A psychological counseling course intelligent teaching pushing method and system

Abstract

The application discloses a kind of psychological counseling course intelligent teaching push method and system, comprising: obtaining original data is disassembled into course node and constructs multidimensional knowledge graph;Learning data is collected, and semantic model generates cognitive impairment error vector and scene efficiency coefficient;According to data, reliability is calculated and input update formula outputs mastery probability and mismatch item;Based on mastery probability and mismatch item, generate and push learning path.The present application effectively alleviates the problem of unclear course push and cognitive overload, significantly improves the accuracy and explainability of knowledge remediation.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence education, specifically to an intelligent teaching delivery method and system for psychological counseling courses. Background Technology

[0002] In higher education psychology programs, there is a clear foundation-application connection between courses such as psychological counseling techniques, psychological assessment, and general psychology. Traditional teaching models, which uniformly deliver course content chapter by chapter, struggle to identify whether students are experiencing difficulties in conceptual memorization, emotion recognition, technology adaptation, or ethical risk assessment. Existing course knowledge graphs do not incorporate the emotional load and ethical risks unique to psychological counseling courses as node attributes, failing to explain cognitive overload and technology misuse observed in high-emotional case questions. Furthermore, the mastery of probability models does not utilize factors such as answer time, accuracy, prompt response time, and semantic error type, making it difficult to distinguish between quick guesses and conceptual deviations. Systems often recommend adjacent knowledge points, failing to provide comprehensive remedial support from foundational courses. Moreover, they primarily cater to individual students, making it difficult to aggregate common learning difficulties across classes to generate specialized content. AI semantic recognition often remains at the evaluation level, without being written back into the knowledge graph. Therefore, a smart teaching system specifically tailored to the characteristics of psychological counseling courses is urgently needed. Summary of the Invention

[0003] This invention proposes an intelligent teaching delivery method for psychological counseling courses, comprising: S1. Obtain the original data and break it down into multiple course nodes to construct a multi-dimensional knowledge graph that includes the quantitative value of emotional load and the level of ethical risk. S2. Collect learning data and scenario factors of target course nodes in the multidimensional knowledge graph, input open-ended answers in the learning data into the semantic model, output cognitive impairment error vector, and fuse the scenario factors into scenario efficiency coefficient; S3. Calculate the reliability of the learning data, input the reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level into the mastery probability update formula, calculate the emotional load mismatch based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, and update the output individual mastery probability based on the emotional load mismatch. S4. Based on the mismatch between the individual's mastery probability and the emotional load, a learning path is generated and pushed to the terminal.

[0004] The multidimensional knowledge graph is constructed in the following manner: Obtain the original materials of the psychological counseling course and break down the course content into concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes; The system receives multiple sets of scoring samples from instructors for concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes. The system calculates the arithmetic mean of the multiple sets of scoring samples and maps them to a preset normalization interval through linear normalization to obtain the initial difficulty, cognitive load requirement, emotional load quantification value, and ethical risk level of each node. Configure directed edges and edge weights between nodes. The directed edges include at least cognitive prerequisite edges, basic support edges, skill transfer edges, emotional load progression edges, case context matching edges, ethical risk constraint edges, easily confused edges, and remedial jump edges. Output the multidimensional knowledge graph composed of the various types of nodes and the directed edges.

[0005] The cognitive impairment error vector is obtained as follows: The open-ended answer text, the standard answer to the knowledge point, and the standard case scenario are input into the semantic model, and the text semantic feature vector is output. The text semantic feature vector is compared with the standard reference answer vector by cosine similarity, and multidimensional error components are calculated and normalized. The multidimensional error components include conceptual confusion error score, visitor emotion recognition deviation score, counseling technology selection and adaptation error score, ethical boundary judgment error score, basic theory missing error score, and vague expression error score. The multidimensional error components are assigned a set norm and aggregated to generate the cognitive impairment error vector.

[0006] The reliability is calculated as follows: Obtain the standard reference time of the target course nodes marked in batches by psychology teaching experts, statistically analyze the distribution of normal thinking and answering time of previous students, and take the lower limit quantile of normal answers as the preset proportion threshold. Determine whether the time taken to answer the question is lower than the product of the preset percentage threshold and the standard reference time, and whether the accuracy rate of the answer is correct: If so, then calculate the quick guess penalty item within the set penalty range based on the degree of deviation in time consumption; Substitute the fast guessing penalty term into the reliability calculation formula, truncate the lower limit of the constraint reliability to a preset lower limit value, and output the reliability.

[0007] The mismatch between the individual mastery probability and the emotional load is calculated as follows: Substitute the correct answer rate, the reliability, the learning behavior data, the cognitive impairment error vector, and the normalized value of the prompt pop-up response time into the evidence calculation formula, and output a learning evidence after normalization by the activation function; The student's pre-class brief emotional scale score and cumulative continuous learning time for the day are obtained. The student's current cognitive capacity is calculated, and the cognitive capacity, the quantitative value of emotional load, and the ethical risk level are subjected to uniform linear scaling normalization. The normalized cognitive capacity, the quantified emotional load, the ethical risk level, and the scenario efficiency coefficient are input into the mismatch calculation formula, and the emotional load mismatch is output. The learning evidence, the emotional load mismatch, the weight of prior knowledge influence, and the weight of the prior edges of the multidimensional knowledge graph are input into the mastery probability update formula, and the individual mastery probability is output.

[0008] The learning path is generated and pushed in the following manner: Calculate the prior comprehensive mastery level of the target course node. If the prior comprehensive mastery level is lower than a preset threshold, intercept the target course node and backtrack along the basic support edge to generate a basic course remedial path. When the individual's mastery probability is lower than the mastery threshold and the error score for missing basic theory or the error score for conceptual confusion in the cognitive impairment error vector is greater than the corresponding threshold, the graph distance from the basic course node to the target course node, the cognitive load of the basic course node, and the semantic similarity between the open-ended answer text and the basic course node are obtained. A comprehensive score for the remedial path is calculated, and a set of candidate remedial paths is generated by sorting them in descending order according to the comprehensive score of the remedial path. The individual mastery probability, the emotional load mismatch, the candidate remedial path set, and the scenario efficiency coefficient are substituted into the push priority calculation formula for weighted sorting, and the result is output and pushed to the student terminal.

