Psychological experiment task-oriented subject response data modeling method and system

Abstract

The invention discloses a psychological experiment task-oriented subject response data modeling method and system, and belongs to the technical field of psychological experiment task data modeling, and the method comprises the following steps: collecting subject response data; performing reaction flow serialization, converting original data into a continuous time sequence, and recognizing a cognitive state transition point through a multi-scale recursive segmentation method based on dynamic entropy perception; dynamic cognitive model construction: adopting an improved drift diffusion model combining dynamic parameter decomposition and hierarchical Bayesian inference to obtain a dynamic cognitive analysis model reflecting cognitive strategy switching, fatigue accumulation and learning effect; and reaction data modeling: establishing a causal relationship among a state label, a cognitive parameter and a reaction behavior, and realizing quantitative modeling and prediction of a psychological experiment task. According to the scheme, fine modeling and dynamic analysis of the multi-dimensional psychological reaction process can be realized, and the interpretability and the application value of psychological experiment data are improved.

Description

Technical Field

[0001] This invention relates to the field of data modeling technology for psychological experimental tasks, specifically to a method and system for modeling subject response data for psychological experimental tasks. Background Technology

[0002] The method and system for modeling subject response data for psychological experiments refers to the process of collecting multimodal response data (such as key presses, mouse movements, eye movements, heart rate, etc.) from subjects during psychological experiments, performing time synchronization and serialization processing on the data, and then constructing a computational model that can characterize the dynamic changes in cognitive state with the task and time. This enables quantitative analysis and causal modeling of psychological mechanisms such as cognitive strategy switching, fatigue accumulation, and learning effects. The role of this method and system is to transform scattered behavioral and physiological responses in psychological experiments into structured, interpretable cognitive parameters and model results. This not only helps improve the depth and accuracy of scientific analysis of experimental data but also provides objective reference and technical support for psychological research, clinical diagnosis, and cognitive training applications.

[0003] However, existing methods for modeling subject response data for psychological experimental tasks generally only collect single-dimensional behavioral responses (such as simple reaction time or accuracy), which makes it difficult to reflect the multimodal procedural characteristics of subjects in complex psychological experimental tasks.

[0004] In existing reaction flow serialization methods, there is a common problem that single time scale segmentation cannot identify the hierarchy of cognitive state transitions, resulting in insufficient accuracy in detecting fatigue and learning effects across trials in tasks such as Stroop and N-Back.

[0005] In existing methods for constructing dynamic cognitive models, the parameters of the drift-diffusion model used for cognitive modeling are usually set to static values, making it difficult to reflect the dynamic changes of an individual's cognitive strategy over time under multiple trial conditions. In particular, in Go / No-Go inhibition control tasks, it is difficult to reveal the phenomenon of gradually increasing thresholds due to fatigue. Summary of the Invention

[0006] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a method and system for modeling subject response data for psychological experimental tasks. Addressing the technical problem that existing methods for modeling subject response data for psychological experimental tasks typically only collect single-dimensional behavioral responses (such as simple reaction time or accuracy), making it difficult to reflect the multimodal procedural characteristics of subjects in complex psychological experimental tasks, this solution creatively employs a comprehensive response data modeling method combining response stream serialization enhancement and dynamic cognitive modeling. This method can simultaneously integrate multimodal trajectories such as key presses, mouse movements, eye movements, and heart rate, and achieve time synchronization, thereby accurately capturing the dynamic evolution of cognitive states in real experimental scenarios and achieving a more comprehensive and fine-grained modeling of the subject's cognitive process. Furthermore, addressing the technical problem that existing response stream serialization methods often fail to identify the hierarchical nature of cognitive state transitions due to single-timescale segmentation, leading to insufficient accuracy in detecting fatigue and learning effects across trials in tasks such as Stroop and N-Back, this solution creatively constructs a multidimensional response stream time series and, based on dynamic entropy perception, multi-dimensional response stream time series. A scale-based recursive response flow sequence segmentation method is used for data analysis and automatic segmentation. By recursively detecting abrupt change points at different scales such as second, minute, and trial levels, it achieves multi-level automatic segmentation of cognitive strategy switching, fatigue accumulation, and learning effects, thereby effectively improving the accuracy and robustness of state transition identification in complex experimental tasks. Addressing the technical problem in existing dynamic cognitive model construction methods where drift-diffusion model parameters are typically set to static values, making it difficult to reflect the dynamic changes in individual cognitive strategies over time under multiple trial conditions, especially in Go / No-Go inhibition control tasks where it fails to reveal the phenomenon of gradually increasing thresholds due to fatigue, this solution creatively adopts an improved drift-diffusion cognitive model combining dynamic parameter decomposition and hierarchical Bayesian inference. This model defines the drift rate and decision threshold as dynamic variables that evolve with each trial, and uses a two-level modeling framework at the group and individual levels for inference, achieving individualized dynamic modeling of the subject's cognitive process under different experimental conditions, effectively improving the explanatory power for cognitive strategy switching, fatigue accumulation, and learning effects.

