Background and significance of the project

Addiction, including substance addiction and behavioral addiction, has become a serious public health and social problem. Whether it is drug abuse (such as opioids, nicotine, alcohol, etc.) or behavioral addiction such as online games and gambling, it involves abnormalities in the brain's reward circuits, manifested as compulsive pursuit of reward stimuli and uncontrollable use. Studies have shown that all addictive substances activate the common mesolimbic dopamine reward pathway, which is the dopamine pathway that projects from the ventral tegmental area (VTA) of the midbrain to regions such as the nucleus accumbens through the midbrain-limbic system. This midbrain-limbic pathway (see figure below) plays a key role in the regulation of motivation, reward, and reinforcement learning, and its dysfunction is thought to be one of the core mechanisms of addiction initiation and maintenance. Figure: The midbrain-limbic dopamine reward pathway shows the purpose (the blue route indicates dopamine projection from VTA to nucleus accumbens, etc.).

Due to the complexity and persistence of addictive behaviors, current withdrawal treatments have limited efficacy and relapse rates remain high. Addiction is seen as a chronic, recurrent brain disorder that requires long-term management rather than a one-time cure. Even after treatment, about 40%~60% of addiction patients will relapse within half a year. Existing drug-assisted treatments (eg, methadone for opioid addiction, naltrexone for alcohol addiction, etc.) alleviate withdrawal symptoms to some extent, but there are also significant limitations and side effects. For example, methadone replacement therapy carries its own risk of addiction, naltrexone may cause hepatotoxicity and adherence problems, and the cigarette withdrawal drug varenicline can cause emotional side effects. These problems have resulted in many patients not being able to adhere to treatment, highlighting the urgency of innovative intervention strategies.

At the same time, the treatment of behavioral addiction (such as game addiction and gambling addiction) lacks specific drugs, mainly relies on psychological interventions, and the effect is easily affected by individual subjective factors. The study found that there are similarities between online game addiction and substance addiction in terms of neural mechanisms, such as impaired value decision-making function and reduced sensitivity to loss aversion. Functional imaging studies have also shown that the reward processing pathways of game addicts function abnormally, and similarly exhibit overreaction to addiction cues and weakened cognitive control in drug addicts. However, the current understanding of the brain circuit mechanism of behavioral addiction is incomplete, and its causal interaction network across brain regions needs to be analyzed urgently.

Clinical needs: In the face of high recurrence rate and limited intervention methods, there is an urgent need to deeply analyze the occurrence and development process of addictive behaviors from the perspective of neural mechanisms, and develop new intervention strategies accordingly. These include: (1) revealing neuromolecular, cellular, and cerebral circuit abnormalities as well as cognitive and emotional dysfunction of addiction at multiple levels; (2) to evaluate the mechanism of action and limitations of existing detoxification drugs and therapies, and to find the causal chain of side effects; (3) the design of novel withdrawal interventions that are more effective and have fewer side effects; (4) Formulate digital and personalized intervention programs for the increasingly prominent problem of behavioral addiction. The National Key R&D Program lists "Neural Mechanism Analysis of Addiction and Withdrawal Intervention Strategies" as a key guideline direction, which is based on the above urgent needs.

New opportunities for artificial intelligence and artificial consciousness: In recent years, the rapid development of artificial intelligence technology has provided new means for the study of complex brain diseases. Among them, the artificial consciousness model has attracted attention, which tries to make AI have the elements of human-like consciousness, so as to better simulate human high-level cognitive and behavioral processes. This project intends to introduce the "Data-Information-Knowledge-Wisdom-Purpose" artificial consciousness model proposed by Chinese scholar Professor Yucong Duan. The model adds the top layer of "Purpose/Purpose" to the traditional DIKW (pyramid) architecture, and realizes two-way feedback and iterative update of semantics at all levels through mesh interaction. This system provides a framework for formalizing and structuring human cognitive processes, enabling AI to simulate human perception, cognition, and motivational decision-making processes in an "explainable and controllable" manner. In particular, the DIKWP model emphasizes the "Purpose" drive: the AI system not only processes the linear process from data to knowledge when sensing and making decisions, but is also led by the built-in goal/purpose layer to achieve self-regulation and iterative optimization. This purpose-driven cyclic structure enables AI to have the ability to self-explanatory and self-optimized prototypes, which is seen as an important path to artificial consciousness. With the help of this model, we are expected to simulate the information processing and motivational driving mechanisms of the brain of addicted individuals at different levels, and establish concepts such as "semantic behavioral causal chain" and "cognitive addiction pathway" to understand addictive behavior from a new perspective.

In short, the significance of this project is to combine the latest artificial consciousness models and semantic technologies to propose new ideas and tools for interdisciplinary integration of the major topic of addiction neural mechanisms and intervention strategies. Guided by clinical needs, the project strives to make original breakthroughs in the analysis of addiction mechanisms and withdrawal intervention methods, and provide scientific support and technical reserves for the prevention and treatment of addiction.

Research status and development trend at home and abroad

Advances in the neurobiological mechanisms of addiction

Circuits and molecular mechanisms of substance addiction: A large number of studies have revealed that substance addiction involves the remodeling of multiple neural circuits in the brain, especially the midbrain-limbic dopamine reward system. Addictive drugs (such as cocaine, morphine, nicotine, etc.) can enhance the release of dopamine from the ventral tegmental area (VTA) of the midbrain to the nucleus accumbens (NAc) to varying degrees, triggering a strong reward effect and disrupting the inhibitory control of behavior in the frontal cortex. As shown in the figure, the midbrain-cortex-limbic circuit, consisting of the prefrontal cortex (PFC), nucleus accumbens (NAc), and ventral tegmental area (VTA), plays a key role in behaviors such as rewarding learning, impulsivity, and addiction. Long-term addiction can lead to changes in synaptic plasticity in this circuit, such as chronic drug exposure inducing a disturbance in the excitatory/inhibitory balance of the cortical-nucleus accumbens pathway, which reinforces drug-seeking behavior. In addition, the addiction process is accompanied by a series of molecular adaptations, such as the accumulation of transcription factors such as ΔFosB, the downregulation of dopamine D2 receptors, and the changes in glutamate receptor subunits, which shape the pathological dependence of the brain on drugs.

In addition to the reward circuit, there has been an increase in attention in recent years to counter the reward/stress pathway. Repeated drug use not only triggers overactivation of the reward system, but the subsequent phase of withdrawal from negative emotions is associated with the stress system in the brain (e.g., activation of the CRF stress pathway in the middle amygdala). Addiction is thought to be the result of a combination of reduced reward system function and increased stress system functioning. This also explains why addicts find it difficult to quit because of the pain of withdrawal, even if they no longer get pleasure.

In terms of domestic research, many teams have used modern neuroscience technology to deeply analyze the addiction circuit. For example, the team of Professor Xin Wenjun of Sun Yat-sen University used optogenetics and in vivo calcium imaging to find two fine neural circuits projected to the anterior and posterior bundle nucleus (IPAC) in the conditional position preference (CPP) model of morphine addiction in rats: one from VTA GABAergic neurons, which is involved in the formation of reward effect; The other is a D1 dopamine receptor-positive neuron from the shell region of the nucleus accumbens, which is involved in the formation of environmental memory. Together, both contribute to the establishment of drug addiction. This achievement reveals the role of a new connection pathway between the midbrain, the nucleus accumbens, and the extended amygdala (IPAC) in addiction, and provides leading evidence for understanding the circuit mechanism of addiction.

In addition, domestic scholars have also focused on the long-term plasticity of addiction. For example, studies have reported the dual roles of neurogenesis and synaptic plasticity in the hippocampus and prefrontal lobes in addiction formation and withdrawal. Studies from Peking University and other institutions have found that chronic morphine exposure causes microglia to release inflammatory factors, which regulate dopaminergic neuronal activity and aggravate addictive behavior. These advances suggest that the neural mechanisms of addiction are complex multi-level networks, including neural circuit reorganization, neuroimmunity, and plasticity changes, which provide key background knowledge for this project.

