An Adaptive Learning Duration Management Method Based on Learning Behavior

Abstract

This invention belongs to the field of educational technology, specifically an adaptive learning time management method based on learning behavior. It compares the current and future cognitive load equivalents of a student's current learning session with the expected load range derived from parental control rules. This identifies immediate or predicted load conflicts, and based on the conflict type and magnitude, triggers and executes corresponding control instructions that integrate the basic duration of a single rest or the selection range of low-cognitive-load knowledge points to adjust the learning process. Finally, the triggering logic of the control instructions is optimized through the feedback data package generated after the execution of the control instructions. This approach quantifies parents' personalized management intentions into identifiable and operable control criteria, and ensures that the triggering of intervention instructions such as rest interruptions or path adjustments matches the student's real-time cognitive resource consumption status, thereby improving the effectiveness and adaptability of the control strategy.

Description

Technical Field

[0001] This invention belongs to the field of educational technology, specifically an adaptive learning time management method based on learning behavior. Background Technology

[0002] Currently, the learning time management solutions adopted by mainstream online education platforms are mainly based on fixed clock mechanisms, which can be summarized into periodic management modes based on fixed time intervals and upper limit management modes based on fixed total time. However, this existing technology has inherent defects, specifically: 1. Existing solutions only trigger management based on preset, constant time parameters, without considering that students' cognitive load changes dynamically in real time when facing learning content of different difficulty and types. The timing of forced breaks or interruptions may not match the student's actual cognitive fatigue point, which may interrupt the efficient learning state or fail to intervene in time when cognitive overload occurs.

[0003] 2. Parents' control intentions are simplified into rigid on / off switches or time limits, lacking the ability to describe dimensions such as learning intensity and content difficulty gradient. This makes it impossible for control behaviors to be dynamically adjusted based on the actual situation of the learning content, resulting in rigid intervention methods, low personalization, and poor user experience and control effects. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, this invention provides an adaptive learning duration control method based on learning behavior, which can effectively solve the problems mentioned in the prior art.

[0005] The objective of this invention can be achieved through the following technical solution: an adaptive learning duration management method based on learning behavior, comprising: determining the current cognitive load equivalent and future cognitive load equivalent of a student's current learning session by querying a knowledge point network that stores the prior dependencies of knowledge points.

[0006] The parent-configured control rules are transformed into a range that includes the expected workload range, the basic duration of a single rest, and the selection range of knowledge points with low cognitive load.

[0007] The current cognitive load equivalent and the future cognitive load equivalent are compared with the expected load range to identify immediate load conflicts or predict load conflicts.

[0008] If a conflict is detected, the corresponding adjustment instruction that integrates the basic duration of the single rest or the range of low cognitive load knowledge points is triggered and executed based on the conflict type and magnitude to adjust the learning process.

[0009] Collect learning efficiency and interaction feedback data after the execution of the control commands, generate feedback data packets, and use them to optimize the triggering logic of the control commands.

[0010] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention calculates the cognitive difficulty gradient of the learning content in the real-time sequence by querying the knowledge point network and combining it with the individualized set of knowledge points mastered by the student, and then accumulates the dynamically changing current cognitive load equivalent and future cognitive load equivalent. This solves the defect of the prior art that only relies on fixed clock triggering control, so that the triggering of intervention commands such as rest interruption or path adjustment can match the student's real-time cognitive resource consumption status, and avoid the disconnect between the control timing and the actual learning rhythm.

[0011] (2) This invention specifically transforms the control rules configured by parents into structured parameters that include the expected load range, the basic duration of a single rest, and the selection range of knowledge points with low cognitive load. This enables parents' personalized management intentions to be quantified into identifiable and operable control criteria, overcoming the problem of the existing technology's extensiveness in only being able to set a single total duration parameter, and providing a data foundation for achieving personalized and refined dynamic control.

[0012] (3) Based on the type and magnitude of real-time or predicted load conflicts, this invention triggers rest commands, low-cognitive-load knowledge point replacement commands, or challenge commands in a differentiated manner. By selecting strategies based on the nature and degree of conflict, it replaces the one-size-fits-all forced interruption of existing technologies. At the same time, by collecting feedback data packets after command execution and using them to optimize the triggering logic, the invention has the ability to self-adjust according to the actual intervention effect, thereby improving the effectiveness and adaptability of the control strategy. Attached Figure Description

[0013] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0014] Figure 1 This is a schematic diagram illustrating the execution flow of the method steps of the present invention.