[0009] It also includes: S5, generating and pushing group-specific content; Specifically, it includes: Using classes as units, the individual mastery probability, cognitive impairment error vector, abnormal answering time, and knowledge base query behavior of students at each course node are summarized to construct a group cognitive impairment matrix. The error rate of knowledge points in a class is calculated by dividing the number of students who made mistakes at the target course node by the total number of students in the class. The mean of semantic errors is obtained by calculating the mean value of the semantic vector setting error between the students' answers and the standard answers. The variance of the ratio of time spent on multiple answers is calculated and the upper limit is truncated to obtain the dispersion of answer time. The persistent knowledge query gap is obtained by combining the number of queries and the weighted scoring of the mastery increase. The group barrier strength is output by linearly normalizing the class knowledge point error rate, the mean semantic error, the answer time dispersion and the stubborn knowledge query gap, and then weighting and summing them. Students with similar cognitive impairment error vector patterns are clustered, and class-specific content is generated based on the group impairment intensity. This class-specific content is then pushed to the corresponding student groups.

[0010] It also includes: S6, adaptive knowledge base expansion; Specifically, it includes: A built-in psychology expertise base is included, which is linked with the semantic model and class group obstacle clustering. When a term or misconception is repeatedly searched by multiple students, repeatedly answered incorrectly, or frequently appears in open-ended answers, candidate knowledge entries, candidate case comparisons, and candidate test questions are automatically generated. Upon receiving the teacher's review instruction, the candidate knowledge items, the candidate case comparisons, and the candidate test questions are added to the psychology professional knowledge base and question bank, and are solidified into the multidimensional knowledge graph as misconception nodes or easily confused edges.

[0011] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent teaching and push method for psychological counseling courses.

[0012] A smart teaching and delivery system for psychological counseling courses includes: The graph construction module is used to acquire the original data of the psychological counseling course, decompose the original data of the psychological counseling course into multiple course nodes containing the emotional load quantification value and the ethical risk level, and construct a multi-dimensional knowledge graph composed of the multiple course nodes and directed edges. The data acquisition and semantic recognition module is used to collect students' learning data and scenario factors at the target course nodes in the multidimensional knowledge graph, input open-ended answer text in the learning data into the semantic model, output cognitive impairment error vector, and fuse the scenario factors into scenario efficiency coefficient. The mastery probability calculation module is used to calculate the reliability based on the answering time in the learning data and the standard reference time. The reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level are input into the mastery probability update formula. Based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, the emotional load mismatch is calculated, and the individual mastery probability is updated and output based on the emotional load mismatch. The path push module is used to generate and push learning paths to student terminals based on the mismatch between the individual's mastery probability and the emotional load.

[0013] This invention proposes an intelligent teaching recommendation method for psychological counseling courses. It constructs a dedicated graph that includes emotional load and ethical risk dimensions and integrates multidimensional behavioral data and semantic error vectors to calculate the mastery probability. This effectively alleviates the problems of unclear recommendation basis and cognitive overload in psychological counseling courses and significantly improves the accuracy and interpretability of knowledge remediation and group feedback. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an intelligent teaching and delivery method for psychological counseling courses proposed in this invention. Detailed Implementation

[0015] refer to Figure 1 This invention proposes an intelligent teaching recommendation method for psychological counseling courses. Its core innovation lies in introducing the unique emotional load and ethical risk dimensions of psychological counseling courses into the node attributes of a knowledge graph. It also constructs an emotional load mismatch item as an independent penalty dimension for updating the mastery probability. This allows the system to not only assess students' mastery based on their answer performance but also simultaneously consider their ability to withstand emotional stimuli and ethical risks within their current cognitive capacity. This results in interpretable recommendation criteria and adaptive paths when generating learning paths. Unlike existing technologies that assess mastery solely based on answer accuracy and recommend only adjacent knowledge points, this invention uses the topological structure of a multi-dimensional knowledge graph to backtrack basic course nodes. It transforms open-ended answers into cognitive impairment error vectors using a semantic model, quantifies environmental interference using a scenario efficiency coefficient, and jointly drives Bayesian updates using the mastery probability update formula. Finally, it outputs the individual's mastery probability and generates a comprehensive remedial path, forming a complete closed loop from individual diagnosis to group feedback and then to the self-evolution of the knowledge base.

[0016] The aforementioned intelligent teaching delivery method for psychological counseling courses specifically includes: S1. Obtain the original data and break it down into multiple course nodes to construct a multi-dimensional knowledge graph that includes the quantitative value of emotional load and the level of ethical risk.

[0017] The original data refers to basic and applied teaching materials covering courses such as basic psychological counseling techniques and psychological assessment; the emotional load quantification value is a normalized value that characterizes the intensity of emotional stimulation brought about by traumatic or negative scenarios in the case; and the ethical risk level is a normalized value that characterizes the degree of privacy or crisis intervention risk involved in the case scenario.

[0018] Specifically, by reconstructing unstructured teaching materials into graph-structured data, implicit psychological characteristics are transformed into explicit graph-structured data, providing structured input for subsequent calculations.

[0019] Based on course materials on psychological counseling techniques, psychological assessment, general psychology, developmental psychology, abnormal psychology, and psychological measurement, the course content is broken down into concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes.

[0020] Each knowledge point node At a minimum, it should include: knowledge point ID, name, course, knowledge type, course segment, set of test questions, and initial difficulty. Cognitive load demand Emotional burden Ethical risk level Applicable Consultation Stage Corresponding emotional labels of visitors Pre-node set Basic Course Remedial Nodes and easily confused node set .

[0021] for The four types of continuous quantitative attributes are uniformly mapped to the interval [0, 1] by taking the arithmetic mean of expert scores and linearly normalizing them.

[0022] The initial difficulty is calculated using a linear normalization formula. :

[0023] Teacher input Experts rated the difficulty. ,in, represent The average of the expert difficulty scores given by the aforementioned sample. represent The minimum value in, represent The maximum value in.