Cognitive neural mechanism of behavioral addiction: Compared with substance addiction, behavioral addiction such as gambling and gaming has no direct exogenous drug effect, and its addiction characteristics are more reflected in the abnormality of cognitive processes and decision-making mechanisms. Despite the lack of drug stimulation, behavioral addicts exhibit similar alterations in brain function as substance addicts. For example, functional MRI studies have shown that when online game addicts complete decision-making tasks, the activation of higher-order cognitive regions such as the orbitofrontal cortex and dorsolateral prefrontal cortex is reduced, and the value assessment function such as loss aversion is impaired. At the same time, reward-related nucleus accumbens and dopamine pathways overreact to game-related cues. This supports the neurological basis for behavioral addictions to share part of substance addiction.

Theoretically, behavioral addiction can be seen as the result of dysfunction of reinforcement learning mechanisms: an individual's preference for immediate reward (instant gratification) is too strong, and there is insufficient consideration for delayed punishment or long-term benefits. This is related to the imbalance between model-based and habit-free decision-making systems in the brain. Normally, model-driven (supported by the prefrontal-hippocampus, etc.) and model-independent habituation systems (supported by the striatum, etc.) work together. However, in the addictive state, the transition from model-driven to model-independent behavior has been observed, and addicts rely more on unconscious habitual responses than on purposeful rational decision-making. Studies have shown that this shift is associated with specific areas of the brain, such as the orbitofrontal cortex and dorsolateral striatum, that change their function during addiction, leading to a shift in behavior from goal-oriented to habit-driven. This finding suggests that we need to pay attention to changes in the decision-making system of addicts and incorporate them into the model.

In conclusion, important progress has been made in the study of the neural mechanism of addiction at home and abroad, which provides a solid foundation for this project. However, most of the existing studies focus on single-level (molecular/cellular or behavioral/cognitive) mechanisms, and lack of cross-level integrated models. Moreover, the understanding of the mechanism of behavioral addiction needs to be deepened. We need a framework that integrates multiple levels of data and takes into account both biological and cognitive factors to comprehensively unravel the mechanisms of addiction.

Withdrawal intervention strategies and limitations

Medication-assisted therapy: Medication-assisted treatment (MAT) is a common strategy for substance addiction. For example, opioid addiction with methadone, buprenorphine replacement therapy, or naltrexone blockade therapy; alcohol addiction with naltrexone, disulfur, and acrosate; Nicotine addiction uses nicotine replacement therapy, bupropion or varenicline, among others. These drugs have been shown to be effective in reducing cravings and reducing withdrawal symptoms. However, its limitations are also obvious: (1) the replacement therapy itself may be addictive (e.g., methadone needs to be taken for a long time and there is a risk of dependence); (2) poor compliance with blockers, and patients often have difficulty in adherence; (3) Significant drug side effects, such as naltrexone can cause liver damage, varenicline may induce depression and abnormal dreams, etc.; (4) Individual differences are large, the efficacy is different, and there is a lack of accurate personalization. As a result, the long-term success rate of drug dependence alone is limited, and other interventions and the exploration of new drugs are needed.

Modern pharmacological research is also looking for new targets for addiction treatment, such as CRF receptor antagonists that regulate stress responses, NMDA receptor modulators that affect memory restructuring, and pro-neuroplasticity pro-molecules. But so far, no revolutionary new drugs have been introduced. On the one hand, the neural adaptation mechanisms of addiction are complex and diverse, and single-target therapy is often difficult to achieve. On the other hand, the R&D cycle of new drugs is long and risky. As a result, some studies have begun to explore drug repurposing strategies, i.e., new uses of existing drugs, such as antiepileptic drugs and anti-inflammatory drugs in reducing addiction relapse. This requires a deep understanding of existing drug mechanisms and finding entry points in their networks of action, which is part of the reason why this project is intended to be implemented using semantic modeling.

Behavioral and psychological interventions: Psychosocial interventions are equally important in addiction treatment, including cognitive behavioral therapy (CBT), motivational enhancement therapy (MET), mindfulness training, and others. These methods aim to change the negative cognitive and behavioral patterns of addicts and improve their ability to cope with cravings and stress. Traditionally, these therapies have been administered face-to-face by therapists, which has the disadvantages of high cost and limited coverage. In recent years, there has been an uptick in digital cognitive interventions, such as the delivery of CBT courses and support through mobile apps and online platforms. Several randomised controlled trials have demonstrated that digital CBT programmes are effective in reducing alcohol and drug use. The US Food and Drug Administration (FDA) has also approved prescription digital therapeutics for the treatment of SUDs, such as reSET applications, indicating that digital interventions are becoming an important adjunct. However, most current digital therapeutics lack personalized and intelligent feedback: i.e., little dynamic adjustment based on the patient's real-time state (cognitive load, mood swings, brain activity, etc.). This is where AI comes into play, by analyzing user behavior and physiological data to intelligently adapt the content and intensity of interventions.

Brain-Computer Interface and Neuromodulation: Emerging neuromodulation technologies offer hope for refractory addiction cases. For example, transcranial magnetic stimulation (TMS) is used to modulate prefrontal cortex activity and has been shown to have some effect on reducing cravings in cocaine and alcohol addiction; Deep brain stimulation (DBS) is tested in patients with refractory drug addiction, targeting the nucleus accumbens to inhibit excessive reward drive. Brain-computer interface (BCI) combined with neurofeedback training is also a hot topic. In recent years, the University of Science and Technology of China and other units have built cognitively guided brain-computer interface neurofeedback systems for nicotine addiction intervention. By allowing the subjects to perform cognitive training under EEG monitoring and real-time feedback on their brain activity patterns, the study found that the subjects' brain responses to smoking cues were significantly reduced after training, and the accuracy of the classifier in discriminating smoking-related EEG reached about 70%. This suggests that BCI-neurofeedback has the potential to reshape the brain response patterns of addicts and reduce sensitivity to addiction cues. It is foreseeable that future interventions will be more combined with non-invasive brain-computer interfaces to achieve monitoring and real-time regulation of brain status. This project intends to combine brain-computer interface with artificial consciousness model to explore a new model of closed-loop intervention: to judge the patient's state through the semantic interpretation of brain signals, and to generate corresponding intervention instructions (such as prompts and stimuli) by artificial intelligence to form a human-computer collaborative withdrawal support system.

Domestic and international trends: In general, addiction intervention strategies are developing in the direction of diversified integration and precise personality. On the one hand, it emphasizes bio-psycho-social integrated interventions, and on the other hand, it introduces advanced technologies such as AI, big data, BCI, etc. to optimize therapies. There are already interdisciplinary collaborations in the world dedicated to the development of so-called "digital vaccines", i.e. the prevention and reduction of addictive behaviors through gamified digital therapeutics. China also attaches great importance to related research, and new methods have emerged that integrate virtual reality exposure therapy, mobile app follow-up, and wearable device monitoring. However, there is currently a lack of a unified platform to integrate brain imaging, brain-computer interfaces, cognitive training, and AI decision support. This project proposes to build an "artificial consciousness and human-machine collaborative intervention platform", which is an innovative measure in line with this development trend, and its ultimate goal is to achieve all-round and whole-process intelligent intervention for addicts.