[0015] Figure 2 This is a schematic diagram illustrating the process for determining the current cognitive load equivalent in this invention.

[0016] Figure 3 This is a schematic diagram of the triggering process of the control command of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 As shown, the present invention provides an adaptive learning duration control method based on learning behavior, including: S1. Determining the current cognitive load equivalent and future cognitive load equivalent of the student's current learning session by querying a knowledge point network that stores the prior dependencies of knowledge points.

[0019] The knowledge point network is a graph data structure used to represent the knowledge system structure of the subject to which the current learning session belongs. In this network, nodes represent independent knowledge points, and directed edges between nodes represent preorder dependencies between knowledge points, with the direction pointing from the preorder knowledge point to the subsequent knowledge point, indicating that learning the subsequent knowledge point requires mastering the preorder knowledge point.

[0020] Reference Figure 2 As shown, the process of determining the current cognitive load equivalent is as follows: First, determine the set of knowledge points that the student has mastered. For each knowledge point in the knowledge point network, calculate the student's historical answer accuracy rate on that knowledge point. Compare the historical answer accuracy rate with a preset mastery judgment threshold. If the historical answer accuracy rate is greater than or equal to the preset mastery judgment threshold, then the knowledge point is determined to be a mastered knowledge point and added to the set.

[0021] The preset mastery judgment threshold is determined based on the historical answer data of all students. Specifically, it is calculated as follows: the third percentile of the historical answer accuracy of all students on each knowledge point is calculated, and then the arithmetic mean of the percentile of all knowledge points is calculated.

[0022] Secondly, for each knowledge point learned in the current learning session, its cognitive difficulty gradient is calculated. Specifically, the prior dependencies of the knowledge point are obtained from the knowledge point network. If the knowledge point belongs to the set of mastered knowledge points, its cognitive distance is set to 0. Otherwise, starting from the knowledge point, the prior dependencies are traced backward until a prior knowledge point belonging to the set of mastered knowledge points is found for the first time. The number of edges of the traced path is recorded as the cognitive distance of the knowledge point.

[0023] The shortest prior dependency path length required to master a knowledge point from the starting point of the knowledge network is defined as the reference cognitive distance of that knowledge point. The cognitive difficulty gradient of the knowledge point is determined by the ratio of the cognitive distance to the reference cognitive distance. The starting point of the knowledge network is the initial set of the most basic or no-prerequisite knowledge points in the subject knowledge system, such as the concept of natural numbers in mathematics or the basic pinyin in Chinese.

[0024] Finally, the cognitive difficulty gradients of all learned knowledge points in the current learning session are summed up, and the sum is the current cognitive load equivalent.

[0025] The process of determining the future cognitive load equivalent is based on the current cognitive load equivalent, and specifically includes: selecting a predetermined number of knowledge points from the sequence of knowledge points that have been planned but not yet learned in the current learning session to form the sequence to be evaluated. The predetermined number of knowledge points can be set to 3 to 5 knowledge points according to the learning pace.

[0026] Determine the cognitive difficulty gradient of each knowledge point in the sequence to be evaluated, and use the sum of the cognitive difficulty gradients of all knowledge points in the sequence as the future cognitive load increment.

[0027] The sum of the future cognitive load increment and the current cognitive load equivalent is taken as the future cognitive load equivalent.

[0028] This invention, through querying a knowledge point network and combining it with a student's individualized set of mastered knowledge points, calculates the cognitive difficulty gradient of the learning content in a real-time sequence, and then accumulates these to obtain the dynamically changing current cognitive load equivalent and future cognitive load equivalent. This overcomes the shortcomings of existing technologies that rely solely on fixed clock triggers for control, ensuring that intervention commands such as rest breaks or path adjustments are matched to the student's real-time cognitive resource consumption status, thus avoiding a disconnect between control timing and actual learning pace.

[0029] S2. The parent-configured control rules are transformed into a range that includes the expected load range, the basic duration of a single rest, and the selection range of low cognitive load knowledge points. The transformation process specifically includes: S21. Transformation of the expected load range: The platform has multiple learning load intensity mode options stored in advance. Each mode is associated with a preset cognitive load equivalent range. Based on the learning load intensity mode selected by the parent, the corresponding cognitive load equivalent range is queried and extracted as the expected load range.