[0024] Cognitive load demand Define the amount of rational cognitive resources required for students to understand and deduce knowledge points, and calculate the cognitive load requirement using a linear normalization formula. :

[0025] Teacher entry Group cognitive load scoring ;in, represent The average value of the cognitive load score sample. represent The minimum value in, represent The maximum value in.

[0026] Emotional burden Define the intensity of emotional stimuli from traumatic and negative scenarios in the accompanying case studies, and calculate the emotional load using a linear normalization formula. :

[0027] Teachers assigned a workload score to each case study related to the knowledge points. for The average load score for The minimum value in the load score, for The maximum value in the load score.

[0028] Ethical Risk Level Define the degree of risk involved in cases and consultation scenarios concerning privacy, crisis intervention, and ethical conflicts, and categorize them into ethical risk levels. The calculation is performed using a linear normalization formula:

[0029] Teachers assign risk scores based on the scenario. The sample mean. for The minimum value in, for The maximum value in.

[0030] Applicable Consultation Stage Then a 5-dimensional 0-1 binary vector is used, without the need for mean and normalization processing, and the matching degree of knowledge points is calculated by the cosine similarity of the two vectors.

[0031] Visitor's emotional labels Using multi-label 0-1 encoding vectors, the bias in student response emotion recognition is quantified by vector Euclidean distance.

[0032] Pre-node set Store the set of upstream knowledge points that must be mastered before learning the current knowledge point.

[0033] The specific formula for calculating the pre-requisite comprehensive mastery is as follows:

[0034] in, This represents the level of comprehensive understanding gained from the prior knowledge. This represents the total number of nodes in the aforementioned set of preceding nodes. This represents the traversed nodes in the aforementioned set of preceding nodes. On behalf of students In the preceding node The probability of an individual mastering something.

[0035] like If the value is below the system's preset threshold, it will be directly blocked. Push notification.

[0036] Basic Course Remedial Node Set Storage support The formula for calculating the score for remedial needs of single basic nodes in the interdisciplinary basic theoretical nodes of understanding is as follows:

[0037] in, This represents the score for the remedial needs. This represents the average similarity between the student's answer text and the standard text of the base node. This represents the average number of hops for the target knowledge point and the basic node in the graph. On behalf of students At the target node The current individual possesses the probability. The error score represents the missing fundamental theory in the semantic recognition output.

[0038] Set of Easily Confused Nodes Storage and For similar knowledge points with blurred conceptual boundaries, the system automatically calculates... The average text similarity with each obfuscated node; the higher the similarity, the greater the penalty weight.

[0039] All node quantification parameters are uniformly output to three major calculation modules: mastery probability update, push priority scoring, and remedial path generation.

[0040] Configure directed edges and their weights between nodes. These directed edges include at least: cognitive prerequisite edges, foundational support edges, skill transfer edges, emotional load progression edges, case scenario matching edges, ethical risk constraint edges, easily confused edges, and remedial jump edges. Output the multidimensional knowledge graph composed of these types of nodes and the directed edges, with all edge weights specified. It belongs to [0, 1] and is obtained by taking the average of the scores from multiple experts and then linearly normalizing it.

[0041] The emotional load progression edge is used to constrain the continuous push of high-emotional cases when cognitive capacity is insufficient; the ethical risk constraint edge is used to limit high-risk cases before the boundaries of crisis intervention are mastered; and the basic support edge is used to review basic course nodes such as general psychology.

[0042] After teachers mark the degree of concept confusion and normalize the weights, if a student answers with confusion about two points, the concept confusion error score will increase; if more than 30% of students in the class are confused and the score is greater than 0.5, a concept confusion topic will be automatically generated and pushed out.

[0043] Furthermore, the original data of the psychological counseling course is obtained, and the course content is broken down into concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes. Multiple sets of scoring samples from the teaching experts are received for each of the concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes. The arithmetic mean of these multiple sets of scoring samples is calculated and mapped to a preset normalization interval using linear normalization to obtain the initial difficulty, cognitive load requirement, emotional load quantification value, and ethical risk level of each node. Directed edges and edge weights are configured between nodes. These directed edges include at least cognitive prerequisite edges, basic support edges, skill transfer edges, emotional load progression edges, case scenario matching edges, ethical risk constraint edges, easily confused edges, and remedial jump edges. The multidimensional knowledge graph composed of these types of nodes and the directed edges is then output.

[0044] The directed edges are topological relationship edges that connect graph nodes and are assigned quantified weights. By explicitly distinguishing 8 types of edges, the unique cognitive prerequisites, basic support, emotional load progression, and ethical risk constraints of the psychological counseling discipline are solidified into computational constraints.

[0045] In one specific implementation: Teachers enter the "empathy techniques" knowledge points from the psychological counseling techniques course into the system. Three experts give difficulty scores of 3, 4, and 3.5 respectively, with an average difficulty score of 3.5, a lower bound of 0, and an upper bound of 5. The initial difficulty is then normalized. Cases involving the same knowledge point and relevance to the client's childhood trauma narrative received emotional load scores of 4.2, 4.5, and 4.0 from three experts, with a mean of 4.23. The normalized emotional load... Regarding ethical risks, due to the involvement of unsensitized traumatic details, experts scored the responses as 4.0, 3.8, and 4.2, with a mean of 4.0. The normalized ethical risk level is as follows: .

[0046] S2. Collect learning data and scene factors of the target course nodes in the multidimensional knowledge graph, input open-ended answers from the learning data into the semantic model, output a cognitive impairment error vector, and fuse the scene factors into a scene efficiency coefficient. The cognitive impairment error vector is a vector generated by aggregating multidimensional semantic comparison errors, derived from the semantic model's comparison calculation of open-ended answer text and standard answer; the scene efficiency coefficient is a numerical value that quantifies the state of environmental interference.

[0047] Specifically, unstructured subjective responses and external environmental interference are transformed into structured data that can participate in numerical calculations.

[0048] The data collected by the system includes: the ratio of the viewing time of knowledge point videos to the standard viewing time. Number of times viewed Number of pauses Response time of pop-up prompts Instant test of answer accuracy Time spent answering single or case study questions and open-ended answer text .