DIKWP Artificial Consciousness Model and Advances in Semantic Technology

The DIKWP artificial consciousness model :D IKWP model is composed of five layers: Data, Information, Knowledge, Wisdom and Purpose, which is an artificial consciousness framework that integrates cognition and semantics. Professor Yucong Duan's team was the first to propose this model, which enables AI to have an intrinsic representation and drive of the ultimate goal by adding a "Purpose" layer to the traditional DIKW architecture. The model adopts a network structure rather than a linear hierarchy, and there is a two-way semantic feedback between the layers: the perception data at the lower level can be gradually upgraded to the high-level abstraction, and the high-level purpose can also guide the selection and processing of low-level information. This feature makes the DIKWP model very suitable for causal chain analysis and decision interpretability studies of complex behaviors. For example, under the DIKWP framework, we can think of addictive behavior as the result of a combination of the data layer (sensory stimuli and physiological signals), the information layer (environmental cues and situations), the knowledge layer (the individual's past experience and memory), the wisdom layer (the assessment and balance of the consequences of the behavior), and the purpose layer (intrinsic motivation and goals). The interaction of each layer forms a semantic behavioral causal chain: for example, environmental triggers (I) trigger memory craving (K), weakens long-term hazard cognition through value judgment (W), and finally executes the seeking behavior (D behavior output) driven by purpose (P). This hierarchical description helps us identify key nodes in the chain of addictive behaviors (e.g., identifying cognitive biases in the "knowledge layer" or value shifts in the "purpose layer" that are driving addiction) and develop targeted interventions.

The DIKWP model is also accompanied by the development of a series of key technologies: (1) Dual-cycle architecture: adding metacognitive loops in addition to the basic cognitive processes to realize AI self-monitoring and self-regulation is considered to be an important path to give AI initial self-awareness. (2) Interpretable Cognitive Operating System (ACOS): The DIKWP model is embedded in the AI system to form a "semantic operating system", and the reasoning process of the large model is divided into five monitorable links: D, I, K, W, and P, and each step has a mathematical definition, so as to realize the traceability of the whole process of AI decision-making. (3) Complex semantic network and energy/information coupling:D IKWP supports the construction of multimodal semantic graphs, coupling the internal cognitive state with the external environment information/energy, and facilitates AI adaptive collaboration in scenarios such as the Internet of Things and brain-computer interfaces. (4) LLM semantic adaptation: To solve the problem of illusion and purpose shift of large language models (LLMs), DIKWP provides structured prompts and multi-round dialogue semantic control to enhance the controllability of LLMs. The above technological developments show that the DIKWP model is becoming practical and systematic, providing new ideas for AI applications in various fields.

In addition to the DIKWP model, other theories (such as global workspace theory IIT, attention mediation theory, etc.) have emerged in the international research of artificial consciousness, but the outstanding feature of DIKWP is the introduction of the Purpose layer and semantic feedback mechanism, giving it a unique edge in cognitive modeling and explainable AI. As the inventor of the model, Professor Yucong Duan, an academician of the International Academy of Advanced Technology and Engineering, has been granted more than 100 patents, covering fields ranging from large model training and cognitive operating systems to AI governance and security. The DIKWP model and its artificial consciousness system were also unveiled on academic platforms such as the first World Artificial Consciousness Conference in 2023, becoming one of the frontier topics in the research of "explainability" and "value alignment" of artificial intelligence. It is foreseeable that in the future evolution of AI to human-like intelligence, the DIKWP model is expected to become a key basic framework, leading the transformation of AI systems from "tools" to "self-aware agents".

Semantic Communication and Semantic-Driven Programming: Some derivative ideas are also being formed around the DIKWP model. Semantic communication refers to the transmission of high-level semantics instead of low-level bits in a communication system, thereby improving efficiency and intelligence. This is a hot topic in the field of 6G communication and human-computer interaction. A DIKWP-based semantic communication system (DIKWP-SC) has been proposed, which uses the subconscious layer (D/I/K) to automatically process semantics, and the consciousness layer (W/P) to make abstract decisions, so as to realize the purpose understanding and active fusion in the process of information transmission. The "Consciousness Bug Theory" has also been used to improve the security mechanism to ensure that the information in semantic communication is transmitted efficiently without introducing semantic distortion or malicious purpose. This is instructive for the data transmission of brain-computer interfaces and the communication of patient semantic feedback in this project, that is, semantic layer understanding should be integrated into the signal exchange, irrelevant data should be filtered, useful information should be extracted, and the purpose should be ensured.

In terms of semantic-driven programming, the DIKWP model is being used to develop a new generation of programming paradigms. Traditional programming regards the problem as a process of input->-processing-> output, while artificial consciousness programming requires a five-layer structure of DIKWP in each stage of input, processing, and output, and the global solution is driven by the top-level purpose. This means that the program is no longer the execution of fixed instructions, but an autonomous solution process with a "purpose". The key steps include mapping all inputs to the DIKWP structure content (aware of the purpose), waking up and satisfying the desired purpose of the output during processing. The Purpose driver makes the program run in a loop: the Purpose guides the parsing of inputs, information extraction, knowledge reasoning, and Wisdom decision-making. This paradigm allows the system to be self-regulating and self-explanatory, which is very suitable for the development of intelligent intervention systems that we need. For example, we can design a semantic-driven decision-making engine for the withdrawal intervention platform: dynamically adjust the intervention strategy W to meet the final rehabilitation goal P according to the patient's different state inputs (physiological data D, behavioral information I, medical history knowledge K, etc.). This kind of program can continuously evaluate its own effect during operation (Wisdom layer self-reflection) and adjust the strategy (Purpose layer correction) to form a closed-loop optimization.

Overall, the development of the DIKWP artificial consciousness model and its related semantic technologies provides a new toolbox for this project. By applying these cutting-edge concepts, we can introduce semantic-level analysis and control into the complex system of addiction neural mechanisms and interventions, which is expected to break through the limitations of traditional methods and achieve more comprehensive and in-depth research and more efficient and intelligent interventions.

Scientific issues, key technical challenges and innovations

Based on the above review, it can be clarified that the core scientific problems and key technical challenges to be solved in this project are as follows:

1. Unified modeling of cross-level causal mechanisms of addictive behavior: Addiction involves multi-level mechanisms from molecular to behavioral, and traditional research is mostly limited to a certain scale. The overarching scientific question for this project is how to integrate neuromolecular**/cellular changes, brain circuit dynamics, and cognitive/behavioral processes within a unified framework** to build a cross-level causal model of addictive behavior? This requires breaking through the semantic divide at different levels of data and linking biological mechanisms to psychological processes. The key challenge is: how can multi-source heterogeneous data (e.g., molecular pathways, physiological signals, cognitive assessments) be semantically integrated? How does brain imaging and behavioral data map to the layers of the DIKWP model? The innovation lies in the introduction of the DIKWP artificial consciousness model as a bridge, and its semantic network is used to associate neural activity with behavioral purpose, so as to construct a semantic causal chain model of addictive behavior for the first time. This model will shed, for example, how the downregulation of dopamine receptors affects cognitive control (knowledge layer) through the meso-circuit and ultimately leads to the development of relapsing behaviors (purpose-driven behaviors).

2. Analysis of the semantic mechanism of existing withdrawal drugs and interventions: At present, the mechanisms and action networks of various detoxification drugs and psychotherapies have not been systematically compared. The second scientific question to be answered in this project is: How can semantic modeling and intelligent reasoning be used to comprehensively unravel the mechanisms and limitations of existing withdrawal interventions? The technical challenge is that it is necessary to construct a semantic knowledge graph covering information such as various drug targets, neural pathways, symptom improvement and side effects, as well as the cognitive process chain affected by different psychological interventions. This requires AI to perform complex causal reasoning, such as "what is the pathway of a drug to reduce craving, and what is the causal basis of its side effects (depression, etc.)". The innovation lies in the use of the knowledge layer and the wisdom layer of the DIKWP model to formally represent and cause and effect the knowledge of these mechanisms, and find out the shortcomings of each intervention (such as only acting on the reward pathway but ignoring the emotional pathway) and the causes of side effects (such as blocking the chain of cognitive decline caused by specific receptors). This is the first time that artificial conscious semantic reasoning has been used for comparative analysis of addiction intervention mechanisms, which can provide a basis for improving existing therapies.

