Assistant psychological interview report generation method and system based on ai conversation and brain wave data

Abstract

This invention discloses a method and system for generating assisted psychological interview reports based on AI dialogue and EEG data, relating to the field of psychological interviewing. The method includes: generating psychological interview questions based on preset interview prompts and a preset question list; acquiring the subject's answers to the psychological interview questions and EEG data during the answering process in real time; determining whether the psychological interview questions have traversed the preset question list; if so, generating a psychological interview dialogue record; assessing the psychological state based on the psychological interview dialogue record and EEG data, obtaining and outputting the psychological state assessment result; if not, adjusting the questioning strategy and answering style of the psychological interview questions based on the EEG data, and generating the next psychological interview question based on it, along with the interview prompts and the preset question list, and repeating the above steps until the psychological interview questions have traversed the preset question list. This invention can quickly generate highly accurate psychological interview reports.

Description

Technical Field

[0001] This invention relates to the field of psychological interview report technology, and in particular to a method and system for generating psychological interview reports based on AI dialogue and EEG data. Background Technology

[0002] In educational settings such as schools, psychological assessments of students through interviews by school counselors are a common method of mental health management. Traditional psychological interviews typically involve one-on-one, face-to-face communication between a counselor and a student, using methods such as questioning and observation to understand the student's psychological state and write an assessment report. This approach is significant in helping students identify and address psychological problems. However, traditional psychological interviews rely heavily on the counselor's personal experience and time commitment, resulting in low efficiency and limited coverage. In recent years, some AI-based psychological assessment tools have begun to be applied to psychological interview scenarios, such as chatbots that analyze student emotions through dialogue. However, these tools often rely solely on verbal data and lack comprehensive analysis of students' physiological signals (such as brainwaves), leading to insufficient depth and accuracy in the assessment results, and failing to fully meet the needs of school counselors in student psychological interviews. Summary of the Invention

[0003] To address the technical problems existing in the background art, this invention proposes a method and system for generating assisted psychological interview reports based on AI dialogue and EEG data.

[0004] In a first aspect, the present invention proposes a method for generating assisted psychological interview reports based on AI dialogue and EEG data, comprising:

[0005] S1. Utilize a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list;

[0006] S2. Send the psychological interview questions to the test subject;

[0007] S3. Real-time acquisition of the EEG data of the test subject's answers to psychological interview questions and the answering process;

[0008] S4. Determine whether the generated psychological interview questions have traversed the preset question list;

[0009] If not, proceed to S5;

[0010] If so, proceed to S6;

[0011] S5. Using a pre-set AI model, adjust the questioning strategy and answering style of the psychological interview questions based on the answers and the EEG data during the answering process; generate the next psychological interview question based on the pre-set interview prompts, the pre-set question list, and the adjusted questioning strategy and answering style of the psychological interview questions, and proceed to S2.

[0012] S6. Using a pre-set AI model, generate a psychological interview dialogue record of the test subject based on the psychological interview questions and answers; conduct a psychological state assessment of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, and obtain the psychological state assessment results of the test subject; output the psychological state assessment results of the test subject in the form of a report.

[0013] Preferably, the AI ​​model used in the psychological state assessment process and the AI ​​model used in the psychological interview process are the same AI model, only the prompt words used are different.

[0014] Preferably, the preset large AI model is fine-tuned using bimodal fine-tuning data.

[0015] Preferably, the bimodal fine-tuning dataset includes multiple fine-tuning samples. Each fine-tuning sample includes psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, and the artificial psychological state assessment results corresponding to the psychological interview questions and answers and the next round of artificial psychological interview questions.

[0016] Preferably, during the bimodal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector representations are combined and input into a preset AI big model, and the AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information.

[0017] Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

[0018] Preferably, a pre-set AI model is used to assess the psychological state of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, obtaining the psychological state assessment result of the test subject, specifically including:

[0019] The brainwave data of the test subject is processed to obtain brainwave feedback information;

[0020] Preprocessing of psychological interview records of the subjects to be tested;

[0021] The brainwave feedback information is converted into cue word adjustment information, and the cue word adjustment information, along with the preprocessed psychological interview dialogue record and the preset psychological state assessment cue words, are input into the preset AI large model to obtain the psychological state assessment results of the test subject.