[0049] Contextual factors include: learning device type, whether learning is continuous, learning time period, classroom / after-class / pre-exam review scenario, level of environmental interference, remaining time for the current learning task, and the student's subjective learning intention. .

[0050] The scenario efficiency coefficient is a combination of multiple dimensions: in the quantification of learning terminal type, the baseline value is 0.9 for computer web terminals, 0.7 for tablet terminals, and 0.4 for mobile terminals; in the quantification of learning time period, the value is 0.9 for daytime classroom time, 0.6 for evening self-study, and 0.3 for late-night fragmentation; the environmental interference level provides a slider of 1-5 levels linearly normalized to [0, 1]; in the quantification of the remaining time for the current learning task, the efficiency coefficient decreases as the remaining time decreases; and in the quantification of students' subjective learning willingness, the front end provides standardized single-selection willingness labels.

[0051] Video viewing time ratio It is quantified as actual viewing time / standard full duration.

[0052] The normalized value for the response time of the pop-up notification is:

[0053] in, This represents the normalized value of the response time of the aforementioned pop-up notification. This represents the response time of the aforementioned pop-up notification. This represents the preset baseline upper limit.

[0054] The accuracy rate is calculated as the ratio of the score on each question to the full score.

[0055] Furthermore, the open-ended answer text, standard answers to knowledge points, and standard case scenarios are input into the semantic model, and a text semantic feature vector is output. The text semantic feature vector is compared with the standard reference answer vector using cosine similarity, and multidimensional error components are calculated and normalized. The multidimensional error components include errors in concept confusion, client emotion recognition bias, selection and adaptation of counseling techniques, judgment of ethical boundaries, lack of basic theory, and vague expression. A set norm is applied to the multidimensional error components, and they are aggregated to generate the cognitive impairment error vector.

[0056] Specifically, the semantic model is a neural network model used to extract deep semantic features of text, transforming unstructured subjective answers into structured error vectors that can participate in numerical calculations. The student semantic vector is compared with the standard reference answer vector using cosine similarity, and six types of error components are calculated and normalized to the [0, 1] interval.

[0057] The total number of semantic errors is calculated using the 1-norm:

[0058] in, Represents the error vector of the aforementioned cognitive impairment. The score represents the error in understanding the concept. This represents the visitor's emotion recognition bias score. This indicates that the consultation technology uses an adaptation error score. The score represents the error in the aforementioned ethical boundary judgment. The score represents the error due to missing fundamental theory. The score represents the vagueness or error of the stated expression.

[0059] When the similarity between the answer text and the standard answer text is extremely low, the error score for conceptual confusion is close to 1. This error is directly used as a penalty item and substituted into the mastery probability update formula to reduce the probability output.

[0060] In one specific implementation: In a case study question on "Visitor Emotion Recognition," a student answers, "The visitor exhibits obvious anger," while the standard answer is, "The visitor exhibits repressed sadness; anger is a surface defense." The semantic model extracts the semantic vectors of the student's answer and compares them with the standard answer vectors using cosine similarity. The deviation score for visitor emotion recognition is then calculated. The calculated result is 0.82, indicating that the consulting technology uses an adaptation error score. The value was 0.65, and the other components were relatively low. The total number of the cognitive impairment error vector 1 norm was 1.72. Based on this, it was determined that the student had a significant cognitive bias in the emotion recognition dimension.

[0061] S3. Calculate the reliability of the learning data, input the reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level into the mastery probability update formula, calculate the emotional load mismatch based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, and update the output individual mastery probability based on the emotional load mismatch.

[0062] The reliability is a coefficient obtained by penalizing abnormal time consumption, used to eliminate invalid response data such as rapid guessing; the individual mastery probability is a normalized value representing the degree of mastery of the node, calculated by fusing multidimensional evidence based on the Bayesian update framework; the emotional load mismatch term is a penalty term that quantifies the difference between the load and cognitive capacity, and the larger the value, the more the current case load exceeds the student's capacity.

[0063] Specifically, a refined estimation of mastery status is achieved by jointly driving Bayesian updates through multi-dimensional data. Standard reference time for the target course nodes, batch-annotated by psychology teaching experts, is obtained. The distribution of normal thinking and answering time for past students is statistically analyzed, and the lower quantile of normal answering is taken as a preset proportion threshold. It is determined whether the answering time is lower than the product of the preset proportion threshold and the standard reference time, and whether the answering accuracy rate is correct. If so, a quick guessing penalty item is calculated based on the degree of deviation in time consumption, placing the answer within a set penalty range.

[0064] The formula for calculating the penalty for quick guessing is:

[0065] in, This represents the penalty for quickly guessing the term. This represents the time taken to answer the question. This represents the preset ratio threshold. This represents the standard reference time.

[0066] Substitute the fast guessing penalty term into the reliability calculation formula, truncate the lower limit of the constraint reliability to a preset lower limit value, and output the reliability:

[0067] in, Represents the aforementioned reliability. This represents the penalty for quickly guessing the term. This represents the preset lower limit value.

[0068] For cases where a correct answer is given after a long period of hesitation, the time taken far exceeds the standard reference time, so the penalty for quick guessing is not triggered, and the reliability remains at 1. However, it will be reflected in the evidence calculation formula through indicators such as the number of pauses in the learning behavior data.

[0069] Furthermore, the standard reference time for the target course nodes, which are batch-annotated by psychology teaching experts, is obtained. The distribution of normal thinking and answering time of previous students is statistically analyzed, and the lower limit quantile of normal answering is taken as a preset proportion threshold. It is determined whether the answering time is lower than the product of the preset proportion threshold and the standard reference time and whether the answering accuracy is correct. If so, a quick guessing penalty item is calculated based on the degree of deviation in time consumption within a set penalty interval. The quick guessing penalty item is substituted into the reliability calculation formula, and the lower limit of the constraint reliability is truncated to a preset lower limit value. The reliability is then output.

[0070] The standard reference time is the benchmark time required for a normal response as labeled by experts. Because responses generated by rapid guessing can lead to an artificially high probability of mastery, it is necessary to remove invalid positive feedback data through anomaly detection in the time dimension, thereby ensuring the quality of evidence input for subsequent Bayesian updates.