3. Semantic-driven design of new withdrawal intervention strategies: Innovative interventions are needed after the limitations of existing therapies are discovered. The third core question is: how to design and optimize new intervention molecules or strategies based on semantic-driven models? This involves two aspects: one is molecular design, that is, the discovery of new compounds with specific semantic functions (e.g., "attenuation of reward memory" or "enhancement of cognitive control purpose"); The second is the design of non-pharmacological strategies, such as new cognitive training paradigms or stimulation protocols. Key technical challenges include translating semantic targets into specific molecular structures or interventions. We need to develop a semantic-> molecular mapping method, which may use knowledge graphs and generative models to identify biological targets that can achieve predetermined semantic effects, and then generate or screen molecules based on them. At the same time, a simulation environment needs to be established to evaluate the effectiveness of these new protocols on animal models or virtual patients. The innovation lies in proposing a "semantic-driven drug innovation model": unlike traditional empirical screening or blind machine learning, we design new drugs/new protocols guided by a clear semantic mechanism of action, and use artificial awareness models for simulation feedback optimization. This will introduce a new paradigm of rational design in the field of addiction treatment.

4. Semantic spatial representation and intervention of behavioral addiction: Compared with substance addiction, game and gambling addiction emphasize more cognitive-affective processes. The fourth question that needs to be solved in this project is: how to construct a semantic space model of behavioral addiction under the DIKWP framework and use artificial consciousness agents to intervene and simulate it? The technical challenge is that behavioral addiction lacks direct physiological drives, so its semantic representation needs to capture factors such as cognitive biases (e.g., the gambler's fallacy), mood swings, and social environment, and map these high-dimensional factors to the state within the artificial consciousness model (knowledge/wisdom layer). At the same time, we want to have AI agents "play" the addict and test the effects of different interventions (e.g., time limits, financial penalties, mindfulness reminders) in a virtual environment. The difficulty that needs to be solved is the simulation of the causal chain network across brain regions, and how artificial consciousness models produce decision-making biases and correction processes similar to those of humans. The innovation lies in the construction of an "artificial consciousness addiction", that is, to let the AI present the cognitive and behavioral characteristics of the addict in the semantic space, and then apply intervention strategies to it to observe the outcome. This approach will provide a safer and more controllable testing ground than real-life experiments for pre-assessing the effectiveness and potential problems of interventions, and is an original tool for behavioral addiction research.

5. The construction of artificial consciousness and human-machine collaborative intervention platform: Ultimately, we hope to integrate the above models and methods into clinical applications. The fifth question is: how to build a collaborative intervention platform that integrates artificial consciousness AI and human therapists**/patients to achieve the integration and closed-loop control of multi-modal information? Technical challenges include: simultaneous acquisition and analysis of multi-source data (brain imaging, EEG, behavior, subjective feedback);** real-time decision-making and safety monitoring of the Artificial Consciousness Operating System (ACOS); and the effectiveness of the human-computer interface (ensuring that AI recommendations are understood and adopted by physicians and patients). In particular, safety and controllability need to be addressed: AI must be explainable and ethical in intervention decision-making, and avoid the adverse effects that may be caused by false feedback. Our innovative approach is based on DIKWP's "semantic operating system" technology, which divides each step of intervention decision into supervised semantic units, which are reviewed by clinical experts or verified by pre-set rules, so as to ensure the safety and reliability of interventions. The platform will integrate the neuroimaging/brain-computer interface input module, semantic behavior analysis module, artificial consciousness decision-making module and feedback execution module to form a closed-loop intervention ecology. This is a first-of-its-kind artificial awareness-driven addiction intervention system that is groundbreaking in terms of technology integration and clinical utility.

In summary, the innovation of this project is reflected in the following: the artificial consciousness (DIKWP) model is comprehensively applied to the study of addiction mechanism and intervention, and the semantic link from the basic neural circuit to the high-level cognitive behavior is opened; develop semantic-driven analysis and design approaches to enable in-depth causal analysis of existing therapies and rational innovation of new therapies; Build a platform for artificial consciousness and human-machine collaboration, and open up a new model of intelligent withdrawal intervention. These are the first of their kind at home and abroad, and are expected to have an important impact at the intersection of addiction science and artificial intelligence.

Research objectives and key questions to be addressed

Focusing on the above scientific issues and challenges, the project establishes the following overarching and decomposition objectives:

The overall goal is to develop an innovative method of addiction mechanism analysis and intervention strategy based on the DIKWP artificial consciousness model, and to build a platform that integrates multi-level semantic modeling and intelligent intervention feedback. Theoretically, the cross-level semantic causal mechanism of addictive behavior is revealed. In terms of methodology, this paper breaks through the bottleneck of existing addiction intervention technology and proposes a new semantic-driven intervention program. In terms of application, the demonstration application of artificial consciousness model in human-machine collaborative withdrawal intervention is realized.

Specific sub-objectives:

• Objective 1: To establish a multi-level semantic model of the neural mechanisms of addiction. The molecular pathways, neural circuits, and cognitive processes related to addiction are mapped into the semantic elements of each layer of the DIKWP model to form a computable behavioral dynamics model and a semantic causal chain. The model was used to explain the internal mechanism of typical addictive behaviors (such as craving and relapse), and its explanatory power for experimental observations was verified. Indicators: Construct at least one semantic model of substance addiction (e.g., opioids) and one set of behavioral addiction (e.g., game addiction); The model was able to reproduce key phenomena reported in the literature (e.g., increased cravings due to dopamine changes caused by a drug).

• Aim 2: To develop a semantic analysis and reasoning system for withdrawal interventions. Based on the knowledge graph and causal reasoning, the DIKWP semantic communication and reasoning framework was introduced to model the mechanism semantic modeling of 3-5 major detoxification drugs and psychotherapies, and find out their respective paths and limitations. Indicators: mapping the semantic causal network for each intervention; At least 3 hypotheses of unspecified causes of side effects or failures should be put forward and consistent with the literature data (e.g., the prediction relationship in the knowledge graph was verified by association analysis).

• Objective 3: To propose new interventions and programmes. Using a semantic-driven approach to drug design, several candidate molecules or interventions are generated to target unmet semantic goals (e.g., alleviate abstinent negative emotions during withdrawal, enhance cognitive control). The effect was evaluated through molecular docking simulation, animal experiments, etc. Indicators: At least one new compound was found to show superior withdrawal efficacy (e.g., 30% reduction in recurrence rate and reduction of side effects) in animal model trials (e.g., reduction in relapse rate > 30%); or design a new cognitive training/brain stimulation regimen that has been shown to be feasible in animal or small clinical trials (e.g., prolonging the latency for relapse of addictive behaviors).

• Objective 4: To build an artificial consciousness simulation and intervention test platform for behavioral addiction. For game addiction and gambling addiction, artificial consciousness agents were developed to simulate their cognitive-emotional-reward processes, and the effects of multiple interventions (such as limiting, reward substitution, and cognitive correction) were tested in a virtual environment. Indicators: The platform can simulate the behavior change curve of addicts under different interventions; Consistent with trends in real-world clinical interventions (e.g., simulations show that time restriction significantly reduces gaming addictive behaviors, which is consistent with clinical observations). The intervention effect evaluation report was output to provide a quantitative basis for the actual intervention strategy.

• Objective 5: Develop a prototype system for human-machine collaborative withdrawal intervention in artificial consciousness. It integrates EEG/brain imaging acquisition module, DIKWP semantic analysis engine, and treatment interactive interface to realize real-time status monitoring, intelligent feedback, and therapy recommendations for addicts. Conduct a small sample trial to evaluate the feasibility of the system. Indicators: The prototype system integrates brain-computer interface to achieve >70% accuracy of brain state recognition (e.g., accurate recognition craving); Individualized feedback to the patient meets safety norms and is assessed by experts to aid withdrawal (e.g., the patient's self-reported cravings are less intense than before the system was used); The software and hardware of the system run stably and pass the technical appraisal.