[0030] S22. Conversion of basic duration for a single rest session: The platform has multiple rest session mode options pre-stored. Each mode defines the rest duration corresponding to the continuous learning period after reaching a preset duration. Based on the rest session mode selected by the parent, the defined rest duration is extracted as the basic duration for the single rest session.

[0031] S23. Transformation of the selection range of low cognitive load knowledge points: Based on the content preference chapters configured by parents, the knowledge points belonging to the content preference chapters within the set of knowledge points already mastered by students are divided into the selection range of low cognitive load knowledge points.

[0032] This invention transforms the parent-configured management rules into structured parameters that include the desired workload range, the basic duration of a single rest session, and the selection range of low cognitive load knowledge points. This allows parents' personalized management intentions to be quantified into identifiable and operable control criteria, overcoming the limitations of existing technologies that can only set a single total duration parameter. It provides a data foundation for achieving personalized and refined dynamic control.

[0033] S3. Compare the current cognitive load equivalent and the future cognitive load equivalent with the expected load range to identify immediate load conflicts or predicted load conflicts.

[0034] The specific process for identifying or predicting immediate load conflicts includes: if the current perceived load equivalent exceeds the upper limit of the expected load range, then it is determined that an immediate load conflict exists.

[0035] If the future perceived load equivalent exceeds the upper limit of the expected load range, then a predicted load conflict is identified.

[0036] S4. If a conflict is detected, the corresponding adjustment instruction that integrates the basic duration of the single rest or the range of low cognitive load knowledge points is triggered and executed based on the conflict type and conflict amplitude to adjust the learning process.

[0037] Reference Figure 3 As shown, the control instructions triggered based on the conflict type and conflict amplitude include: when there is an immediate load conflict, it indicates that the student's current cognitive load has exceeded their individual tolerance range, and a rest instruction needs to be triggered immediately to relieve the pressure. In order to adapt to the severity of the overload, the difference between the current cognitive load equivalent and the upper limit of the expected load range is defined as the conflict amplitude. The product of the conflict amplitude and the preset duration adjustment coefficient is added to the basic duration of a single rest to obtain the specific execution duration of the rest instruction.

[0038] It should be understood that the preset duration adjustment coefficient is used to convert the difference in cognitive load into an actual extension of rest time, and its value should match the student's experiential learning endurance at the current grade level. For example, this coefficient can be calibrated to 0.5 minutes per unit of load through experiments at the corresponding grade level, meaning that for every unit exceeding the upper limit of load, the rest time is increased by 0.5 minutes on top of the base duration.

[0039] When a predicted load conflict exists, it indicates that continuing learning along the current path will lead to overload in the future, requiring a proactive adjustment of the learning path. The processing logic is based on a tiered approach according to the severity of the conflict: the difference between the future cognitive load equivalent and the upper limit of the expected load range is defined as the conflict magnitude.

[0040] If the conflict magnitude is greater than or equal to the path adjustment threshold set for the subject to which the current learning session belongs, it is judged as a severe conflict, triggering a low cognitive load knowledge point replacement instruction for the sequence to be evaluated, which aims to reduce the objective difficulty of subsequent paths.

[0041] If the conflict magnitude is less than the path adjustment threshold, it is determined to be a minor conflict, triggering a challenge command. This aims to enhance students' subjective ability to cope with challenges and maintain their original learning goals by providing learning support.

[0042] It should be noted that the process for adjusting the target path threshold for the subject to which the current learning session belongs is as follows: For the subject to which the current learning session belongs, trace data from multiple historical learning sessions are collected. The trace data for each historical session includes: the conflict amplitude calculated during the session, and the student's average correct answer rate on subsequent assessment tasks related to the knowledge point of the session within a preset time window after the session ends.

[0043] The average correct answer rate data of all historical learning sessions is used as input to the K-means clustering algorithm. In this embodiment, the necessary parameter, the number of clusters K, is set to 2. The output is the clusters. Historical learning sessions contained in the cluster with higher cluster center values ​​are defined as efficient sessions, and those contained in the cluster with lower cluster center values ​​are defined as inefficient sessions. The specific implementation process of the K-means clustering algorithm is a well-known technique and will not be described in detail here.

[0044] It should be noted that the K-means clustering algorithm is only one algorithm for clustering measurement points provided in this embodiment. Under the premise that the purpose of clustering can be achieved, implementers may also use other feasible clustering algorithms in the prior art as other implementation methods, and this invention does not limit them.