[0007] The technical solution adopted by this invention is as follows: This invention provides a method for modeling subject response data for psychological experimental tasks, which includes the following steps:

[0008] Step S1: Data collection of subject responses;

[0009] Step S2: Reaction flow serialization;

[0010] Step S3: Construction of a dynamic cognitive model;

[0011] Step S4: Reaction data modeling.

[0012] Furthermore, in step S1, the subject response data acquisition is used to collect multi-dimensional, high-granular procedural data of the psychological experiment task. Specifically, it involves synchronously collecting the discrete behavioral events and continuous behavioral trajectories of the subject in the psychological experiment task, and synchronizing them in time through a programming interface to obtain a multimodal raw dataset.

[0013] Further, in step S2, the response stream serialization is used to transform the raw data into a continuous time series and identify the critical points of transition in the subject's internal cognitive state. Specifically, the multimodal raw dataset is arranged in chronological order of trials, and secondary features are extracted to construct a multidimensional response stream time series. Using a multi-scale recursive response stream sequence segmentation method based on dynamic entropy perception, the multidimensional response stream time series is analyzed and automatically segmented to identify the time points representing cognitive strategy switching, fatigue accumulation, and learning effects, thus obtaining state transition labeled response stream sequence data. This includes the following steps:

[0014] Step S21: Multidimensional reaction flow feature construction, specifically, extracting reaction features from the multimodal original dataset and aligning and standardizing the features to obtain a multidimensional reaction flow feature matrix;

[0015] The response characteristics specifically include trial number, stimulus type, task type, reaction time, accuracy, eye movement fixation characteristics, mouse displacement characteristics, and heart rate characteristics.

[0016] Step S22: Dynamic entropy perception calculation optimization, specifically by calculating the change in feature information within the time window and performing significance analysis, detecting state change points, and obtaining a set of candidate segmentation points;

[0017] The formula for calculating the change in the feature information is:

[0018] ;

[0019] In the formula, It is the change in feature information, H t H is the comprehensive information entropy within the t-th time window, where t is the time window index. t-1 It is the comprehensive information entropy within the (t-1)th time window, d is the total number of feature dimensions, i is the feature dimension index, and p t,i It represents the probability of the value of the i-th feature dimension within the t-th time window;

[0020] Step S23: Multi-scale recursive segmentation, specifically by recursively calculating significance and performing segmentation at second, minute and trial time scales, screening candidate points and recursively segmenting to obtain multi-level segmentation results;

[0021] Step S24: State transition labeling, specifically by combining segment frequency, feature change trend and experimental event information to automatically generate labels that represent cognitive strategy switching, fatigue accumulation and learning effect;

[0022] Step S25: Reaction flow serialization, specifically by integrating segment boundaries, time sequence and automatically generated state labels, and outputting in a unified format to obtain state transition labeled reaction flow sequence data.

[0023] Further, in step S3, the dynamic cognitive model construction is used to establish an individualized computational model that can reflect the evolution of cognitive parameters of subjects in psychological experimental tasks over time and with internal states. Specifically, based on the state transition labeled response flow sequence data, an improved drift-diffusion cognitive model combining dynamic parameter decomposition and hierarchical Bayesian inference is constructed. The drift rate and decision threshold parameters are defined as dynamic variables that can evolve with each trial. A two-level modeling mechanism at the group and individual levels is introduced. The Markov chain Monte Carlo sampling method is used to estimate the parameters of the model, and the posterior cognitive parameter distribution of each subject in each time period is output. A dynamic cognitive analysis model that can be used to analyze cognitive strategy switching, fatigue accumulation, and learning effects is obtained, including the following steps:

[0024] Step S31: Model structure initialization, used to clarify the model framework. Specifically, by introducing dynamic parameter decomposition into the drift diffusion model, the drift rate and decision threshold are defined as dynamic variables that change with the number of trials, thus obtaining the dynamic cognitive model structure.

[0025] The calculation formula for the dynamic parameter decomposition is as follows:

[0026] ;

[0027] In the formula, v c is the drift rate parameter of the subject in the c-th trial. The calculation object is the estimated rate of evidence accumulation in the response data of that trial. v0 is the subject's baseline drift rate value. It is the drift rate offset of the subject relative to the baseline in the c-th trial, a c is the decision threshold parameter for the subject in the c-th trial. The calculation object is the estimated value of the decision boundary in the response data of this trial, and a0 is the baseline value of the subject's decision threshold. It is the threshold offset of the c-th trial relative to the baseline;

[0028] Step S32: Set up a hierarchical modeling mechanism to take into account both group-level patterns and individual differences. Specifically, by setting the prior distribution of the group level and combining it with the experimental data of the individual level, a hierarchical Bayesian modeling framework is established to obtain the two-level parameter mapping relationship between the group and the individual.

[0029] Step S33: Dynamic parameter driving and correction, which is used to enable the model to reflect the real cognitive process. Specifically, by combining the state transition labeled response flow sequence, the direction of parameter change is driven and corrected under the conditions of cognitive strategy switching, fatigue accumulation and learning effect, so as to obtain the dynamic evolution parameter trajectory.