[0022] Preferably, psychological indicators include depression, anxiety, stress, concentration, meditative state, flow state, and overall mood index.

[0023] Preferably, the psychological state assessment results output in the form of a report include an analysis of the student's psychological state, potential problems, and future recommendations.

[0024] Secondly, this invention also proposes an auxiliary psychological interview report generation system based on AI dialogue and EEG data, comprising:

[0025] The electroencephalogram (EEG) monitoring module is used to acquire EEG data of the subject during psychological interviews and dialogues.

[0026] The digital human dialogue module utilizes a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list, and pushes these questions to the test subject. It acquires real-time data on the test subject's responses to the questions and the resulting EEG data. The module determines whether the generated questions have traversed the pre-set question list. If not, it uses the pre-set AI model to adjust the questioning strategy and response style based on the responses and EEG data. It then generates the next psychological interview question based on the pre-set interview prompts, the pre-set question list, and the adjusted questioning strategy and response style, repeating the process until the question list has been traversed. If yes, it uses the pre-set AI model to generate the psychological interview dialogue record of the test subject based on the questions and responses.

[0027] The data processing and analysis module is used to use a pre-set AI model to assess the psychological state of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, and obtain the psychological state assessment results of the test subject.

[0028] The report generation module is used to output the psychological state assessment results of the test subjects in the form of reports.

[0029] Preferably, the preset AI model in the digital human dialogue module and the data processing and analysis module is the same AI model, only the prompt words used are different.

[0030] Preferably, the pre-defined AI model is fine-tuned in a bimodal manner using bimodal fine-tuning data; wherein, the bimodal fine-tuning dataset includes multiple fine-tuning samples, each fine-tuning sample including psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, and the artificial psychological state assessment results corresponding to the psychological interview questions and answers and the next round of artificial psychological interview questions.

[0031] Preferably, during the bimodal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector tables are combined and input into a preset AI big model, and the AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information.

[0032] Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

[0033] Preferably, a pre-set AI model is used to assess the psychological state of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, obtaining the psychological state assessment result of the test subject, specifically including:

[0034] The brainwave data of the test subject is processed to obtain brainwave feedback information;

[0035] Preprocessing of psychological interview records of the subjects to be tested;

[0036] The brainwave feedback information is converted into cue word adjustment information, and the cue word adjustment information, along with the preprocessed psychological interview dialogue record and the preset psychological state assessment cue words, are input into the preset AI large model to obtain the psychological state assessment results of the test subject.

[0037] The proposed method and system for generating assisted psychological interview reports based on AI dialogue and EEG data can quickly generate highly accurate psychological interview reports by combining AI dialogue and EEG data, thereby reducing the workload of psychological counselors and improving the accuracy and coverage of psychological interviews. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the structure of an AI-based dialogue and EEG data-assisted psychological interview report generation system in one embodiment of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] In a first aspect, the present invention proposes a method for generating assisted psychological interview reports based on AI dialogue and EEG data, comprising:

[0041] S1. Utilize a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list;

[0042] S2. Send the psychological interview questions to the test subject;

[0043] S3. Real-time acquisition of the EEG data of the test subject's answers to psychological interview questions and the answering process;

[0044] S4. Determine whether the generated psychological interview questions have traversed the preset question list;

[0045] If not, proceed to S5;

[0046] If so, proceed to S6;

[0047] S5. Using a pre-set AI model, adjust the questioning strategy and answering style of the psychological interview questions based on the answers and the EEG data during the answering process; generate the next psychological interview question based on the pre-set interview prompts, the pre-set question list, and the adjusted questioning strategy and answering style of the psychological interview questions, and proceed to S2.

[0048] S6. Using a pre-set AI model, generate a psychological interview dialogue record of the test subject based on the psychological interview questions and answers; conduct a psychological state assessment of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, and obtain the psychological state assessment results of the test subject; output the psychological state assessment results of the test subject in the form of a report.