[0071] Furthermore, the correct answer rate, reliability, learning behavior data, cognitive impairment error vector, and normalized response time of the prompt pop-up are substituted into the evidence calculation formula, and after normalization by the activation function, a learning evidence is output. The student's pre-class brief emotion scale score and cumulative continuous learning time for the day are obtained, the student's current cognitive capacity is calculated, and the cognitive capacity, the emotional load quantification value, and the ethical risk level are subjected to uniform linear scaling normalization. The normalized cognitive capacity, emotional load quantification value, ethical risk level, and scenario efficiency coefficient are input into the mismatch calculation formula, and the emotional load mismatch is output. The learning evidence, the emotional load mismatch, the weight of prior knowledge influence, and the weight of the prior edge of the multidimensional knowledge graph are input into the mastery probability update formula, and the individual mastery probability is output.

[0072] Specifically, the formula for calculating one-time learning evidence is: in, This represents one instance of learning evidence. Represents the Sigmoid normalization function. This represents the weight of the correct answer rate. This represents the accuracy rate of the answers given. This represents the reliability weight of the answer. Represents the aforementioned reliability. This represents the weight of the comprehensive indicator of learning input. This represents the normalized number of views. This indicates a prompt to understand the speed weight. This represents the normalized value of the response time of the aforementioned pop-up notification. This represents the weight of the single-dimensional semantic error item penalty. The weight representing the total penalty for the L1 semantic errors is... The 1-norm represents the error vector of the cognitive impairment. This represents the semantic error categories for each dimension.

[0073] The Sigmoid function output naturally maps to [0, 1]. For skewed data in psychological counseling case responses, teachers can configure a global offset coefficient in the background to adjust it to fit the subject data distribution.

[0074] The learning evidence calculation formula is set with six adjustable weights, and the weight allocation is tailored to the learning characteristics of psychological counseling courses. Prioritize the acceptance of answers based on deep thinking, and reduce the interference of false high scores caused by rapid guessing; The maximum weight is applied to all texts, and stronger penalties are imposed on cognitive biases related to concepts, emotions, ethics, and basic theories, targeting numerous open-ended subjective questions such as case analyses and simulated dialogues in the course. With moderate auxiliary weight, a logarithmic function is used to smooth out the fluctuations in the number of review times, avoiding excessive review that could inflate the evaluation score. It has the smallest value and serves only as a weak auxiliary correction term. Constraints: and Each penalty coefficient is configured independently and has no normalization constraint. All indicators with original scores in the range of 0 to 5 must be linearly normalized to the [0, 1] interval before participating in this formula and all mathematical model calculations to eliminate dimensional differences. All weights can be customized and adjusted by teachers in the backend to adapt to different teaching scenarios.

[0075] Obtain the student's pre-class brief mood scale score and the cumulative continuous learning time for the day, and calculate the student's current cognitive capacity. The original cutoff range is [1, 5], which needs to be mapped to the [0, 1] interval through linear scaling. The normalized cognitive capacity, the quantified emotional load, the ethical risk level, and the scenario efficiency coefficient are input into the formula for calculating the emotional load mismatch, and the emotional load mismatch is output:

[0076] in, This represents the emotional load mismatch item. Representing the emotional load, This represents the level of ethical risk. This represents the normalized current cognitive capacity of the student. This represents the efficiency coefficient of the described scenario. This represents the weight coefficient of the corresponding item.

[0077] The intensity of emotional stimulation inherent in the knowledge points / cases is a static fixed value, which is quantified and assigned a grade of 0-5. Non-traumatic theory is scored 0 points, and severely traumatic material is scored 4.15 points. The risk weight of the material causing vicarious trauma and ethical conflict is quantified on a scale of 0-5, with 0 points for no privacy or details and 45 points for severe trauma cases that have not been desensitized; It is the cognitive capacity of student i at time t, defined as the upper limit of the psychological capacity that student can bear in terms of emotional and information processing at that time. It is a dynamic variable, and the input is the score of the pre-class simple emotion scale (0~5 points) and the cumulative continuous learning time of the day.

[0078] The learning evidence, the emotional load mismatch, the weight of prior knowledge influence, and the weight of the prior edges of the multidimensional knowledge graph are input into the mastery probability update formula, and the individual mastery probability is output.

[0079] The probability update formula is as follows:

[0080] in, This represents the updated probability of the individual possessing the knowledge. This represents the probability of the individual possessing the knowledge before the update. This represents the weight of the evidence in this learning session. This represents one instance of learning evidence. Represents the emotional load penalty weight. This represents the emotional load mismatch item. This represents the weight of the aforementioned prior knowledge. Represents the preceding node To the current node Pre-edge weights, On behalf of students In the preceding node The probability of an individual mastering something.

[0081] Refers to students completing knowledge points The contribution weight of immediate learning behavior evidence such as post-learning responses, classroom feedback, and emotional self-assessment is set in the range of [0.1, 0.8]. The higher the test accuracy and classroom interaction completion, the closer it is to the upper bound of 0.8. When students are passively avoiding or emotionally collapsing, the weight is reduced to 0.1~0.4. This is used to control the influence of a single learning behavior, avoid a single poor or excellent performance directly overturning the overall mastery probability, and smooth out fluctuations in single data.

[0082] This is used to measure the constraining effect of prior knowledge mastery on the understanding of current knowledge points. It is adapted to the progressive logic of psychological counseling knowledge. The benchmark value range is [0, 1], and the interval is not set to 1. For knowledge points with high emotional trauma, the value is 0.8, and for general theories, the value is 0.3-0.5. This means that prior knowledge is only a constraint rather than a determining factor. Even if the foundation is weak, students can still partially master new knowledge through intensive classroom guidance and teacher guidance, and will not directly reduce the probability of mastery to zero.

[0083] Prerequisite knowledge points in the representative graph With target knowledge points The strength of the association dependency is in the range [0, 1]. The value is 0 for no related prerequisite knowledge, 0.71 for strong dependent core prerequisite knowledge points, and 0.1 or 0.4 for weakly related auxiliary knowledge points.