[0030] Step S34: Parameter inference estimation, used to obtain the confidence interval of dynamic parameters. Specifically, the model parameters are estimated by posterior distribution using the Markov chain Monte Carlo sampling method to obtain the dynamic cognitive parameter estimation results for each subject in each data segment and each trial.

[0031] Step S35: Dynamic cognitive analysis model generation, used to output interpretable cognitive analysis results. Specifically, by integrating parameter trajectories and experimental condition information, a dynamic cognitive analysis model that can reveal cognitive strategy switching, fatigue accumulation, and learning effects is generated.

[0032] Further, in step S4, the response data modeling is used to realize causal modeling of the subject's response data. Specifically, based on the dynamic cognitive analysis model, the state transition labeled response flow sequence is combined with the corresponding dynamic cognitive parameters to establish the causal modeling relationship of the subject's response; by comparing the parameter changes and response behavior differences under different state labels, the causal effects of cognitive strategy switching, fatigue accumulation, and learning effects on the subject's response behavior are analyzed; and combined with the individualized parameter estimation results of hierarchical Bayesian inference, the group level pattern and individual level difference are distinguished to obtain reference data for subject response modeling analysis that can characterize the causal relationship of the psychological experimental task.

[0033] The reference data for the modeling and analysis of the subjects' responses specifically includes: dynamic cognitive parameter trajectory data, state segmentation causal effect indicators, group and individual stratification results data, and response behavior prediction and fitting results.

[0034] The present invention provides a subject response data modeling system for psychological experimental tasks, including a data acquisition module, a response flow serialization module, a cognitive model construction module, and a response data modeling module;

[0035] The data acquisition module is used to collect subject response data, obtain a multimodal raw dataset through subject response data collection, and send the multimodal raw dataset to the response stream serialization module;

[0036] The reaction stream serialization module is used for reaction stream serialization. Through reaction stream serialization, state transition labeled reaction stream sequence data is obtained, and the state transition labeled reaction stream sequence data is sent to the cognitive model construction module.

[0037] The cognitive model construction module is used for dynamic cognitive model construction. Through dynamic cognitive model construction, a dynamic cognitive analysis model is obtained, and the dynamic cognitive analysis model is used in the reaction data modeling module.

[0038] The reaction data modeling module is used for reaction data modeling, and through reaction data modeling, reference data for subject reaction modeling analysis is obtained.

[0039] The beneficial effects achieved by the present invention using the above solution are as follows:

[0040] (1) In view of the technical problem that existing methods for modeling subject response data for psychological experimental tasks generally only collect single-dimensional behavioral responses (such as simple reaction time or accuracy), which are difficult to reflect the multimodal procedural characteristics of subjects in complex psychological experimental tasks, this solution creatively adopts a comprehensive response data modeling method that combines response flow serialization enhancement and dynamic cognitive modeling. It can simultaneously integrate multimodal trajectories such as key presses, mouse, eye movements and heart rate, and achieve time synchronization, thereby accurately capturing the dynamic evolution of cognitive state in real experimental scenarios and realizing a more comprehensive and finer-grained modeling of the subject's cognitive process.

[0041] (2) In response to the common problem in existing reaction flow serialization methods that the single time scale segmentation cannot identify the hierarchical nature of cognitive state transitions, resulting in insufficient accuracy in detecting fatigue and learning effects across trials in tasks such as Stroop and N-Back, this solution creatively adopts the construction of a multi-dimensional reaction flow time series and a multi-scale recursive reaction flow sequence segmentation method based on dynamic entropy perception for data analysis and automatic segmentation. By recursively detecting mutation points at different scales such as second, minute and trial levels, it realizes multi-level automatic segmentation of cognitive strategy switching, fatigue accumulation and learning effects, thereby effectively improving the accuracy and robustness of state transition identification in complex experimental tasks.

[0042] (3) In response to the technical problem that in existing methods for constructing dynamic cognitive models, the parameters of the drift-diffusion model used for cognitive modeling are usually set to static values, which makes it difficult to reflect the dynamic changes of an individual's cognitive strategy over time under multiple trials, especially in the Go / No-Go inhibition control task, it is impossible to reveal the phenomenon of gradually increasing threshold due to fatigue. This solution creatively adopts an improved drift-diffusion cognitive model that combines dynamic parameter decomposition and hierarchical Bayesian inference to construct a dynamic cognitive model. The drift rate and decision threshold are defined as dynamic variables that can evolve with each trial. Inference is performed through a two-level modeling framework at the group level and the individual level, which realizes individualized dynamic modeling of the cognitive process of subjects under different experimental conditions and effectively improves the ability to explain cognitive strategy switching, fatigue accumulation and learning effects. Attached Figure Description

[0043] Figure 1 A flowchart illustrating a method for modeling subject response data for psychological experimental tasks provided by this invention;

[0044] Figure 2 A schematic diagram of a subject response data modeling system for psychological experimental tasks provided by the present invention;

[0045] Figure 3 This is a schematic diagram of the reaction flow serialization process in step S2;

[0046] Figure 4 A flowchart illustrating the process of constructing the dynamic cognitive model in step S3.