[0049] This invention combines AI dialogue and EEG data to quickly generate highly accurate psychological interview reports, reducing the workload of school counselors and improving the accuracy and coverage of psychological interviews.

[0050] It is important to understand that the AI ​​model used in the psychological state assessment process and the AI ​​model used in the psychological interview process are the same AI model, only the prompts are different. This ensures the consistency between the process and indicators of psychological state assessment based on the responses to each psychological interview question and the process and indicators of psychological state assessment in generating the next psychological interview question, thereby improving the accuracy of the psychological interview.

[0051] During the psychological interview, a pre-set AI model is used to interact with the interviewee in real time through voice, text, and video based on a pre-set list of questions. This guides the interviewee to express their psychological state and simulates a psychological interview scenario to ensure the interview is targeted and smooth.

[0052] In this embodiment, the preset large AI model is fine-tuned using bimodal fine-tuning data.

[0053] The bimodal fine-tuning dataset is constructed based on psychological interview dialogue records, cue word modulation information obtained from EEG data during psychological interview dialogues, artificial psychological interview dialogue records, and psychological state assessment results determined by artificial psychological interviews.

[0054] The bimodal fine-tuning dataset in this embodiment includes multiple fine-tuning samples. Each fine-tuning sample includes psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, and the artificial psychological state assessment results corresponding to the psychological interview questions and answers and the next round of artificial psychological interview questions.

[0055] This embodiment guides the AI ​​model to learn to dynamically adjust its questioning strategies and answering styles based on EEG feedback during the bimodal fine-tuning process by fine-tuning the structure of the samples. It then generates the next psychological interview questions based on the questioning methods and response strategies under different psychological states, thereby achieving the ability to provide emotional empathy and adaptive guidance during the psychological interview process.

[0056] It is important to understand that the questions for the next round of human psychological interviews will be standardized psychological interview texts designed by psychological counselors or psychology teachers, covering typical psychological scenarios such as learning pressure, interpersonal relationships, emotion regulation, and self-awareness.

[0057] The results of the artificial psychological state assessment are psychological state labels determined by psychological counselors based on student interview content and EEG characteristics. These labels are used to guide the model in learning questioning methods and response strategies under different psychological states.

[0058] In this embodiment, the prompt word adjustment information is generated by transforming EEG data into prompt words that reflect changes in the student's psychological state after signal denoising, spectrum analysis, and psychological feature extraction. For example, when a change in EEG data is detected, the system automatically generates the prompt word: "Increased anxiety, decreased attention."

[0059] During the dual-modal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector tables are combined and input into a preset AI big model. The AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information.

[0060] Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

[0061] For example, when the input is "Student: I've been feeling anxious lately" and the EEG indicates "anxiety ↑", the next round of human psychological interview question is "AI: What's been bothering you lately?". If the model outputs "AI: Is there something you can't let go of?" and the semantic similarity is 0.93, which is greater than 0.9 (the first preset threshold), the psychological state is identified as "anxiety", which is consistent with the human, and then the fine-tuning is considered to have met the standard.

[0062] In this embodiment, a pre-set AI model is used to assess the psychological state of the test subject based on the subject's psychological interview dialogue records, EEG data, and pre-set psychological state assessment prompts, obtaining the psychological state assessment result of the test subject, specifically including:

[0063] The brainwave data of the test subject is processed to obtain brainwave feedback information;

[0064] Preprocessing of psychological interview records of the subjects to be tested;

[0065] The brainwave feedback information is converted into cue word adjustment information, and the cue word adjustment information, along with the preprocessed psychological interview dialogue record and the preset psychological state assessment cue words, are input into the preset AI large model to obtain the psychological state assessment results of the test subject.

[0066] In the process of psychological state assessment, EEG data is used to dynamically regulate the semantic generation of the AI ​​model. By reflecting the real-time psychological state of the test subject, the model adjusts the attention distribution and language style during the generation process, so that the AI ​​model can synchronously reflect the physiological and emotional changes of the test subject when generating questions or assessment results, thereby improving the accuracy and empathy of psychological state assessment.

[0067] The psychological indicators include depression, anxiety, stress, focus, meditation level, flow state, and overall mood index.