[0084] This is a decay coefficient designed to address the emotional load mismatch in psychological counseling courses, with a value range of [0.2, 1]. The larger The smaller the value, the lower the value, with a minimum of 0.2. Knowledge points without emotional load are fixed at a value of 1. This is a separately set penalty item, distinct from ordinary subject models, and is specifically adapted to scenarios of vicarious trauma and emotional exhaustion in psychological counseling. Penalties for content without emotional load can be automatically canceled without interfering with the calculation of regular theoretical knowledge points.

[0085] Because the probability update formula integrates three types of variables—in-class learning performance, prior knowledge logic, and emotional psychological load—it can simultaneously consider the mismatch between students' current learning performance, prior knowledge mastery, and the emotional load unique to psychological counseling courses. This addresses the shortcoming of simply using a general mastery probability formula, which makes it difficult to clearly explain the technical effects. By capturing changes in students' emotions and fatigue levels within a single class using time-series dynamic variables, and achieving dynamic real-time updates, this model can dynamically adjust and monitor probabilities in real time, and provide early warnings of vicarious trauma risks, compared to static general models. Relying on... Quantifying pre-existing dependencies can also accurately distinguish between two different learning obstacles: "weak foundation leading to inability to learn" and "emotional overload leading to inability to learn." This allows the system to push supplementary courses or emotional guidance interventions separately, rather than uniformly repeating practice questions.

[0086] In one specific implementation: when a student answers a question correctly within 5 seconds that takes a standard reference time of 60 seconds, a preset proportion threshold k=0.3, the system determines... And if the answer is correct, the calculation of the quick guessing penalty item will be triggered. Reliability This significantly reduces the contribution weight of the answer in a single learning evidence calculation, preventing an artificially inflated probability of mastery. When students are in a pre-exam review scenario with low environmental interference, the scenario efficiency coefficient... The higher level reduces the overall emotional load mismatch, allowing the system to push relatively high-load cases without triggering interception.

[0087] S4. Based on the mismatch between the individual's mastery probability and the emotional load, a learning path is generated and pushed to the terminal. The learning path is a data stream containing a sequence of course nodes and remedial jump directions, dynamically calculated by the system based on multi-dimensional quantitative constraints.

[0088] Specifically, the prior comprehensive mastery level of the target course node is calculated. If the prior comprehensive mastery level is lower than a preset threshold, the target course node is intercepted, and the basic course nodes are traced back along the basic support edge to generate a basic course remedial path. When the individual mastery probability is lower than the mastery threshold and the error score for missing basic theory or conceptual confusion in the cognitive impairment error vector is greater than the corresponding set threshold, the graph distance from the basic course node to the target course node, the cognitive load of the basic course node, and the semantic similarity between the open-ended answer text and the basic course node are obtained to calculate the comprehensive score of the remedial path.

[0089] in, This represents the overall score of the remedial path. Represents the basic node To the target node The basic support edge weight, This represents the open-ended answer text and the basic course nodes. semantic similarity, Represents the distance of the aforementioned spectrum. This represents the probability that the individual possesses the knowledge. The score represents the error due to missing fundamental theory.

[0090] A candidate remedial path set is generated by sorting the remedial path comprehensive scores in descending order. The individual mastery probability, the emotional load mismatch, the candidate remedial path set, and the scenario efficiency coefficient are then weighted and sorted using the push priority calculation formula.

[0091] in, This represents the push priority. This represents the probability that the individual possesses the knowledge. This represents the emotional load mismatch item. This represents the efficiency coefficient of the described scenario. This represents the overall score of the remedial path. This represents the weight coefficient of the corresponding item.

[0092] Output and push the learning path to the student's terminal. If the student's mastery probability is less than 0.6 and The system automatically skips the recommendations of adjacent nodes and directly backtracks along the basic support edge to the ordinary psychology basic node with a distance of 2 hops in the graph for remediation.

[0093] Further, the prior comprehensive mastery of the target course node is calculated. If the prior comprehensive mastery is lower than a preset threshold, the target course node is intercepted, and the basic course nodes are traced back along the basic support edge to generate a basic course remedial path. When the individual mastery probability is lower than the mastery threshold and the error score of missing basic theory or the error score of conceptual confusion in the cognitive impairment error vector is greater than the corresponding threshold, the graph distance from the basic course node to the target course node, the cognitive load of the basic course node, and the semantic similarity between the open-ended answer text and the basic course node are obtained. A comprehensive score for the remedial path is calculated, and a candidate remedial path set is generated by sorting the candidate remedial paths in descending order of the comprehensive score. The individual mastery probability, the emotional load mismatch, the candidate remedial path set, and the scenario efficiency coefficient are substituted into the push priority calculation formula for weighted sorting, output, and pushed to the student terminal.

[0094] The basic course remedial path is a jump sequence that goes back from the current course to underlying courses such as general psychology, and uses graph topological distance and semantic similarity to accurately locate the root cause of knowledge gaps.

[0095] Furthermore, taking classes as units, the individual mastery probability, cognitive impairment error vector, abnormal answering time, and knowledge base query behavior of students at each course node are summarized to construct a group cognitive impairment matrix. The class knowledge point error rate is calculated by dividing the number of incorrect answers in the target course node by the total number of students in the class. The mean semantic error is obtained by calculating the mean error between the semantic vector setting of students' answers and the standard answer. The variance of the ratio of multiple answering times is calculated and the upper limit is truncated to obtain the answering time dispersion. The persistent knowledge query gap is obtained by combining the number of queries and the mastery increase with a weighted score. The class knowledge point error rate, the mean semantic error, the answering time dispersion, and the persistent knowledge query gap are linearly normalized and then weighted and summed to output the group impairment strength. Students with similar cognitive impairment error vector patterns are clustered, and class-specific content is generated based on the group impairment strength. The class-specific content is then pushed to the corresponding student groups.

[0096] Specifically, taking classes as units, the individual mastery probability, cognitive impairment error vector, abnormal answering time, and knowledge base query behavior of students at each course node are summarized to construct a group cognitive impairment matrix. This group cognitive impairment matrix is ​​a feature matrix that aggregates the overall learning behavior data of the class, moving from individual diagnosis to group commonality analysis, thus assisting teachers in implementing targeted interventions.