[0047] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0050] Example 1, see Figure 1 This invention provides a method for modeling subject response data for psychological experimental tasks, which includes the following steps:

[0051] Step S1: Data collection of subject responses;

[0052] Step S2: Reaction flow serialization;

[0053] Step S3: Construction of a dynamic cognitive model;

[0054] Step S4: Reaction data modeling.

[0055] By performing the above operations, this solution addresses the technical problem that existing methods for modeling subject response data for psychological experimental tasks generally only collect single-dimensional behavioral responses (such as simple reaction time or accuracy), making it difficult to reflect the multimodal procedural characteristics of subjects in complex psychological experimental tasks. This solution creatively adopts a comprehensive response data modeling method that combines response flow serialization enhancement and dynamic cognitive modeling. It can simultaneously integrate multimodal trajectories such as key presses, mouse movements, eye movements, and heart rate, and achieve time synchronization. This allows for the accurate capture of the dynamic evolution of cognitive states in real experimental scenarios, achieving a more comprehensive and granular modeling of the subject's cognitive process.

[0056] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S1, the subject response data collection is used to collect multi-dimensional, high-granularity procedural data of the psychological experiment task. Specifically, it involves synchronously collecting the discrete behavioral events and continuous behavioral trajectories of the subject in the psychological experiment task, and performing time synchronization through a programming interface to obtain a multimodal raw dataset.

[0057] The psychological experimental tasks specifically include a choice response task, a Stroop color word task, an N-Back working memory task, and a Go / No-Go inhibitory control task;

[0058] The discrete behavioral events specifically include key press events, key release events, task start events, task end events, and error feedback trigger events; wherein: the key press event includes key value encoding and press time;

[0059] The continuous behavioral trajectory specifically includes mouse movement trajectory, eye movement trajectory, and heart rate change trajectory; wherein: the mouse movement trajectory is represented by a two-dimensional coordinate sequence corresponding to a timestamp; the eye movement trajectory includes the two-dimensional coordinates of the fixation point, fixation duration, and saccade path; and the heart rate change trajectory records electrocardiogram and pulse wave data using a sampling time series.

[0060] Preferably, Table 1 is an example table of data fields for the multimodal raw dataset. As shown in the table, the multimodal raw data includes discrete behavioral event data and continuous behavioral trajectory data. The discrete behavioral event data mainly includes key events, task control events, and error feedback events, which are used to characterize the explicit operation and task response process of the subjects in a single trial. The continuous behavioral trajectory data includes mouse trajectory, eye movement trajectory, and heart rate change trajectory, which are used to characterize the subjects' action characteristics, visual attention, and physiological responses during task execution.

[0061] Table 1. Example of data fields in the original multimodal dataset.

[0062]

[0063] The programming interface specifically includes high-precision clock synchronization, software event trigger synchronization, hardware interface synchronization, and buffer queue synchronization. Specifically: the high-precision clock synchronization timestamps all data streams using the global clock signal of the experimental computer system; the software event trigger synchronization achieves unified recording of stimulus presentation and response acquisition through experimental control software event triggers; the hardware interface synchronization aligns EEG data, eye-tracking data, and behavioral data through parallel or serial port trigger signals; and the buffer queue synchronization ensures that all modal data are stored in timestamp order through a unified queue mechanism during the data writing process.

[0064] Example 3, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the response flow serialization is used to transform the raw data into a continuous time series and identify the critical points of transition in the subject's internal cognitive state. Specifically, the multimodal raw dataset is arranged in chronological order of trials, and secondary features are extracted to construct a multidimensional response flow time series. The multidimensional response flow time series is then analyzed and automatically segmented using a multi-scale recursive response flow sequence segmentation method based on dynamic entropy perception to identify the time points representing cognitive strategy switching, fatigue accumulation, and learning effects, thus obtaining state transition labeled response flow sequence data. This includes the following steps:

[0065] Step S21: Multidimensional reaction flow feature construction, specifically, extracting reaction features from the multimodal original dataset and aligning and standardizing the features to obtain a multidimensional reaction flow feature matrix;

[0066] The response characteristics specifically include trial number, stimulus type, task type, reaction time, accuracy, eye movement fixation characteristics, mouse displacement characteristics, and heart rate characteristics.

[0067] Preferably, the formula for calculating the multidimensional reactive flow characteristic matrix is:

[0068] ;

[0069] In the formula, F is the multidimensional reactive flow feature matrix, t is the time window index, d is the number of feature dimensions, and T is the number of sliding windows. It is the first-dimensional feature value in the t-th time window;

[0070] Table 2 is an example table of the definition of the feature dimensions. As shown in the table, the feature dimensions are used to clarify the specific composition and correspondence of multimodal features in the response flow matrix. In this embodiment, different feature dimensions correspond to different response channels and signal sources of the subjects in the psychological experimental task. Among them, behavioral features (such as reaction time and correctness) are used to describe the subjects' explicit response performance at the task execution level; physiological features (such as eye movement fixation features and heart rate features) are used to reflect the subjects' implicit physiological changes during the task; interactive features (such as mouse displacement and speed) are used to reveal the subjects' movement trajectory and response fluency during the operation process; and experimental condition features (such as task type and stimulus type) are used to provide contextual constraints and condition identifiers.