[0068] In one specific embodiment, the AI ​​large model uses the Tongyi Qianwen QwQ-32B as the base model and performs bimodal fine-tuning on a self-collected dataset of student psychological interviews. This bimodal fine-tuning process not only utilizes the psychological interview dialogue records but also introduces EEG feedback information. The EEG feedback is converted into cue word modulation information and input into the model along with the dialogue text, thereby enabling the model to learn its response capabilities under different psychological states.

[0069] The psychological state assessment results output in the form of a report in this embodiment include an analysis of the student's psychological state, potential problems, and future suggestions for reference by the school counselor.

[0070] Secondly, such as Figure 1 As shown, this invention also proposes an assisted psychological interview report generation system based on AI dialogue and EEG data, comprising:

[0071] The electroencephalogram (EEG) monitoring module is used to acquire EEG data of the subject during psychological interviews and dialogues.

[0072] The digital human dialogue module utilizes a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list, and pushes these questions to the test subject. It acquires real-time data on the test subject's responses to the questions and the resulting EEG data. The module determines whether the generated questions have traversed the pre-set question list. If not, it uses the pre-set AI model to adjust the questioning strategy and response style based on the responses and EEG data. It then generates the next psychological interview question based on the pre-set interview prompts, the pre-set question list, and the adjusted questioning strategy and response style, repeating the process until the question list has been traversed. If yes, it uses the pre-set AI model to generate the psychological interview dialogue record of the test subject based on the questions and responses.

[0073] The data processing and analysis module is used to use a pre-set AI model to assess the psychological state of the test subject based on the psychological interview dialogue record, EEG data, and pre-set psychological state assessment prompts, and obtain the psychological state assessment results of the test subject.

[0074] The report generation module is used to output the psychological state assessment results of the test subjects in the form of reports.

[0075] In this embodiment, the preset AI large model is fine-tuned using bimodal fine-tuning data.

[0076] The bimodal fine-tuning dataset includes multiple fine-tuning samples. Each fine-tuning sample includes psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, the artificial psychological state assessment results corresponding to the psychological interview questions and answers, and the next round of artificial psychological interview questions.

[0077] During the dual-modal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector tables are combined and input into a preset AI big model. The AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information.

[0078] Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

[0079] Specifically, a pre-set AI model is used to assess the psychological state of the test subjects based on their psychological interview dialogue records, EEG data, and pre-set psychological state assessment prompts, yielding the psychological state assessment results, which include:

[0080] The brainwave data of the test subject is processed to obtain brainwave feedback information;

[0081] Preprocessing of psychological interview records of the subjects to be tested;

[0082] EEG feedback information is converted into cue word adjustment information and input into a pre-set AI model along with the psychological interview dialogue record. This allows for dynamic adjustment of the cue words, guiding the model to adopt adaptive empathic expressions and questioning methods in subsequent interviews. This improves the adaptability and effectiveness of the psychological interview process and yields the psychological state assessment results of the test subjects.

[0083] The preset AI model in the digital human dialogue module and the data processing and analysis module is the same AI model, only the prompt words used are different.

[0084] In this embodiment, the digital human dialogue module and the brainwave monitoring module work in parallel to collect dialogue data and brainwave data, respectively.

[0085] The present invention will now be described in conjunction with specific embodiments.

[0086] Example 1

[0087] This invention proposes an assisted psychological interview report generation system based on AI dialogue and EEG data, comprising:

[0088] The digital human dialogue module is used to deploy a digital human interface on Android devices. It supports natural language dialogue, collects student user dialogue data, and its main functions are: using a preset AI model to generate psychological interview questions based on preset interview prompts and a preset question list, and pushing the psychological interview questions to the test subject; acquiring the test subject's answers to the psychological interview questions and the EEG data of the answering process in real time; determining whether the generated psychological interview questions have traversed the preset question list; if not, adjusting the questioning strategy and answering style of the psychological interview questions based on the answers and the EEG data of the answering process using the preset AI model; generating the next psychological interview question based on the preset interview prompts, the preset question list, and the adjusted questioning strategy and answering style, repeating the steps until the question list has been traversed; if so, generating the psychological interview dialogue record of the test subject based on the psychological interview questions and answers using the preset AI model.