[0045] The conflict amplitudes corresponding to all inefficient sessions are screened out to form a set of amplitudes to be analyzed, and the statistical center value of all conflict amplitudes is calculated. The statistical center value is then determined as a candidate value for the path adjustment threshold.

[0046] Using the candidate values ​​as decision boundaries, a backtracking simulation is performed on the traceability data of all historical learning sessions. The backtracking simulation logic is as follows: if the conflict magnitude is greater than or equal to the candidate value, it is determined that a path adjustment should be triggered; if the conflict magnitude is less than the candidate value, it is determined that a path adjustment should not be triggered.

[0047] The simulation decision results are compared with the inefficient and efficient session categories: for historical data marked as inefficient sessions, a simulation decision that the session should be triggered is counted as a correct judgment; for historical data marked as efficient sessions, a simulation decision that the session should not be triggered is counted as a correct judgment.

[0048] The percentage of all correct judgments out of the total number of historical sessions is used as the decision accuracy rate.

[0049] If the decision accuracy reaches a preset standard, such as exceeding 90%, then the candidate value is determined as the path adjustment threshold. Otherwise, with the candidate value as the center, a numerical search interval is determined by expanding in both positive and negative directions with twice the standard deviation of the set of amplitudes to be analyzed as the floating radius. Iterative search is performed within this interval to obtain the value with the highest decision accuracy, which is then determined as the path adjustment threshold.

[0050] To ensure that the replaced path can effectively eliminate the predicted load conflict, the low cognitive load knowledge point replacement instruction for the sequence to be evaluated is an iterative optimization process, which specifically includes: selecting the knowledge point with the highest cognitive difficulty gradient from the sequence to be evaluated as the target knowledge point.

[0051] The knowledge point network is queried to obtain all adjacent knowledge points that have a direct dependency relationship with the target knowledge point. Among them, knowledge points whose cognitive difficulty gradient is lower than that of the target knowledge point and whose value is within the selection range of low cognitive load knowledge points are selected to form a set of candidate replacement knowledge points.

[0052] It should be noted that the direct dependency specifically refers to all other knowledge point nodes in the knowledge point network that are directly connected to the target knowledge point node through a directed edge. The direction of the dependency represented by the directed edge includes a preorder dependency from the adjacent knowledge point to the target knowledge point, or a postorder dependency from the target knowledge point to the adjacent knowledge point.

[0053] For each knowledge point in the candidate replacement knowledge point set, calculate its replacement utility score. The replacement utility score is calculated by a linear weighted summation function, which considers the following two factors: (i) based on the difference between the cognitive difficulty gradient of the target knowledge point and the cognitive difficulty gradient of the current candidate knowledge point, determine the difference in cognitive load gradient that the knowledge point can reduce after replacing the target knowledge point, which can be recorded as the reduction contribution.

[0054] (ii) The semantic correlation degree between the current candidate knowledge point and the other knowledge points in the sequence to be evaluated in the knowledge point network. The process of determining the semantic correlation degree is as follows: In the knowledge point network, calculate the shortest path length, counted by edges, between the current candidate knowledge point and each other knowledge point in the sequence to be evaluated except for the target knowledge point.

[0055] The arithmetic mean of all calculated shortest path lengths is taken, and the reciprocal of this arithmetic mean is normalized to obtain a value between 0 and 1, which is used as the semantic association metric. The shorter the shortest path length, the stronger the conceptual connection between knowledge points in the network, and the higher its semantic association metric.

[0056] In one specific implementation, the weight of the contribution to reducing cognitive load can be set to 0.7 and the weight of semantic relevance can be set to 0.3, so that the optimization focuses more on reducing cognitive load while taking into account the coherence of the knowledge structure.

[0057] The knowledge point with the highest replacement utility score is selected as the optimal replacement knowledge point to replace the target knowledge point.

[0058] The future cognitive load equivalent is recalculated based on the replaced sequence to be evaluated, and the predicted load conflict is determined again.

[0059] If the predicted load conflict still exists, repeat the above steps until the predicted load conflict is eliminated, or all knowledge points in the sequence to be evaluated have been replaced.

[0060] The challenge instructions include: starting from the beginning of the sequence to be evaluated, sequentially accumulating the cognitive difficulty gradient of each knowledge point, and calculating the sum of the accumulated value and the current cognitive load equivalent in real time. The first knowledge point that causes the accumulated sum to exceed the upper limit of the expected load interval is marked as a knowledge point of concern, to ensure that the intervention is located at the starting node that triggers the prediction conflict.