[0071] By defining a unified feature dimension, structured fusion of multi-source data can be achieved within a time window, and a unified time series data framework can be established, enabling subsequent dynamic entropy calculation and state segmentation to be completed in the same feature space, thereby improving the time alignment accuracy of multimodal features and the consistency of model input.

[0072] Furthermore, the feature dimensions, during the definition process, can support the expansion of different experimental paradigms (such as Stroop tasks, N-Back tasks, and Go / No-Go tasks). When a certain type of modality feature is added or removed in an experimental task, it is only necessary to expand or hide the corresponding dimension in the feature matrix to achieve the scalable configuration of the model structure, thereby ensuring the universality and portability of the present invention in different psychological experimental environments.

[0073] Table 2 Example of Feature Dimension Definitions

[0074]

[0075] Step S22: Dynamic entropy perception calculation optimization, specifically by calculating the change in feature information within the time window and performing significance analysis, detecting state change points, and obtaining a set of candidate segmentation points;

[0076] The formula for calculating the change in the feature information is:

[0077] ;

[0078] In the formula, It is the change in feature information, H t H is the comprehensive information entropy within the t-th time window, where t is the time window index. t-1 It is the comprehensive information entropy within the (t-1)th time window, d is the total number of feature dimensions, i is the feature dimension index, and p t,i It represents the probability of the value of the i-th feature dimension within the t-th time window;

[0079] Step S23: Multi-scale recursive segmentation, specifically by recursively calculating significance and performing segmentation at second, minute and trial time scales, screening candidate points and recursively segmenting to obtain multi-level segmentation results;

[0080] In a preferred embodiment, to quantify the significance of entropy mutations at different time scales, the significance is calculated by constructing a significance indicator function, the formula for which is:

[0081] ;

[0082] In the formula, It is a significance indicator function, used to indicate whether the k-th time scale is a significant segmentation point at time t. It is the change in feature information at the k-th time scale. It is the mean value of the feature information change at the k-th time scale. It is the standard deviation of the feature information at the k-th time scale. It is the significance threshold. It is an indicator function, and the significance threshold is adaptively determined based on the statistical distribution of historical or baseline data;

[0083] Step S24: State transition labeling, specifically by combining segment frequency, feature change trend and experimental event information to automatically generate labels that represent cognitive strategy switching, fatigue accumulation and learning effect;

[0084] The specific steps for generating the automated tags include:

[0085] Step S241: Feature change trend classification, specifically, based on the multi-level segmentation results, take an equal number of time windows before and after the time window corresponding to the segmentation candidate time, calculate the trend feature index, and perform state classification based on the state judgment logic to obtain the preliminary state label classification result.

[0086] The state determination logic includes cognitive strategy switching determination logic, fatigue accumulation determination logic, and learning effect determination logic; the cognitive strategy switching determination logic specifically refers to a sudden change in the direction of drift rate estimation; the fatigue accumulation determination logic specifically refers to a significant increase in reaction time characteristics and a decrease in fixation concentration characteristics; the learning effect determination logic specifically refers to a significant increase in accuracy characteristics and a decrease in reaction time characteristics.

[0087] Step S242: Experimental event coordination, specifically, based on the preliminary state label classification results, if multiple state label conditions are met within a single time window, then based on the label confidence score, the label with the highest confidence score is selected as the label classification result to obtain the state labels representing cognitive strategy switching, fatigue accumulation, and learning effects;

[0088] Step S25: Reaction flow serialization, specifically by integrating segment boundaries, time sequence and automatically generated state labels, and outputting in a unified format to obtain state transition labeled reaction flow sequence data.

[0089] By performing the above operations, this solution addresses the common technical problem in existing reaction flow serialization methods where single-timescale segmentation fails to identify the hierarchical nature of cognitive state transitions, leading to insufficient accuracy in detecting fatigue and learning effects across trials in tasks such as Stroop and N-Back. This solution creatively constructs a multi-dimensional reaction flow time series and uses a multi-scale recursive reaction flow sequence segmentation method based on dynamic entropy perception for data analysis and automatic segmentation. By recursively detecting abrupt change points at different scales such as second, minute, and trial levels, it achieves multi-level automatic segmentation of cognitive strategy switching, fatigue accumulation, and learning effects, thereby effectively improving the accuracy and robustness of state transition identification in complex experimental tasks.

[0090] Example 4, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S3, the dynamic cognitive model construction is used to establish an individualized computational model that can reflect the evolution of cognitive parameters of subjects in psychological experimental tasks over time and internal state. Specifically, based on the state transition labeled response flow sequence data, an improved drift-diffusion cognitive model combining dynamic parameter decomposition and hierarchical Bayesian inference is constructed. The drift rate and decision threshold parameters are defined as dynamic variables that can evolve with each trial. A two-level modeling mechanism at the group level and the individual level is introduced. The Markov chain Monte Carlo sampling method is used to estimate the parameters of the model, and the posterior cognitive parameter distribution of each subject in each time period is output. A dynamic cognitive analysis model that can be used to analyze cognitive strategy switching, fatigue accumulation and learning effects is obtained, including the following steps:

[0091] Step S31: Model structure initialization, used to clarify the model framework. Specifically, by introducing dynamic parameter decomposition into the drift diffusion model, the drift rate and decision threshold are defined as dynamic variables that change with the number of trials, thus obtaining the dynamic cognitive model structure.