[0097] The formula for calculating the error rate of knowledge points in a class is:

[0098] in, This represents the error rate of the knowledge points in the class. This represents the number of people who answered the questions incorrectly. This represents the total number of students in the class.

[0099] The formula for calculating the mean semantic error is:

[0100] in, This represents the mean of the semantic errors. Represents the total number of students in the class. On behalf of students Response semantic vector This represents the semantic vector of the standard answer.

[0101] The formula for calculating the dispersion of answering time is:

[0102] in, This represents the dispersion of the time taken to answer the question. This represents the preset upper limit of variance cutoff. This represents the variance of the ratio of the time taken to respond to multiple responses.

[0103] The formula for calculating the persistent knowledge gap is:

[0104] in, This represents the persistent knowledge gap. Represents the weight of the number of queries. Represents the number of queries. This represents the weighting of price increases. This indicates that the price increase is being monitored.

[0105] The group obstacle strength is output by linearly normalizing and weighting the class knowledge point error rate, the mean semantic error, the answer time dispersion, and the persistent knowledge query gap. Students with similar cognitive obstacle error vector patterns are clustered, and class-specific content is generated based on the group obstacle strength.

[0106] In one specific implementation: In a class of 40 students, 28 students answered incorrectly on the knowledge point of "selection and adaptation of consulting techniques." The class's error rate for this knowledge point is... Mean of semantic errors Dispersion of answering time ; persistent knowledge search gap After linear normalization of the four indicators, the weighted summation yields a group impairment intensity of 0.55, which exceeds the preset threshold of 0.5. The system automatically generates a "Consultation Technology Adaptation Error Topic" and pushes it to students in the class with similar cognitive impairment error vector patterns.

[0107] Furthermore, a built-in psychology knowledge base is included, which is linked with the semantic model and class group obstacle clustering. When a term or misconception is repeatedly queried, repeatedly answered incorrectly, or frequently appears in open-ended answers by multiple students, candidate knowledge entries, candidate case comparisons, and candidate test questions are automatically generated. Upon receiving teacher review instructions, the candidate knowledge entries, candidate case comparisons, and candidate test questions are added to the psychology knowledge base and question bank, and are solidified into the multidimensional knowledge graph as misconception nodes or easily confused edges.

[0108] Specifically, the psychology knowledge base is a dynamic database storing psychological concepts, cases, and test questions. Through a closed-loop feedback mechanism, the knowledge base self-evolves as the teaching process progresses. When the system detects that a term appears more frequently in the knowledge base query log than a preset threshold, or appears frequently as an error feature in open-ended answer texts, an entry generation mechanism is triggered. The generated candidate content is then officially added to the database after being reviewed by teachers on the web management platform, and corresponding misconception nodes are added to the knowledge base graph, or existing easily confused edge weights are reinforced.

[0109] In one specific implementation: multiple students repeatedly exposed the ethical error of "the counselor making value judgments on the client" in case analysis. The system automatically generated candidate entries for "comparison of cases based on the principle of value neutrality". After review by the teacher, these entries were solidified into the graph as ethical risk nodes and the interception priority of these nodes was increased in subsequent pushes.

[0110] Furthermore, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent teaching and delivery method for psychological counseling courses.

[0111] Furthermore, this application also proposes an intelligent teaching and recommendation system for psychological counseling courses, comprising: a graph construction module, used to acquire original data of psychological counseling courses, decompose the original data of psychological counseling courses into multiple course nodes including emotional load quantification values ​​and ethical risk levels, and construct a multidimensional knowledge graph composed of the multiple course nodes and directed edges; and a data acquisition and semantic recognition module, used to collect students' learning data and scene factors on target course nodes in the multidimensional knowledge graph, input open-ended answer text in the learning data into a semantic model, output a cognitive impairment error vector, and integrate the scene factors into scene efficiency. The module includes a coefficient and a mastery probability calculation module, used to calculate reliability based on the answering time in the learning data and the standard reference time. It inputs the reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level into the mastery probability update formula. Based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, it calculates emotional load mismatches and updates the output individual mastery probability based on these mismatches. A path push module is used to generate and push learning paths to student terminals based on the individual mastery probability and the emotional load mismatches.

[0112] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent teaching delivery of psychological counseling courses, characterized in that, include: S1. Obtain the original data and break it down into multiple course nodes to construct a multi-dimensional knowledge graph that includes the quantitative value of emotional load and the level of ethical risk. S2. Collect learning data and scenario factors of target course nodes in the multidimensional knowledge graph, input open-ended answers in the learning data into the semantic model, output cognitive impairment error vector, and fuse the scenario factors into scenario efficiency coefficient; S3. Calculate the reliability of the learning data, input the reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level into the mastery probability update formula, calculate the emotional load mismatch based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, and update the output individual mastery probability based on the emotional load mismatch. S4. Based on the mismatch between the individual's mastery probability and the emotional load, a learning path is generated and pushed to the terminal.

2. The intelligent teaching and delivery method for psychological counseling courses according to claim 1, characterized in that, The multidimensional knowledge graph is constructed in the following manner: Obtain the original materials of the psychological counseling course and break down the course content into concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes; The system receives multiple sets of scoring samples from instructors for concept nodes, skill nodes, case nodes, ethical risk nodes, emotion recognition nodes, basic course nodes, and misconception nodes. The system calculates the arithmetic mean of the multiple sets of scoring samples and maps them to a preset normalization interval through linear normalization to obtain the initial difficulty, cognitive load requirement, emotional load quantification value, and ethical risk level of each node. Configure directed edges and edge weights between nodes. The directed edges include at least cognitive prerequisite edges, basic support edges, skill transfer edges, emotional load progression edges, case context matching edges, ethical risk constraint edges, easily confused edges, and remedial jump edges. Output the multidimensional knowledge graph composed of the various types of nodes and the directed edges.