[0089] The EEG monitoring module is used to collect EEG data in real time through wearable devices and process the EEG data to obtain seven indicators, including depression, anxiety, stress, concentration, meditation level, flow, and comprehensive mood index, to monitor the students' physiological and psychological state during the interview process, form a real-time quantitative monitoring of the students' psychological state, and provide multi-dimensional data support.

[0090] The data processing and analysis module is used for preprocessing psychological interview dialogue records, including data cleaning and standardization; processing the EEG data of the test subjects to obtain EEG feedback information; preprocessing the psychological interview dialogue records of the test subjects; converting the EEG feedback information into prompt word adjustment information, and inputting the prompt word adjustment information, the preprocessed psychological interview dialogue records, and the preset psychological state assessment prompt words into the preset AI large model to obtain the psychological state assessment results of the test subjects.

[0091] The report generation module is used to output structured psychological state assessment results; the psychological state assessment results output in report form include student psychological state analysis, potential problems and future suggestions.

[0092] The digital human dialogue module includes a pre-set AI model. This model generates natural and friendly responses in real time based on 14 core questions from the "Elementary and Secondary School Students' Psychological Assessment Interview Outline Record Form." For example, when a student says "learning is difficult," the digital human dialogue module dynamically adjusts its questioning based on EEG feedback. If anxiety levels rise, it generates a gentler, more reassuring question: "Which subject is difficult for you?" This helps to delve deeper into the source of the student's learning pressure, enhancing empathy and interview depth. This dynamically adjusted dialogue strategy ensures the interview's relevance and fluency.

[0093] Example 2

[0094] Using the AI-assisted psychological interview report generation system based on dialogue and EEG data described in Example 1, a total of 100 psychological interview experiments were conducted on 50 students in two schools. In each experiment, a student conducted a complete interview with the system. The data were averaged to ensure the scientific validity and reliability of the results.

[0095] In this embodiment, five professional psychology teachers were invited to score the psychological interview reports generated in each experiment (the scoring range was 0-100%, based on the consistency between the report and the student's actual psychological state), and the average value was taken as the report accuracy index. The experiment compared human psychological interviews, AI psychological interviews (voice dialogue only), and AI psychological interviews combined with EEG (this invention), and the results are shown in Table 1.

[0096] Table 1

[0097]

[0098] As shown in Table 1, the interview and report generation times of this invention are significantly reduced. Manual psychological interviews require 33 minutes (interview) and 27 minutes (report generation), which is inefficient due to reliance on individual questioning by the counselor and manual report writing. AI-based psychological interviews (voice-only dialogue) shorten the interview time to 18 minutes and the report generation time to 3 minutes through automated dialogue. AI-based psychological interviews incorporating EEG (this invention) slightly increase the interview time to 19 minutes (due to the need for simultaneous EEG data analysis), but the report generation time is only 4 minutes, far shorter than manual methods. This is thanks to the efficient data processing capabilities of the AI ​​large-scale model and the automated report generation mechanism, resulting in an overall efficiency improvement of approximately 8 times and a significant reduction in the workload of counselors.

[0099] In this embodiment, the accuracy of the report was calculated by averaging the scores given by five professional psychological counselors for each interview report (based on the consistency between the report content and the student's actual psychological state). Human psychological interviews, due to the experience and judgment of professionals, achieved an average accuracy of 85%; AI psychological interviews (voice-only) had an accuracy of 75% due to a lack of physiological data support; and AI psychological interviews incorporating EEG data (this invention) improved the objectivity and comprehensiveness of the report by integrating EEG data (such as indicators of depression and anxiety), achieving an average accuracy of 82%, approaching the level of human assessment.