[0061] Based on the attributes of the knowledge point of interest, such as conceptual or computational, at least one comprehension-aid resource is selected and bound to the knowledge point of interest. The comprehension-aid resource includes step-by-step analysis, example demonstration, or analogy explanation. As an example, if the knowledge point of interest is conceptual, then analogy explanation or example demonstration resources are preferentially matched and bound, aiming to promote the understanding of abstract concepts through concrete examples or vivid analogies.

[0062] If the knowledge point you are interested in is computational, prioritize matching and binding step-by-step analysis resources, aiming to reduce cognitive load by breaking down the operation process or logical steps.

[0063] It should be noted that this mapping relationship is not an absolute one-to-one correspondence. The specific complexity of a knowledge point can be judged based on whether the number of its direct preorder dependencies in the knowledge point network exceeds the number of direct preorder dependencies of 80% of the knowledge points in the network. This allows for the binding of various types of combined resources to provide multi-faceted support.

[0064] Configure immediate positive feedback rules for the key knowledge points. These rules are defined as providing corresponding levels of incentive feedback when students achieve a comprehension goal while learning the knowledge point. Specifically, the conceptual comprehension goal can be set as correctly reciting key concepts, and the computational comprehension goal can be set as completing a basic variation exercise.

[0065] The incentive feedback can be exemplified by visual encouragement signs or progress point rewards.

[0066] Generate an independent learning task module that integrates the aforementioned knowledge points of interest, binding auxiliary understanding resources, and incentive rules, as the challenge scheme.

[0067] The proposed solution is inserted before the original learning order of the knowledge points to be tackled in the current learning path and then executed.

[0068] If the frequency of student requests for help or the number of consecutive incorrect answers exceeds the preset number during the execution of the challenge program, the program will be exited and a low-cognitive-load knowledge point replacement instruction will be triggered for the sequence to be evaluated.

[0069] S5. Collect the learning efficiency and interaction feedback data after the execution of the control command, generate a feedback data package and use it to optimize the triggering logic of the control command.

[0070] The generation of the feedback data packet is mainly to provide a structured and traceable empirical data basis for subsequent optimization. Specifically, it includes: collecting the average correct answer rate and average time spent per question of students on subsequent related learning tasks within a predetermined time window after the execution of the control instruction, thus forming short-term efficiency data.

[0071] Record students' acceptance or rejection of changes in learning content or path resulting from the execution of the aforementioned control instructions, as feedback on compliance.

[0072] The identification information of the control command, the context state at the time of triggering, the short-term efficiency data, and the compliance feedback are associated and encapsulated to form a feedback data packet.

[0073] The triggering logic of the optimized control command includes: based on the identification information in the feedback data packet, backtracking and extracting the context state when the corresponding control command was triggered.

[0074] For each type of control instruction, the following targeted optimization is performed: When the control instruction is a rest instruction, the average correct answer rate and average time per question in the short-term efficiency data are compared with the student's historical baseline level. The historical baseline level refers to the student's average correct answer rate and average time per question on the same knowledge point over the past 30 days. If the student's average correct answer rate is lower than a certain percentage of their personal historical baseline level, such as lower than 80% of the baseline value, it is determined that the student's recovery is insufficient, and the preset time adjustment coefficient needs to be adjusted upward by a preset step size to increase the rest time compensation.

[0075] If a student's average accuracy rate is significantly higher than their historical baseline (e.g., above 115% of the baseline) and accompanied by an abnormal increase in average time spent answering questions, it indicates possible excessive delay or inattention, suggesting over-recovery. In this case, the preset time adjustment coefficient should be adjusted downwards by a preset step size to reduce rest time compensation. All adjustments to the preset time adjustment coefficient are made within preset safety boundaries.

[0076] When the control instruction is a knowledge point replacement instruction with low cognitive load, the weight allocation used to calculate the replacement utility score is dynamically updated based on the compliance feedback and short-term efficiency data in the feedback data packet. Specifically, historical successful control cases are retrieved as positive samples, and the gradient descent optimization algorithm is used to find the weight combination that can better filter out such samples in the scoring function, and then updated.