3. The intelligent teaching and delivery method for psychological counseling courses according to claim 2, characterized in that, The cognitive impairment error vector is obtained as follows: Input open-ended answer text, standard answers to knowledge points, and standard case scenarios into the semantic model, and output the text semantic feature vector; The text semantic feature vector is compared with the standard reference answer vector by cosine similarity, and multidimensional error components are calculated and normalized. The multidimensional error components include conceptual confusion error score, visitor emotion recognition deviation score, counseling technology selection and adaptation error score, ethical boundary judgment error score, basic theory missing error score, and vague expression error score. The multidimensional error components are assigned a set norm and aggregated to generate the cognitive impairment error vector.

4. The intelligent teaching and delivery method for psychological counseling courses according to claim 3, characterized in that, The reliability is calculated as follows: Obtain the standard reference time of the target course nodes marked in batches by psychology teaching experts, statistically analyze the distribution of normal thinking and answering time of previous students, and take the lower limit quantile of normal answers as the preset proportion threshold. Determine whether the time taken to answer questions is lower than the product of the preset percentage threshold and the standard reference time, and whether the accuracy rate is correct: If so, then calculate the quick guess penalty item within the set penalty range based on the degree of deviation in time consumption; Substitute the fast guessing penalty term into the reliability calculation formula, truncate the lower limit of the constraint reliability to a preset lower limit value, and output the reliability.

5. The intelligent teaching and delivery method for psychological counseling courses according to claim 4, characterized in that, The mismatch between the individual mastery probability and the emotional load is calculated as follows: Substitute the answer accuracy, reliability, learning behavior data, cognitive impairment error vector, and normalized value of prompt pop-up response time into the evidence calculation formula, and output a learning evidence after normalization by the activation function; The student's pre-class brief emotional scale score and cumulative continuous learning time for the day are obtained. The student's current cognitive capacity is calculated, and the cognitive capacity, the quantitative value of emotional load, and the ethical risk level are subjected to uniform linear scaling normalization. The normalized cognitive capacity, the quantified emotional load, the ethical risk level, and the scenario efficiency coefficient are input into the mismatch calculation formula, and the emotional load mismatch is output. The learning evidence, the emotional load mismatch, the weight of prior knowledge influence, and the weight of the prior edges of the multidimensional knowledge graph are input into the mastery probability update formula, and the individual mastery probability is output.

6. The intelligent teaching and delivery method for psychological counseling courses according to claim 5, characterized in that, The learning path is generated and pushed in the following manner: Calculate the prior comprehensive mastery level of the target course node. If the prior comprehensive mastery level is lower than a preset threshold, intercept the target course node and backtrack along the basic support edge to generate a basic course remedial path. When the individual's mastery probability is lower than the mastery threshold and the error score for missing basic theory or the error score for conceptual confusion in the cognitive impairment error vector is greater than the corresponding threshold, the graph distance from the basic course node to the target course node, the cognitive load of the basic course node, and the semantic similarity between the open-ended answer text and the basic course node are obtained. A comprehensive score for the remedial path is calculated, and a set of candidate remedial paths is generated by sorting them in descending order according to the comprehensive score of the remedial path. The individual mastery probability, the emotional load mismatch, the candidate remedial path set, and the scenario efficiency coefficient are substituted into the push priority calculation formula for weighted sorting, and the result is output and pushed to the student terminal.

7. The intelligent teaching and delivery method for psychological counseling courses according to claim 6, characterized in that, Also includes: S5. Generate and push group-specific content; Specifically, it includes: Using classes as units, the individual mastery probability, cognitive impairment error vector, abnormal answering time, and knowledge base query behavior of students at each course node are summarized to construct a group cognitive impairment matrix. The error rate of knowledge points in a class is calculated by dividing the number of students who made mistakes at the target course node by the total number of students in the class. The mean of semantic errors is obtained by calculating the mean value of the semantic vector setting error between the students' answers and the standard answers. The variance of the ratio of time spent on multiple answers is calculated and the upper limit is truncated to obtain the dispersion of answer time. The persistent knowledge query gap is obtained by combining the number of queries and the weighted scoring of the mastery increase. The group barrier strength is output by linearly normalizing the class knowledge point error rate, the mean semantic error, the answer time dispersion and the stubborn knowledge query gap, and then weighting and summing them. Students with similar cognitive impairment error vector patterns are clustered, and class-specific content is generated based on the group impairment intensity. This class-specific content is then pushed to the corresponding student groups.

8. The intelligent teaching and delivery method for psychological counseling courses according to claim 7, characterized in that, Also includes: S6. Knowledge base adaptive expansion; Specifically, it includes: A built-in psychology expertise base is included, which is linked with the semantic model and class group obstacle clustering. When a term or misconception is repeatedly searched by multiple students, repeatedly answered incorrectly, or frequently appears in open-ended answers, candidate knowledge entries, candidate case comparisons, and candidate test questions are automatically generated. Upon receiving the teacher's review instruction, the candidate knowledge items, the candidate case comparisons, and the candidate test questions are added to the psychology professional knowledge base and question bank, and are solidified into the multidimensional knowledge graph as misconception nodes or easily confused edges.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent teaching and push method for psychological counseling courses as described in any one of claims 1-8.

10. A smart teaching and delivery system for psychological counseling courses, characterized in that, include: The graph construction module is used to acquire the original data of the psychological counseling course, decompose the original data of the psychological counseling course into multiple course nodes containing the emotional load quantification value and the ethical risk level, and construct a multi-dimensional knowledge graph composed of the multiple course nodes and directed edges. The data acquisition and semantic recognition module is used to collect students' learning data and scenario factors at the target course nodes in the multidimensional knowledge graph, input open-ended answer text in the learning data into the semantic model, output cognitive impairment error vector, and fuse the scenario factors into scenario efficiency coefficient. The mastery probability calculation module is used to calculate the reliability based on the answering time in the learning data and the standard reference time. The reliability, the learning data, the cognitive impairment error vector, the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level are input into the mastery probability update formula. Based on the scenario efficiency coefficient, the emotional load quantification value, and the ethical risk level, the emotional load mismatch is calculated, and the individual mastery probability is updated and output based on the emotional load mismatch. The path push module is used to generate and push learning paths to student terminals based on the mismatch between the individual's mastery probability and the emotional load.

Images