[0100] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for generating assisted psychological interview reports based on AI dialogue and EEG data, characterized in that, include: S1. Utilize a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list; S2. Send the psychological interview questions to the test subject; S3. Real-time acquisition of the test subject's answers to psychological interview questions and EEG data during the answering process; S4. Determine whether the psychological interview questions have traversed the preset question list; if not, proceed to S5; if yes, proceed to S6. S5. Using a pre-set AI model, adjust the questioning strategy and answering style of the psychological interview questions based on EEG data; generate the next psychological interview question based on the interview prompts, the pre-set question list, and the adjusted questioning strategy and answering style of the psychological interview questions, and proceed to S2; S6. Using a pre-set AI model, generate psychological interview dialogue records of the test subject based on psychological interview questions and answers; process the EEG data of the test subject to obtain EEG feedback information; preprocess the psychological interview dialogue records of the test subject. EEG feedback information is converted into cue word adjustment information; the cue word adjustment information, along with the preprocessed psychological interview dialogue record and the preset psychological state assessment cue words, are input into the preset AI large model to obtain the psychological state assessment results of the test subject; the psychological state assessment results of the test subject are output in the form of a report. The pre-set AI model is fine-tuned in two modes using bimodal fine-tuning data. The bimodal fine-tuning dataset includes multiple fine-tuning samples. Each fine-tuning sample includes psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, and the artificial psychological state assessment results corresponding to the psychological interview questions and answers and the next round of artificial psychological interview questions. During the dual-modal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector representations are combined and input into a preset AI big model. The AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information. Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

2. The method for generating assisted psychological interview reports based on AI dialogue and EEG data according to claim 1, characterized in that, The AI ​​model used in the psychological state assessment process and the AI ​​model used in the psychological interview process are the same AI model.

3. A system for generating assisted psychological interview reports based on AI dialogue and EEG data, characterized in that: include: The electroencephalogram (EEG) monitoring module is used to acquire EEG data of the subject during psychological interviews and dialogues. The digital human dialogue module utilizes a pre-set AI model to generate psychological interview questions based on pre-set interview prompts and a pre-set question list, and pushes these questions to the test subject. It acquires real-time data on the test subject's responses to the questions and the resulting EEG data. The module determines whether the generated questions have traversed the pre-set question list. If not, it uses the pre-set AI model to adjust the questioning strategy and response style based on the responses and EEG data. It then generates the next psychological interview question based on the pre-set interview prompts, the pre-set question list, and the adjusted questioning strategy and response style, repeating the process until the question list has been traversed. If yes, it uses the pre-set AI model to generate the psychological interview dialogue record of the test subject based on the questions and responses. The data processing and analysis module is used to process the EEG data of the test subject to obtain EEG feedback information; Preprocessing of psychological interview records of the subjects to be tested; EEG feedback information is converted into cue word adjustment information, and the cue word adjustment information, along with the preprocessed psychological interview dialogue record and the preset psychological state assessment cue words, are input into the preset AI large model to obtain the psychological state assessment results of the test subject. The report generation module is used to output the psychological state assessment results of the test subjects in the form of a report; The pre-set AI model is fine-tuned in two modes using bimodal fine-tuning data. The bimodal fine-tuning dataset includes multiple fine-tuning samples. Each fine-tuning sample includes psychological interview questions and answers, cue word modulation information obtained from EEG data in the psychological interview questions and answers, the artificial psychological state assessment results corresponding to the psychological interview questions and answers, and the next round of artificial psychological interview questions. During the dual-modal fine-tuning process, the psychological interview questions and answers are encoded to obtain text vector representations; the cue word modulation information is encoded to obtain EEG vector representations; the text vector representations and EEG vector representations are combined and input into a preset AI big model. The AI ​​big model generates the next round of psychological interview questions by learning the correspondence between the psychological interview dialogue records and the cue word modulation information. Based on the generated next-round psychological interview questions and the corresponding next-round human psychological interview questions, calculate semantic similarity; based on the psychological state prompts in the prompt word adjustment information and the corresponding human psychological state assessment results, calculate the consistency of psychological state labels; if the semantic similarity reaches the first preset threshold and the consistency of psychological state labels reaches the second preset threshold, the fine-tuning ends; otherwise, continue fine-tuning.

4. The AI-assisted psychological interview report generation system based on dialogue and EEG data as described in claim 3, characterized in that, The preset AI model in the digital human dialogue module and the data processing and analysis module is the same AI model.

Images