Through the realization of the above sub-goals, the scientific questions raised by the project will be gradually answered, and the expected innovation breakthroughs of the project will be achieved. The objectives of each part of the project support each other: the mechanism model provides the theoretical basis, the semantic analysis finds the direction of improvement, the new scheme design and simulation verify its effectiveness, and finally integrates it into the intervention system to apply the verification. This closed loop from theory, > method, > application ensures the implementation of project objectives and the transformation of results.

Research content and technical route

In order to achieve the above goals, this project intends to set up five research directions (tasks), corresponding to the research content described in the guide: (1) multi-level analysis of addiction mechanisms, (2) semantic modeling of withdrawal drugs, (3) semantic-driven intervention innovation, (4) semantic modeling and simulation of behavioral addiction, and (5) integration of artificial awareness intervention platform. The main contents, technical routes and innovation points of each research direction are as follows:

Research Direction 1: Analysis of the neural mechanism of addiction behavior based on artificial consciousness model

Main contents: Using the DIKWP artificial consciousness framework, this paper explores the neuromolecular, cellular and circuit mechanisms of substance addiction and behavioral addiction, and establishes behavioral dynamics and semantic causal chain models. Specifically, it includes: (1) collecting and collating neurobiological data related to addiction (neuronal firing, changes in neurotransmitter levels, strength of brain connectivity, etc.) and behavioral data (craving scores, behavior frequency, etc.); and (2) mapping these data to the five-layer elements of the DIKWP model. For example, electrophysiology and imaging data are used as "data (D) layers", which are extracted into "information (I) layers" (such as enhanced activity in specific brain regions) through patterns, and then associated with existing knowledge bases (known knowledge of addiction circuits) to form "knowledge (K) layers" representations, and value evaluation and trade-offs are given to form "Wisdom (W) layers" to make judgments, and finally correspond to the individual's motivational Purpose "P layer" (such as the strong purpose of continuing to seek medicine). (3) Based on this, the addiction semantic network model is constructed, with nodes representing the concepts of each layer of DIKWP and edges representing causality or correlation (e.g., "reduced hippocampal plasticity (K)→ reduced long-term return evaluation (W) → weakened abstinence determination (P)"); (4) The system dynamics method is used to quantitatively simulate the semantic network to analyze the key causal paths and its influence on behavior.

Technical route: We will combine knowledge graph and differential equation modeling. Firstly, the knowledge graph of addiction mechanism was constructed, and the biological mechanism and cognitive role reported in the literature were woven into a net. Then, causal inference algorithms (such as structural equation model or dynamic Bayesian network) are introduced to give the graph quantitative relationship, so as to deduce the overall effect from a certain node change. Semantic-driven simulation: Set different initial states of "purpose" (e.g., strong craving vs. no craving), observe the evolution of each layer of state in the model, and verify whether it reproduces the actual characteristics of addictive behavior. If necessary, reinforcement learning algorithms are introduced to allow AI agents to learn strategies on the model to better fit the trajectory of real behavior.

Innovation: For the first time, the artificial consciousness model is applied to the unified modeling of addiction mechanism, and the integrated simulation of bio-psychological processes is realized. Traditionally, neural circuit models and behavioral models have been in different domains, but we use DIKWP as a bridge to integrate the two into a complete chain from data to purpose. This not only facilitates understanding, but also provides a new means of validating causality: by manipulating nodes at a certain level in the model, it is possible to predict the impact of the corresponding intervention on the overall behavior. This lays the foundation for the design of follow-up intervention strategies.

Expected Results: 1 report on the semantic model of addiction mechanism (substance addiction and behavioral addiction) was formed, and related papers were published. The model will reveal at least two or more previously unidentified causal chains of addiction and validate them through animal experiments or big data analysis (e.g., the model predicts that the chain of "decreased prefrontal lobe function leads to weak willpower (P)", which can be supported by data from the prefrontal inhibition test of addicts).

Research Direction 2: DIKWP Semantic Modeling and Intelligent Reasoning for Withdrawal Drugs and Therapies

Main contents: To establish a semantic model of DIKWP for the existing major withdrawal interventions (including drugs and psychotherapy), and to analyze its mechanism of action, scope of application and causal chain of side effects. Typical cases: drugs such as methadone, buprenorphine, naltrexone; Psychological aspects such as Cognitive Behavioral Therapy (CBT), Mindfulness Meditation, etc. Specific steps: (1) collect mechanistic information (drug target receptors, physiological effects, therapeutic effects on cognition, etc.) and efficacy data (success rate, side effect rate, etc.) of each intervention method; (2) The intervention means were regarded as some intervention input to the semantic model of addiction, and the DIKWP model was used to model its process. For example, methadone replacement therapy can be represented as the input of a stimulus that activates μ-opioid receptors at the data layer, which in turn produces a judgment in the Wisdom layer that alleviates withdrawal pain, but may not change the motivation for addiction at the Purpose layer; CBT is modeled as the adjustment of the knowledge layer and the wisdom layer, that is, correcting the over-perception of drug benefits, enhancing the trade-off of withdrawal benefits, etc., so as to change the decision-making tendency of the Purpose layer. (3) The inference algorithm was used to deduce the limitations of intervention methods and the formation mechanism of side effects. For example, through model simulation, it was found that naltrexone blocking reward may also block the natural reward pathway, resulting in anhedonia. CBT requires a high degree of patient cooperation, and its effect depends on the initiative of the Purpose layer, and the efficacy of the model will be reduced if the motivation of the Purpose layer is insufficient.

Technical route: Construct a knowledge graph of withdrawal intervention, with nodes including drugs, targets, symptoms, cognitive processes, etc., and relationships including "acting on", "inducing", "alleviating" and "side effects". Combining logical reasoning and causal analysis: using rule-based reasoning (e.g., descriptive logic) to obtain the direct impact of the intervention, and then using the causal network approach to find out the chain of indirect effects. The key is to use the DIKWP model to clarify how the intervention acts on which layer of the five-layer structure, so as to locate its effects and shortcomings. For example, methadone acts on physiological D and I layers (reducing data input for withdrawal symptoms), but does not affect the memory of K layer and the value judgment of W layer, so the recurrence rate is still high; Conversely, a perfect intervention should work at the W and P levels (change values and wills). We formalized this concept as a criterion for evaluating each intervention.

Innovation: Traditionally, the evaluation of withdrawal drugs/therapies has mostly been based on empirical statistics and single-point mechanism analysis. This study introduces a new method of semantic causal analysis, and comprehensively examines the chain of action and blind spots of interventions through artificial consciousness models. This approach can explain why certain interventions are working/not working and predict how they can be improved. For example, it can explain why patients may still crave naltrexone even with naltrexone (because the Purpose layer is unaffected) and predict that "success may be increased if the combination of psychological interventions strengthens Purpose control". Such reasoning has not been shown in existing studies.

Expected Results: Construction of semantic knowledge base and model of withdrawal intervention; Write a semantic analysis report of the intervention mechanism and make recommendations for improvement for each therapy. At least one set of intelligent question-answering or decision support archetypes: input patient characteristics and optional interventions, and the system can suggest a recommended strategy based on semantic reasoning (e.g., for patients with high anxiety, model inference emphasizes mindfulness training to alleviate stress and help boost willingness). These results will be published/applied in the form of papers and patents, and will provide guidance for the subsequent design of new therapies.