[0092] The calculation formula for the dynamic parameter decomposition is as follows:

[0093] ;

[0094] In the formula, v c is the drift rate parameter of the subject in the c-th trial. The calculation object is the estimated rate of evidence accumulation in the response data of that trial. v0 is the subject's baseline drift rate value. It is the drift rate offset of the subject relative to the baseline in the c-th trial, a c is the decision threshold parameter for the subject in the c-th trial. The calculation object is the estimated value of the decision boundary in the response data of this trial, and a0 is the baseline value of the subject's decision threshold. It is the threshold offset of the c-th trial relative to the baseline;

[0095] Step S32: Set up a hierarchical modeling mechanism to take into account both group-level patterns and individual differences. Specifically, by setting the prior distribution of the group level and combining it with the experimental data of the individual level, a hierarchical Bayesian modeling framework is established to obtain the two-level parameter mapping relationship between the group and the individual.

[0096] The specific steps for setting the hierarchical modeling mechanism include:

[0097] Step S321: Group level setting, specifically, based on the drift rate parameter and drift rate baseline value in the dynamic cognitive model structure, by calculating the average drift rate and average decision threshold parameter of all subjects, a group level parameter set is constructed, and the prior distribution of the group level parameter set is set to a normal distribution;

[0098] Step S322: Within-group standard deviation modeling, specifically, based on the group level parameter set, introduce group level standard deviation parameters and use conjugate prior distributions to construct the within-group standard deviation parameter set;

[0099] Step S323: Individual parameter modeling and mapping, specifically, based on the group level parameter set and the within-group standard deviation parameter set, for each subject individual, the individual drift rate and threshold parameter are respectively distributed normally with the group level parameter set as the mean and the within-group standard deviation parameter set as the standard deviation, to obtain the individual mapping parameter set;

[0100] Step S324: Setting modeling input data, specifically, constructing observation data parameters based on the individual mapping parameter set, the state transition labeled response flow sequence data and the corresponding actual response data, and using the observation data parameters as modeling data input;

[0101] Step S325: Hierarchical modeling output. Specifically, based on the set modeling data input, the hierarchical Bayesian joint modeling stage is entered. By establishing a hierarchical Bayesian generative model, hierarchical modeling output is performed to obtain the two-level parameter mapping relationship between groups and individuals. The calculation formula is as follows:

[0102] ;

[0103] In the formula, These are the joint model objective parameters, used to represent the joint distribution of observed data and model parameters, and are used for parameter inference and estimation. It is the joint prior distribution of a set of level parameters, where, It is the average drift rate parameter of the group level. It is the average decision threshold parameter for the group levels. It is a set of horizontal drift rate standard deviation parameters. It is the standard deviation parameter of the group level decision threshold. This is the per-subject multiplication operator, where u is the subject index. Let be the conditional probability of the individual parameter of the u-th subject relative to the group level parameter, where It is the individual mapping drift rate of the u-th subject. It is the individual mapping decision threshold for the u-th subject. These are probability conditions used to represent the set of group level parameters and the set of within-group standard deviation parameters. It is a trial-by-trial and trial-by-subject multiplication operator. It is the likelihood probability of the observed data based on the state labels. These are the actual observation data from the u-th subject in the c-th trial. It is the drift rate parameter of the u-th subject in the c-th trial. It is the decision threshold parameter for the u-th subject in the c-th trial. It is the status label of the u-th subject in the c-th trial;

[0104] The group and individual two-level parameter mapping relationship specifically includes the group-level parameter set, the within-group standard deviation parameter set, the individual mapping parameter set, and the observation mapping dataset;

[0105] Step S33: Dynamic parameter driving and correction, which is used to enable the model to reflect the real cognitive process. Specifically, by combining the state transition labeled response flow sequence, the direction of parameter change is driven and corrected under the conditions of cognitive strategy switching, fatigue accumulation and learning effect, so as to obtain the dynamic evolution parameter trajectory.

[0106] In a preferred embodiment, to establish a quantitative mapping relationship between psychological states and model parameters, a state-driven weight matrix is ​​introduced, and a dynamic parameter-driven correction equation is constructed. The calculation formula is as follows:

[0107] ;

[0108] In the formula, W is the state-driven weight matrix, used to measure the direction and intensity of the influence of each state on the parameter shift, and is specifically calculated using Bayesian estimation. It is the state label identified as a cognitive strategy switch in the c-th trial. It is the state label identified as fatigue accumulation in the c-th trial. The state label identified as a learning effect in the c-th trial is... It is a random disturbance term that follows a zero-mean Gaussian distribution and is used to reflect non-systematic fluctuations;

[0109] Step S34: Parameter inference estimation, used to obtain the confidence interval of dynamic parameters. Specifically, the model parameters are estimated by posterior distribution using the Markov chain Monte Carlo sampling method to obtain the dynamic cognitive parameter estimation results for each subject in each data segment and each trial.