[0077] When the control command is a challenge command, if the compliance feedback in the feedback data packet shows rejection, it indicates that the incentive feedback has not achieved the expected effect. The trigger intensity of the reward in the instant positive feedback rule is adjusted. If the student's average correct answer rate is lower than a certain percentage of their personal historical baseline level, the type of auxiliary understanding resource bound to this type of knowledge point is increased.

[0078] This invention, based on the type and magnitude of real-time or predicted load conflicts, differentiates the triggering of rest commands, low-cognitive-load knowledge point replacement commands, or challenge commands. This strategy selection based on the nature and severity of the conflict replaces the one-size-fits-all forced interruption of existing technologies. Furthermore, by collecting feedback data packets after command execution and using them to optimize the triggering logic, the invention possesses the ability to self-adjust based on actual intervention effects, thereby improving the effectiveness and adaptability of the control strategy.

[0079] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. An adaptive learning duration control method based on learning behavior, characterized in that, include: By querying the knowledge point network that stores the prior dependencies of knowledge points, the current cognitive load equivalent and future cognitive load equivalent of the student's current learning session can be determined. The process for determining the current cognitive load equivalent includes: determining the set of knowledge points that students have mastered based on their historical answer data; For each knowledge point in the sequence of knowledge points already learned in the student's current learning session, obtain the preorder dependencies of that knowledge point from the knowledge point network; If the knowledge point belongs to the set of knowledge points that the student has mastered, then the cognitive distance of the knowledge point is set to 0. Otherwise, starting from the knowledge point, the knowledge point network is traversed in reverse along the preorder dependency relationship until the first preorder knowledge point that belongs to the set of knowledge points that the student has mastered is encountered. The number of dependency relationship edges traversed is recorded as the cognitive distance of the knowledge point. The shortest prior dependency path length required to master a knowledge point from zero knowledge level will be used as the reference cognitive distance for that knowledge point. The cognitive difficulty gradient of the knowledge point is determined based on the ratio of the cognitive distance to the reference cognitive distance. The cognitive difficulty gradients of all learned knowledge points in the current learning session are summed to obtain the current cognitive load equivalent. The process of determining the future cognitive load equivalent includes: selecting a predetermined number of knowledge points from the sequence of knowledge points that have been planned but not yet learned in the current learning session to form a sequence to be evaluated; Determine the cognitive difficulty gradient of each knowledge point in the sequence to be evaluated, and use the sum of the cognitive difficulty gradients of all knowledge points in the sequence as the future cognitive load increment; The sum of the future cognitive load increment and the current cognitive load equivalent is taken as the future cognitive load equivalent; The control rules configured by parents are transformed into a range that includes the expected workload range, the basic duration of a single rest, and the selection range of knowledge points with low cognitive load. The current cognitive load equivalent and the future cognitive load equivalent are compared with the expected load range to identify immediate load conflicts or predicted load conflicts. If a conflict is detected, the corresponding adjustment instruction that integrates the basic duration of the single rest or the range of low cognitive load knowledge points is triggered and executed based on the conflict type and magnitude to adjust the learning process. Control commands triggered based on conflict type and conflict amplitude include: If an immediate load conflict exists, a rest instruction is triggered. The difference between the current perceived load equivalent and the upper limit of the expected load range is defined as the conflict amplitude. The product of the conflict amplitude and the preset duration adjustment coefficient is added to the basic duration of a single rest to obtain the specific execution duration of the rest instruction. If a predicted load conflict exists, the difference between the future cognitive load equivalent and the upper limit of the expected load range is defined as the conflict amplitude. If the conflict amplitude is greater than or equal to the path adjustment threshold set by the subject to which the current learning session belongs, a low cognitive load knowledge point replacement instruction is triggered for the sequence to be evaluated; otherwise, a challenge instruction is triggered. Collect learning efficiency and interaction feedback data after the execution of the control commands, generate feedback data packets, and use them to optimize the triggering logic of the control commands.

2. The adaptive learning duration control method based on learning behavior according to claim 1, characterized in that, The process of determining the set of knowledge points already mastered includes: Extract any knowledge point from the knowledge point network and mark it as the target knowledge point. Filter all answer data related to the target knowledge point from the students' historical answer data and calculate the historical answer accuracy rate of the target knowledge point. The historical answer accuracy rate of the target knowledge point is compared with the preset mastery judgment threshold. If the historical answer accuracy rate is greater than or equal to the preset mastery judgment threshold, the target knowledge point is determined to be a mastered knowledge point; otherwise, it is determined not to be a mastered knowledge point. All mastered knowledge points are regularly organized according to a fixed cycle to form a set of mastered knowledge points.