Research Direction 3: Design and validation of semantic-driven novel intervention molecules and protocols

Main content: To explore new intervention strategies based on the identification of existing therapeutic limitations. This includes the design of new drug molecules and the development of new non-drug interventions, all guided by semantic goals. Steps: (1) According to the results of research direction 2, identify the unmet key therapeutic semantic needs, such as: "selectively inhibit craving memory reconsolidation", "enhance the regulation of impulses in the prefrontal lobe", "alleviate anxiety and depression in withdrawal period", etc.; (2) Determine the biological target or intervention pathway hypothesis for each need. For example, inhibition of memory reconsolidation could consider blocking specific signaling pathways between the hippocampus-amygdala; To enhance cognitive control, consider boosting prefrontal dopamine or utilizing transcranial stimulation; (3) New drug design: using the method of knowledge graph + model generation, the candidates who meet the conditions are screened and optimized in the huge compound space. Knowledge graphs provide target-pathway-action relationship constraints, and generative models (such as deep generative networks or reinforcement learning) produce molecules with desired properties. For example, in order to achieve the effect of "increasing prefrontal dopamine without overactivating the nucleus accumbens", we designed molecules that specifically excitate D1 receptors and moderately cross the blood-brain barrier; (4) Non-drug program design: propose a new cognitive training or stimulation program based on artificial consciousness simulation. For example, having addicts perform specific tasks in a VR environment to reshape reward estimates, or designing brain-computer interface closed-loop stimulation strategies that automatically give negative frontal feedback to inhibit impulses when a craving signal is detected. Virtual addiction agents are used to test the effects of these regimens and select the best ones for validation in animal experiments or clinical trials.

Technical route: For the design of new drug molecules, a cross-cutting approach is adopted: first, knowledge-driven screening (using known pharmacological knowledge to narrow the search scope, such as locking a certain class of receptor ligands), and then AI generative optimization (generating molecules that meet multi-parameter optimization through deep learning models). In particular, a semantic scoring function is introduced: unlike the traditional optimization of IC50 or ADMET alone, we will define a semantic score for the molecule, such as "both inhibiting craving without diminishing natural reward", which requires a synthesis of multi-target modes of action and calculation using knowledge graph reasoning. During the optimization process, the AI tries different structures and evaluates them with semantic scores until the polymer is found. For the design of the new scheme, human-computer confrontation simulation was adopted: the artificial consciousness addiction agent interacted with the intervention AI, and the strategy was continuously improved in the simulated environment. This is similar to adversary learning in reinforcement learning, allowing the intervention AI to learn the optimal strategy. The whole route takes semantic goals as the core evaluation criteria.

Innovation: The traditional research and development of new drugs takes a long time and has no clear direction, but the "semantic-driven drug design" proposed in this study deduces drug characteristics from the semantic requirements of efficacy, which significantly narrows the search space and improves the pertinence. At the same time, we use artificial consciousness models for intervention strategy inventions, which break through the limitations of human experience and may generate unexpected new solutions. For example, an AI may find that a unique training sequence or stimulus waveform is more effective than artificially assumed. This idea of "creating" an intervention method by AI is the first of its kind in the field of withdrawal and is a disruptive innovation.

Expected Results: Acquisition of at least 1-2 promising new compounds (patent application completed) and 1 novel intervention program (e.g., VR+BCI training program, application for software copyright or patent). Small-scale animal trials verified that the new molecule showed significantly better effects than existing drugs in animal models of addiction (statistically significant p<0.05), or that the new behavioral intervention program showed better adherence and preliminary efficacy in volunteers. Publish high-level papers to explain the design methods and experimental results, and lay the foundation for further clinical research.

Research Direction 4: DIKWP Semantic Space Modeling and Intervention Simulation of Game and Gambling Addiction

Main contents: The DIKWP artificial consciousness model was applied to the field of behavioral addiction, and the semantic space model of online game addiction and gambling addiction was constructed, and artificial intelligence simulation experiments were developed to test the effect of different intervention strategies at the semantic level. Specific work: (1) Construct cognitive-affective semantic models of game addiction and gambling addiction. Relevant research data were collected, such as abnormal brain function areas of game addicts, types of cognitive biases (such as probability distortions) of gambling addicts, etc., and expressed as various layers of DIKWP elements: the data layer included sensory stimuli (game rewards, gambling stimuli) and physiological responses (dopamine release, etc.) in addiction situations; The information layer is the extracted features (win-loss feedback frequency, reward uncertainty, etc.); The knowledge layer includes the individual's understanding of rules and probabilities, and the memory of past experience. The Wisdom layer involves its assessment of risk and benefit, as well as emotional responses (e.g., excitement or frustration); The purpose layer is the motivation to continuously engage in behavior (escapism or pleasure pursuit); (2) Build an artificial consciousness simulation environment: make an artificial intelligence agent act in a simulated game or gambling environment with the DIKWP model as the "psychological framework". Agents have addiction-like preferences and limitations (by adjusting their knowledge/wisdom layer parameters such as high discount rate, low loss aversion, etc.). Engaging agents in simulated gaming/gambling tasks that create addictive behavioural traits (e.g., a cycle of losing and gambling); (3) Introduce different intervention strategy simulations, such as setting betting limits (external constraints, equivalent to limiting the intensity of stimuli at the data layer), increasing risk reminders (providing real probability information at the knowledge layer), and introducing a delay reward mechanism (changing the evaluation of instant rewards at the Wisdom layer). Observe the agent's behavior and internal state changes, such as the number of bets, stop probability, sentiment indicators, etc.; (4) Analyze the links of each intervention that work within the model and compare their effectiveness. If a strategy is found to significantly reduce the purpose of continuing to gamble on the P layer, it is considered effective.

Technical route: Agent-based modeling and simulation method is adopted. Develop a simulation environment (which can be reduced to a classic Iowa gambling task or reinforcement learning environment) to engage agents with DIKWP decision architecture. The agent's decision-making is determined by the internal five layers: we will design brain-like sub-modules for it (e.g., the "reward evaluation module" corresponds to the Wisdom layer, and the "impulse control module" corresponds to the Purpose layer regulation). Trans-brain causal models were introduced into the simulation, such as setting "when the reward appears, the surrogate nucleus accumbens (simulated reward) activation is enhanced, while the prefrontal control is weakened" to realistically reproduce the addiction decision. Different interventions are achieved by changing the parameters or inputs of these modules. Run a large number of simulation statistics and apply reinforcement learning algorithms to find the best combination of interventions. Finally, the simulation conclusions are validated with the necessary human data, such as comparing them with experimental results in the literature.

Innovation: This is the first time that the artificial consciousness model has been used in an in vivo simulation experiment for behavioral addiction intervention. Traditional psychological research can only be done through trial and error in real people, but this task establishes a "digital testing ground". The combination of human-like cognitive biases and machine-observable internal states allows us to open the "black box" to see how interventions affect the agent's intrinsic semantic layer changes (e.g., how prompt information changes their knowledge-level cognition). This provides unique causal insights and predictive capabilities. Especially for interventions with ethical or practical limitations, we can first evaluate them in simulation. For example, the effects of coercive measures on agents in extreme situations can be simulated without the need for real human experimentation. This innovative approach is expected to become a new paradigm in behavioral addiction research.

Expected Results: Completed the development of 1 set of software of "Artificial Awareness Simulation Platform for Behavioral Addiction"; Simulation models and intervention effect reports for at least two addictive behaviors were given. The report will list the magnitude of the impact of various strategies on the agent behavior indicators and recommend the best solution. For example, we may find that the combination of "limiting game time + immediate feedback of real loss" is more effective in correcting game addiction behaviors, which can inform policymaking. The results were published in the form of academic papers and conference reports, which enhanced the influence in the field of behavioral addiction modeling in China.