[0110] Step S35: Dynamic cognitive analysis model generation, used to output interpretable cognitive analysis results. Specifically, by integrating parameter trajectories and experimental condition information, a dynamic cognitive analysis model that can reveal cognitive strategy switching, fatigue accumulation, and learning effects is generated.

[0111] By performing the above operations, this solution addresses the technical problem in existing dynamic cognitive model construction methods where drift-diffusion model parameters are typically set to static values, making it difficult to reflect the dynamic changes in an individual's cognitive strategy over time under multiple trials, especially in Go / No-Go inhibition control tasks where it fails to reveal the phenomenon of gradually increasing thresholds due to fatigue. This solution creatively employs an improved drift-diffusion cognitive model combining dynamic parameter decomposition and hierarchical Bayesian inference to construct a dynamic cognitive model. Drift rate and decision threshold are defined as dynamic variables that evolve with each trial, and inference is performed through a two-tiered modeling framework at both the group and individual levels. This achieves individualized dynamic modeling of the cognitive process of subjects under different experimental conditions, effectively improving the explanatory power for cognitive strategy switching, fatigue accumulation, and learning effects.

[0112] Example 5, see Figure 1 , Figure 2 This embodiment is based on the above embodiment. In step S4, the response data modeling is used to realize causal modeling of the subject's response data. Specifically, based on the dynamic cognitive analysis model, the state transition labeled response flow sequence is combined with the corresponding dynamic cognitive parameters to establish the causal modeling relationship of the subject's response; by comparing the parameter changes and response behavior differences under different state labels, the causal effects of cognitive strategy switching, fatigue accumulation, and learning effects on the subject's response behavior are analyzed; and combined with the individualized parameter estimation results of hierarchical Bayesian inference, the group level pattern and individual level difference are distinguished to obtain reference data for subject response modeling analysis that can characterize the causal relationship of the psychological experimental task.

[0113] The reference data for the modeling and analysis of the subjects' responses specifically includes: dynamic cognitive parameter trajectory data, state segmentation causal effect index, group and individual stratification results data, and response behavior prediction and fitting results;

[0114] The dynamic cognitive parameter trajectory data specifically refers to the drift rate curve, decision threshold curve, and changes in non-decision time for the subjects at different time periods and different trials.

[0115] The state segmentation causal effect index specifically includes the drift rate mutation magnitude under cognitive strategy switching conditions, the threshold rise rate under fatigue accumulation conditions, and the drift rate growth rate under learning effect conditions.

[0116] The group and individual stratification results data specifically include the mean parameter estimates and confidence intervals at the group level, as well as the dynamic parameter posterior distribution at the individual level, used to compare differences among subjects.

[0117] The reaction behavior prediction and fitting results specifically include the evaluation of the model fit for changes in reaction time distribution, accuracy distribution, and error rate, as well as the prediction of future trial reaction trends based on the dynamic model.

[0118] Example 6, see Figure 1 and Figure 2 Based on the above embodiments, this embodiment provides a subject response data modeling system for psychological experimental tasks, including a data acquisition module, a response flow serialization module, a cognitive model construction module, and a response data modeling module.

[0119] The data acquisition module is used to collect subject response data, obtain a multimodal raw dataset through subject response data collection, and send the multimodal raw dataset to the response stream serialization module;

[0120] The reaction stream serialization module is used for reaction stream serialization. Through reaction stream serialization, state transition labeled reaction stream sequence data is obtained, and the state transition labeled reaction stream sequence data is sent to the cognitive model construction module.

[0121] The cognitive model construction module is used for dynamic cognitive model construction. Through dynamic cognitive model construction, a dynamic cognitive analysis model is obtained, and the dynamic cognitive analysis model is used in the reaction data modeling module.

[0122] The reaction data modeling module is used for reaction data modeling, and through reaction data modeling, reference data for subject reaction modeling analysis is obtained.

[0123] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0125] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for modeling subject response data for psychological experimental tasks, characterized in that: The method includes the following steps: Step S1: Collect subject response data to obtain the multimodal raw dataset; Step S2: Response flow serialization. The multimodal raw dataset is arranged in the order of trial time, and secondary features are extracted to construct a multidimensional response flow time series. The multidimensional response flow time series is analyzed and automatically segmented by a multi-scale recursive response flow sequence segmentation method based on dynamic entropy perception. The time points representing cognitive strategy switching, fatigue accumulation and learning effect are identified to obtain state transition labeled response flow sequence data. Step S3: Dynamic cognitive model construction. Based on the state transition labeled response flow sequence data, an improved drift-diffusion cognitive model combining dynamic parameter decomposition and hierarchical Bayesian inference is constructed. The drift rate and decision threshold parameters are defined as dynamic variables that can evolve with each trial. A two-level modeling mechanism at the group level and the individual level is introduced. The Markov chain Monte Carlo sampling method is used to estimate the parameters of the model. The posterior cognitive parameter distribution of each subject in each time period is output to obtain the dynamic cognitive analysis model. Step S4: Response data modeling. Based on the dynamic cognitive analysis model, the state transition labeled response flow sequence is combined with the corresponding dynamic cognitive parameters to establish the causal modeling relationship of the subject's response. Combined with the individualized parameter estimation results of hierarchical Bayesian inference, the group level pattern and individual level difference are distinguished to obtain reference data for subject response modeling analysis that can characterize the causal relationship of the psychological experimental task.

