaning, which corresponds to the knowledge layer and information layer in the DIKWP model: language concepts can be seen as semantic network nodes, and the paths between different meanings correspond to the process of information flow and knowledge generation. Researchers can use knowledge graphs or cognitive maps to specify the knowledge structure in the DIKWP model, enabling computational systems to have a "cognitive library" of human language semantics.
In terms of artificial consciousness evaluation, the DIKWP model provides an interpretable framework: by retaining semantic markers (such as the semantic explanation of diagnosis and treatment recommendations) in each step of calculation, the traceability and controllability of artificial intelligence decisions can be achieved. Organizations such as the World Artificial Consciousness (WAC) committee proposed artificial consciousness evaluations, exploring indicators based on the DIKWP model such as "semantic consistency" or "purpose completion," used to measure the consciousness level and safety of AI systems.
Furthermore, the DIKWP model shares commonalities with existing computational theories. For example, deep learning models such as feedforward neural networks and transformers can be mapped to DIKWP substructures (data layer to information layer to knowledge layer), while graph computing and knowledge graphs correspond to the interaction between the knowledge layer and the purpose layer. Through interdisciplinary fusion, we can build hybrid models that combine brain maps, language models, and energy field theories to evaluate artificial and natural consciousness from multiple perspectives.
In summary, the networked DIKWP semantic model provides a multi-level, interpretable framework for understanding cognition and consciousness. Compared with mainstream consciousness theories such as IIT, GNWT, and predictive coding, DIKWP emphasizes the unity of the information field and energy field, as well as the cyclic feedback of semantic levels, which enables it to cover multiple aspects such as information integration, broadcast accessibility, and high-level purpose. At the neural mapping level, the model maps the cerebral cortex, limbic system, and neural networks to the five DIKWP semantic layers one by one, providing new ideas for systematically analyzing brain function. In terms of disease modeling, the DIKWP framework can describe the information loss and feedback breakdown at different levels in cognitive impairment, and has potential diagnostic and predictive value. Future research needs to further verify the predictions of DIKWP based on neuroimaging and cognitive experiments, for example, examining the correspondence between semantic levels and brain region activity, as well as the specific damage patterns of the DIKWP cycle in patients with mental illness. At the same time, interpretable computational models should be developed, avoiding the use of abstract concepts that are difficult to understand intuitively, such as high-dimensional tensors or entropy, and turning to analysis methods based on semantic graph structures and network mapping, thereby enhancing the transparency and safety of artificial consciousness systems.
The networked DIKWP*DIKWP semantic model provides a unified perspective for studying the mechanisms of consciousness and cognitive impairment by viewing the cognitive process as a dynamic cyclic network between semantic levels. This model not only has a corresponding relationship with modern consciousness theories, but can also be mapped to the neural structures and functional networks of the brain, thereby achieving interpretable cognitive modeling. Preliminary applications to diseases such as depression, schizophrenia, and Alzheimer's show that the DIKWP framework can reveal abnormalities in information flow and feedback at different semantic levels under these conditions, providing inspiration for multimodal diagnosis and treatment plan design. With the development of interdisciplinary methods, based on tools such as graph neural networks, brain maps, psycholinguistics, and artificial consciousness evaluation, the DIKWP model is expected to become a bridge connecting cognitive science and artificial intelligence, achieving a deeper understanding of consciousness and cognitive impairment.
References: This article cites multiple cutting-edge research results, including those from "Nature Neuroscience," "Neuron," "Trends in Cognitive Sciences," etc. See the citations in the text for details.
[1] Instability of default mode network connectivity in major depression: a two-sample confirmation study | Translational Psychiatry https://www.nature.com/articles/tp201740?error=cookies_not_supported&code=9f8432b8-1e60-41b6-b6a0-8a07a0acc5ce [2] Connectome-constrained networks predict neural activity across the fly visual system | Nature https://www.nature.com/articles/s41586-024-07939-3?error=cookies_not_supported&code=61eacb18-32cc-4ebd-9567-b1efc1ca1e81
玩透DeepSeek：认知解构+技术解析+实践落地
人工意识概论：以DIKWP模型剖析智能差异，借“BUG”理论揭示意识局限
人工智能通识 2025新版 段玉聪 朱绵茂 编著 党建读物出版社
邮箱|duanyucong@hotmail.com