Research Direction 5: Construction of Addiction Intervention Mechanism Platform of Artificial Consciousness and Human-Computer Collaboration

Main contents: Based on the above research results, a human-computer intervention platform driven by artificial consciousness is developed to realize the human-machine collaboration of multiple intervention methods. The platform includes: data collection layer (physiological and behavioral data obtained by neuroimaging, EEG, and wearable sensing), semantic analysis layer (DIKWP artificial consciousness operating system, multi-level semantic analysis of data), decision-making layer (giving personalized intervention decisions according to semantic causal models), and execution layer (acting on patients through APP, VR, brain stimulation equipment, etc., and collecting feedback). The platform will integrate specific functional modules such as: neuroimaging + AI monitoring module (using fMRI/EEG to detect brain craving-related patterns, which are interpreted into semantic signals by AI in real time), brain-computer interface feedback module (when high craving states are detected, automatically triggering such as transcranial electrical stimulation to reduce excitability, or reminding patients to perform mindfulness exercises through visual/auditory signals), and digital cognitive training module(gamified tasks are provided for patients to improve cognitive control, and AI adjusts the difficulty of tasks based on patient completion), therapist collaboration interface (AI analysis and recommendations are presented to clinicians in an interpretable form, such as annotating the current stage of addiction and recommended intervention strategies). Through the collaboration of these modules, a closed-loop intervention is established: the cycle of monitoring-analysis-intervention-re-monitoring is similar to the sensing-planning-execution closed-loop in autonomous driving, but applied to withdrawal support.

Technical roadmap: The platform development will follow a modular architecture. Firstly, the Artificial Consciousness Operating System Kernel (ACOS) is designed, which is embedded in the DIKWP model, which is responsible for coordinating the data and semantic flow of each module. ACOS converts the data from the sensor into DIKWP content (corresponding to the perception D/I layer) and runs semantic programs (such as the models and strategies in tasks 1-4 above) to derive the Wisdom decision W and Purpose P layer outputs. To achieve this, we will use the relevant solution in Professor Yucong Duan's patent to break down the AI decision-making process into five steps that can be monitored. The modules are connected via a semantic communication interface, i.e. the information exchanged is tagged with DIKWP. For example, the EEG module outputs not only waveforms, but events labeled as "high thirst states (layer I)"; The cognitive training module does not require simple instructions, but the goal of "improving willpower (P-level)", which is translated by ACOS into specific task parameters. This semantic interface greatly improves the efficiency and intelligence of the collaboration of all parts. In terms of implementation, a microservice architecture will be adopted, with each functional component developed independently and communicated through a semantic bus.

In terms of test evaluation, we will first conduct a simulation test: use the artificial consciousness agent of task 4 to simulate the patient, interact with the platform, and check the correctness and safety of the platform's decision-making logic. Then conduct a small-scale clinical trial: invite addiction survivors to participate, let the platform assist the withdrawal process for 1-3 months, and observe their experience and effect. In the process, the human-computer interaction interface and algorithm parameters are continuously improved.

Innovation: The platform is a groundbreaking integration: for the first time, artificial consciousness (explainable AI), brain-computer interfaces, semantic communication, and digital therapeutics are fused together to serve addiction intervention. Compared with traditional treatment, it is more real-time (real-time monitoring of brain status), **more personalized (**AI tailors strategies for each patient), and better synergy (human-machine participation in treatment decision-making). At the same time, unlike general AI systems, every decision-making step within the platform is transparent and traceable, and therapists and patients can understand what the AI did and why, thereby increasing trust and compliance. This kind of artificial consciousness-driven explainable intelligent intervention is a cutting-edge exploration in the medical field and has important demonstration significance.

Expected Results: Completed the prototype of the "Artificial Awareness Platform for Addiction Intervention" and passed the test; Apply for corresponding patents or software copyrights for platform software and hardware; Formation of platform usage guidelines and clinical evaluation reports. It is expected that the platform will significantly help patients improve some indicators, such as reducing subjective craving scores, prolonging the number of days of uninterrupted abstinence, etc., in the trial, with a statistically significant difference compared with the control group without using the platform. The project will summarize the experience of this innovative model and propose further large-scale clinical applications. Ultimately, we hope that the platform will lay the foundation for subsequent industrialization and become a new prototype of intelligent addiction rehabilitation equipment/systems in China.

Project implementation plan and schedule

In order to ensure the realization of the project objectives, we formulate scientific and reasonable implementation plans and schedules. The project period is planned to be 5 years, which is divided into three stages: basic research stage (1-2 years), integrated research stage (3-4 years) and experimental verification stage (5 years). The main tasks and milestones of each phase are as follows:

• Year 1-2: Basic Research Stage. Focus on research directions 1 and 2 to consolidate the theoretical and model foundation. In the first year, the initial construction and validation of the addiction semantic model was completed (Task 1); In the second year, the semantic analysis of the withdrawal intervention was completed and the limitations of the existing therapy were reported (Task 2). Milestones: Formation of the semantic model of addiction mechanisms version 1.0, covering major loops and behavioral causal chains, construction of a knowledge graph of withdrawal intervention, and identification of at least 3 improvement needs. Publish 2-3 high-level papers to consolidate the foundation of follow-up research of the project.

• Year 3-4: Integration stage. Combined with the results of basic research, we will promote research directions 3, 4 and some directions 5. Year 3: Redesign and Simulation (Tasks 3 and 4): Complete the screening of new drug candidates and carry out animal experiment validation; Develop an artificial awareness simulation platform for behavioral addiction and generate recommendations for intervention strategies. In the fourth year, we entered the integrated development of the platform (task 5): realized the parallel debugging of the functions of each module, and preliminarily completed the prototype of the artificial consciousness intervention platform. Milestone: Obtain animal experimental data of at least one effective new compound, and report of behavioral intervention simulation results; The prototype of the platform was successfully run in a closed-loop environment in a simulated environment. He has applied for 2-3 invention patents and published several papers.

• Year 5: Pilot Verification Phase. Small-scale hands-on trial validation and optimization of the platform. Select partner hospitals/drug rehabilitation centers, invite volunteers from addiction patients to try out the platform (with ethical and safe permissions), and collect feedback to improve the system. Complete the acceptance of the overall technical indicators of the project. Milestones: Provide platform trial data to prove that its functionality is effective and secure; Complete the project summary report and technical specifications to meet the expected indicator requirements.

During the whole implementation process, the project implements a sub-project responsibility system and a regular evaluation mechanism. Each research direction is set up with sub-topics, led by experts in the corresponding field, and the outputs and time nodes are clarified. The Project Management Office checks the progress on a quarterly basis, conducts a stage review every year, and adjusts resources and technical routes according to expert opinions to ensure that tasks are completed on time and with high quality.

The project team and the existing work base

The project is undertaken by a highly qualified interdisciplinary team. The core members of the team include experts in the field of artificial intelligence and artificial consciousness, neuroscience and addiction medicine, software engineering and system integrators, etc., with the knowledge and experience in multiple fields required to achieve the project goals. The project leader, Professor Duan Moumou (hypothetically the core of Yucong Duan's team), has profound attainments in the field of artificial consciousness DIKWP model, has published more than 260 related papers, and has been authorized 85 invention patents, and his DIKWP theory has laid a methodological foundation for the project. Professor Li, the co-leader, is a neurobiologist who has been engaged in the study of the brain mechanism of addiction for a long time, and has made rich achievements in dopamine circuits and animal models of addictive behavior. The team also includes Associate Professor Yang Moumou of the School of Computer Science (expert in artificial intelligence and knowledge graph), Professor Wang Moumou of the School of Psychology (cognitive intervention expert), Chief Physician Zhang Moumou of the Affiliated Hospital (clinical expert of addiction), etc., many of whom have presided over national scientific research projects and have good cooperation experience.

Existing research foundation: In recent years, the team has carried out preliminary exploration and achieved results in the relevant directions of the project, which provides a solid foundation for the implementation of the project

• In terms of artificial consciousness model, the project team members systematically studied the theory and application of the DIKWP model. Professor Duan's team built a DIKWP physiological and artificial consciousness prototype system, which mapped the natural language interaction and subjective cognitive process in the doctor-patient consultation scene into the DIKWP model, and realized the unified computing simulation of data, information, knowledge, wisdom and purpose in the dialogue. The DIKWP brain region mapping theory proposed by them clarifies the corresponding brain regions and roles of the five levels in the cognitive process, and uses fusion transformation technology to solve the problem of inconsistent and imprecise information in actual interaction. The prototype system successfully visualized the DIKWP semantic graph and demonstrated the interpretability advantage of the artificial consciousness model in complex cognitive tasks. These results demonstrate our ability to apply the DIKWP model to the more complex scenario of addiction, with a preliminary software and algorithmic foundation.