2. The method for modeling subject response data for psychological experimental tasks according to claim 1, characterized in that: In step S1, the subject response data acquisition is used to collect multi-dimensional, high-granular procedural data of the psychological experiment task. Specifically, it involves synchronously collecting the discrete behavioral events and continuous behavioral trajectories of the subject in the psychological experiment task, and synchronizing them in time through a programming interface to obtain a multimodal raw dataset.

3. The method for modeling subject response data for psychological experimental tasks according to claim 2, characterized in that: In step S2, the reaction flow serialization includes the following steps: multidimensional reaction flow feature construction, dynamic entropy-aware computation optimization, multi-scale recursive segmentation, state transition annotation, and reaction flow serialization.

4. The method for modeling subject response data for psychological experimental tasks according to claim 3, characterized in that: The dynamic entropy perception calculation optimization specifically involves calculating the change in feature information within a time window and performing a significance analysis to detect state change points and obtain a set of candidate segmentation points. The formula for calculating the change in the feature information is: ； In the formula, It is the change in feature information, H t H is the comprehensive information entropy within the t-th time window, where t is the time window index. t-1 It is the comprehensive information entropy within the (t-1)th time window, d is the total number of feature dimensions, i is the feature dimension index, and p t,i It represents the probability of the value of the i-th feature dimension within the t-th time window; The multi-scale recursive segmentation specifically involves recursively calculating saliency and performing segmentation at second, minute, and trial time scales to screen candidate points and obtain multi-level segmentation results; the state transition labeling specifically involves automatically generating labels by combining segmentation frequency, feature change trends, and experimental event information to obtain state labels representing cognitive strategy switching, fatigue accumulation, and learning effects.

5. The method for modeling subject response data for psychological experimental tasks according to claim 4, characterized in that: In step S3, the construction of the dynamic cognitive model includes the following steps: model structure initialization, hierarchical modeling mechanism setting, dynamic parameter driving and correction, parameter inference estimation, and dynamic cognitive analysis model generation.

6. The method for modeling subject response data for psychological experimental tasks according to claim 5, characterized in that: The model structure initialization is specifically achieved by introducing dynamic parameter decomposition into the drift-diffusion model, defining the drift rate and decision threshold as dynamic variables that change with the number of trials, thus obtaining a dynamic cognitive model structure. The calculation formula for the dynamic parameter decomposition is as follows: ； In the formula, v c is the drift rate parameter of the subject in the c-th trial. The calculation object is the estimated rate of evidence accumulation in the response data of that trial. v0 is the subject's baseline drift rate value. It is the drift rate offset of the subject relative to the baseline in the c-th trial, a c is the decision threshold parameter for the subject in the c-th trial. The calculation object is the estimated value of the decision boundary in the response data of this trial, and a0 is the baseline value of the subject's decision threshold. It is the threshold offset of the c-th trial relative to the baseline; The hierarchical modeling mechanism is specifically designed by establishing a hierarchical Bayesian modeling framework by setting a prior distribution at the group level and combining it with experimental data at the individual level, thereby obtaining a two-level parameter mapping relationship between groups and individuals. The dynamic parameter driving and correction mechanism is specifically designed by driving and correcting the direction of parameter change under conditions of cognitive strategy switching, fatigue accumulation, and learning effect by combining state transition labeled response flow sequences, thereby obtaining a dynamically evolving parameter trajectory.

7. The method for modeling subject response data for psychological experimental tasks according to claim 6, characterized in that: In step S4, the reference data for subject response modeling analysis specifically includes: dynamic cognitive parameter trajectory data, state segmentation causal effect index, group and individual stratification result data, and response behavior prediction and fitting results.

8. A subject response data modeling system for psychological experimental tasks, used to implement the subject response data modeling method for psychological experimental tasks as described in any one of claims 1-7, characterized in that: It includes a data acquisition module, a reaction flow serialization module, a cognitive model construction module, and a reaction data modeling module.

9. A subject response data modeling system for psychological experimental tasks according to claim 8, characterized in that: The data acquisition module is used to collect subject response data, obtain a multimodal raw dataset through subject response data collection, and send the multimodal raw dataset to the response stream serialization module; The reaction stream serialization module is used for reaction stream serialization. Through reaction stream serialization, state transition labeled reaction stream sequence data is obtained, and the state transition labeled reaction stream sequence data is sent to the cognitive model construction module. The cognitive model construction module is used for dynamic cognitive model construction. Through dynamic cognitive model construction, a dynamic cognitive analysis model is obtained, and the dynamic cognitive analysis model is used in the reaction data modeling module. The reaction data modeling module is used for reaction data modeling, and through reaction data modeling, reference data for subject reaction modeling analysis is obtained.

Images