3. The adaptive learning duration control method based on learning behavior according to claim 1, characterized in that, The identification of immediate load conflicts or prediction of load conflicts includes: If the current cognitive load equivalent exceeds the upper limit of the expected load range, then an immediate load conflict is determined to exist; If the future perceived load equivalent exceeds the upper limit of the expected load range, then a predicted load conflict is identified.

4. The adaptive learning duration control method based on learning behavior according to claim 1, characterized in that, The instructions for replacing low-cognitive-load knowledge points in the sequence to be evaluated include: From the sequence to be evaluated, the knowledge point with the highest cognitive difficulty gradient is selected as the target knowledge point; The knowledge point network is queried to obtain all adjacent knowledge points that have a direct dependency relationship with the target knowledge point. Among them, knowledge points whose cognitive difficulty gradient is lower than that of the target knowledge point and whose value is within the selection range of low cognitive load knowledge points are selected to form a set of candidate replacement knowledge points. For each knowledge point in the candidate replacement knowledge point set, a replacement utility score is calculated by linear weighting based on the difference in cognitive load gradient that it can reduce after replacing the target knowledge point, and its semantic relevance with the other knowledge points in the sequence to be evaluated. The knowledge point with the highest replacement utility score is selected as the optimal replacement knowledge point, and the target knowledge point is replaced. The future cognitive load equivalent is recalculated based on the replaced sequence to be evaluated, and the predicted load conflict is determined again. If the predicted load conflict still exists, the operation of selecting target knowledge points from the sequence to be evaluated, forming a set of candidate replacement knowledge points, calculating the replacement utility score, and selecting the optimal replacement knowledge point to replace the target knowledge point is repeated until the predicted load conflict is eliminated, or all knowledge points in the sequence to be evaluated have been traversed by the replacement operation.

5. The adaptive learning duration control method based on learning behavior according to claim 4, characterized in that, The challenge instructions include: Identify the first knowledge point in the sequence to be evaluated whose cognitive difficulty gradient accumulation causes the future cognitive load equivalent to exceed the upper limit of the expected load range, and mark it as a knowledge point of concern; Select at least one auxiliary comprehension resource and bind it to the knowledge point of concern. The auxiliary comprehension resource includes step-by-step analysis, example demonstration or analogy explanation. Configure immediate positive feedback rules for the knowledge points that students are interested in. The immediate positive feedback rules are defined as providing corresponding levels of incentive feedback when students achieve the understanding goal in the process of learning the knowledge points. Generate an independent learning task module that integrates the aforementioned knowledge points of interest, binding auxiliary understanding resources, and incentive rules, as a challenge scheme; The proposed solution is inserted before the original learning order of the knowledge points to be explored in the current learning path and then executed. If the frequency of student requests for help or the number of consecutive incorrect answers exceeds the preset number during the execution of the challenge program, the program will be exited and a low-cognitive-load knowledge point replacement instruction will be triggered for the sequence to be evaluated.

6. The adaptive learning duration control method based on learning behavior according to claim 5, characterized in that, The feedback data packet generation process includes: Within a predetermined time window after the execution of the control command, the average correct answer rate and average time spent per question of students on subsequent related learning tasks are collected as short-term efficiency data. Record students' acceptance or rejection of changes in learning content or path resulting from the execution of the aforementioned control instructions, as feedback on compliance; The identification information of the control command, the context state at the time of triggering, the short-term efficiency data, and the compliance feedback are associated and encapsulated to form a feedback data packet.

7. The adaptive learning duration control method based on learning behavior according to claim 6, characterized in that, The triggering logic for the optimized control command includes: Based on the identification information in the feedback data packet, the context state when the corresponding control command was triggered is retrieved back; If the control command is a rest command, the preset duration adjustment coefficient is adaptively adjusted based on the short-term efficiency data in the feedback data packet. If the control instruction is a low cognitive load knowledge point replacement instruction, then the weight allocation used in calculating the replacement utility score is dynamically updated based on the compliance feedback and short-term efficiency data in the feedback data packet. If the control command is a challenge command, then the binding type of the auxiliary understanding resource or the strength of the instant positive feedback rule will be dynamically adjusted based on the short-term efficiency data and compliance feedback in the feedback data packet.

Images