• In terms of the study of the neural mechanism of addiction, the laboratory of Professor Li, a member of the team, is the first in China to use optogenetics to describe the circuit mechanism of morphine addiction, revealing the dual effects of the VTA-IPAC pathway on reward and memory. The results were published in journals such as Addiction Biology, indicating that the team has experience in exploring microscopic circuits. In addition, we maintain cooperation with a number of addiction research centers in China, and have access to valuable experimental data and biological samples to support model construction and validation.

• In terms of knowledge graph and semantic reasoning, Associate Professor Yang's team has built medical knowledge graphs, studied the application of knowledge graphs in drug discovery, and participated in the formulation of the IEEE international standard for financial knowledge graphs. They are familiar with biopharmaceutical ontology and knowledge mining, and have published papers in the direction of knowledge graph-driven drug repositioning. This will directly contribute to the construction of the knowledge graph of withdrawal intervention and the development of semantic reasoning algorithms in the project.

• In terms of brain-computer interface and digital intervention, Professor Wang's team has experience in developing VR and mobile health applications for psychological interventions, and Director Zhang's clinical institution has carried out transcranial magnetic stimulation to treat drug addiction, accumulating experience in patient recruitment and clinical evaluation. They also participate in the cognitive neurofeedback BCI project of USTC and have a good understanding of its technical details. These provide a practical basis for platform integration and trial implementation of the project.

The preliminary research of each member of the team provides a rich data, method reserve and validation environment for this project. For example, we have a large number of behavioral and brain recordings of animal models of addiction that can be used for model training; There is a preliminary DIKWP programming framework that can be used for rapid prototyping; Ethics approvals and patient cohorts from partner hospitals are in place to prepare for pilot applications. A good foundation increases the reliability of the project's success.

In addition, the project relies on the fully equipped experimental and calculation conditions of the unit. We have a neuroimaging center (3T MRI, optogenetic imaging device), a brain-computer interface laboratory (high-density EEG, transcranial electrogram/magnetic stimulator), a high-performance computing cluster and a secure data storage environment, which provides the necessary conditions for the implementation of the project.

Expected results, form of results and assessment indicators

The project is expected to produce a series of important results in both scientific research and technology development, including:

1. Theoretical and methodological achievements: The construction of semantic models of addiction neural mechanisms and intervention semantic reasoning methods have enriched and developed the application of artificial consciousness theory in the field of cognitive neuroscience. It is expected to publish more than 10 SCI papers (including no less than 3 journals in the first district of the Chinese Academy of Sciences) and several Chinese core journal papers, covering addiction semantic models, semantic-driven drug design, new intervention strategy simulation, human-computer collaboration platform, etc. This project will form a monograph or 1 chapter, systematically summarize the experience of the application of the DIKWP artificial consciousness model in brain science, and provide reference for the academic community.

2. Core technology and software: Develop a prototype system of "Artificial Awareness Intervention Platform for Addiction", including related software and hardware interfaces. In terms of software, one set of DIKWP semantic simulation platform (including addiction behavior simulation module and intervention strategy optimization module) and one intervention decision support system were delivered, with visual interface and interactive functions. In terms of hardware, it integrates 1 set of portable EEG acquisition and stimulation linkage equipment. It is expected to apply for more than 5 invention patents (including new compounds, new methods, new systems, etc.) and more than 3 software copyrights. The technical indicators are satisfied: the real-time monitoring accuracy of the platform is ≥ 70%, the intervention response delay is <1 second, and the coverage rate of explainable rules is 100%. We will also formulate draft technical standards or specifications, such as "Data Interface Specification for Artificial Consciousness-Assisted Withdrawal Intervention System" and "Semantic-Driven Cognitive Training Process Specification", to lay the foundation for subsequent industry promotion.

3. Experimental data and empirical evidence: A large number of valuable experimental data are generated, including the database of addictive behavior and brain activity, and the database of intervention simulation results. Specifically, build a database containing multimodal data of at least 100 addicts (or animal experiments), and open and share part of it with scientific research peers; Complete preliminary data on efficacy and safety of new compounds; A small sample clinical trial report was formed, and the indicators such as the reduction in recurrence rate and the reduction in thirst score under the assistance of the platform were quantified, and it was expected to increase by at least 20% compared with the control. Through these data, the effectiveness of the proposed method of the project is demonstrated. For example, in a trial of about 10 people, the average number of consecutive days of abstinence among platform users was significantly longer than the historical baseline (assuming an increase from 20 days to more than 30 days).

4. Talent training and team development: The program will cultivate a group of interdisciplinary talents, including 10-15 postdoctoral fellows, doctoral students and engineers. These people will master new ways to combine artificial intelligence and neuroscience, and become the backbone of related fields in the future. It is expected that during the implementation of the project, 2-3 young backbones will grow into deputy senior professional titles, and the influence of the team in the intersection of artificial consciousness and brain science in China will be significantly improved. The project will also consolidate industry-university-research partnerships, such as establishing links with drug rehabilitation institutions, hospitals, and AI companies, to create conditions for the transformation of results.

5. Socio-economic impact: In the long run, the project will have a positive impact on both the scientific and social levels. Scientifically, it has opened up a new direction for the application of artificial consciousness models to the research of brain diseases, and enhanced China's international status in the field of brain science and brain-like intelligence. In society, if the project technology is further matured, it can be applied to drug rehabilitation centers, rehabilitation centers and community addiction intervention, improve the success rate of withdrawal, reduce the harm of addiction to public health, and have great social welfare value. Economically, the intelligent withdrawal intervention system is expected to form an emerging industry, and related products and services can be promoted to the mental health market, generating potential economic benefits. During the implementation of the project, attention will also be paid to the patent layout and follow-up incubation, so as to lay a good foundation for the transformation of achievements.

In summary, the expected results of this project are rich and innovative, and its completion will mark an important original breakthrough in the field of addiction research and promote the application of artificial intelligence technology in medical and health scenarios. We will evaluate the progress of the project in strict accordance with the above indicators to ensure that the output of each task meets or exceeds expectations, so as to ensure the successful conclusion of the project and the sustainable development of the results.

Primary references

1. Everitt, B. J., & Robbins, T. W. (2016). **Drug addiction: updating actions to habits.**Annual Review of Psychology, 67, 23-50.

2. Volkow, N. D., Koob, G. F., & McLellan, A. T. (2016). **Neurobiologic advances from the brain disease model of addiction.**New England Journal of Medicine, 374(4), 363-371.

3. Fiore, V. G., et al. (2023). **Computational models of behavioral addictions: state of the art and future directions.**Addictive Behaviors, 134, 107441.

4. Duan, Y., et al. (2025). DIKWP Semantic-Driven Programming for Artificial Consciousness and Case StudyResearch Report.

5. Yucong Duan . (2025). DIKWP Artificial Consciousness Model Research ReportResearch Report

6. Xin Wenjun et al. (2023). Neural Circuit Mechanisms of Morphine Addiction Reward and Environmental MemorySun Yat-sen University Medical Center News

7. Bu, J., et al. (2021). **A Novel Cognition-Guided Neurofeedback BCI Dataset on Nicotine Addiction.**Frontiers in Neuroscience, 15, 647844.

8. James, T., & Hennig, H. (2024). **Knowledge Graphs and Their Applications in Drug Discovery.**Methods in Molecular Biology, 2716, 203-221.

9. Wu Jian et al. (2021). Reward Processing Deficits and Their Neural Mechanisms in Online Game AddictionAdvances in Biochemistry and Biophysics

10. Daniel Zhang et al. (2022). Application and prospect of digital therapeutics in substance addictionChinese Journal of Psychiatry, 55(4), 300-305