A project-oriented learning quality traction and credible constraint formative evaluation method and system

Abstract

The application discloses a quality traction and credible constraint type formative evaluation method and system for project-based learning, and the method comprises the following steps: determining an evaluation framework, publishing problem quality weight evaluation standards, making Rubrics scales clear, and publishing evaluation rules to students; in the whole process of implementing project-based learning, problem data processed by each student is continuously collected, and each problem is scored in four stages according to the Rubrics scales, thereby forming a process evidence library of problem levels; the effective value of each problem is calculated; the comprehensive score of the student is calculated; the final score of each student is output, and an evaluation report containing problem level details is generated. Through tracking the complete life cycle of all problems of the student in the course project, the number, quality, solution effect and credibility of the problem proposer and other multi-dimensional information of the problem are fused into a comprehensive score, so that the accurate, objective and quantifiable formative evaluation of the professional ability growth of the student in the project-based learning is realized.

Description

Technical Field

[0001] This invention belongs to the field of educational evaluation technology, and relates to formative evaluation technology for the growth of students' professional abilities in project-based learning. Specifically, it relates to a quality-driven and credible constraint-based formative evaluation method and system for project-based learning. Background Technology

[0002] Project-based learning (PBL) is a teaching methodology that uses real-world problems or projects as a vehicle and emphasizes students' independent exploration in an open-ended problem space. Students experience the entire process from problem definition, solution conception, detailed design to outcome implementation, constructing knowledge and developing skills through "learning by doing." Formative assessment is an evaluation paradigm that runs throughout the learning process and aims to continuously improve teaching and learning. In PBL, it is defined as: a systematic activity that continuously collects student learning evidence during project implementation to diagnose learning status, provide feedback for improvement, and adjust teaching strategies, thereby promoting students' skill development.

[0003] However, existing PBL formative assessment methods suffer from numerous technical shortcomings in practical applications, hindering the effectiveness of formative assessment. Through a systematic review of domestic and international PBL formative assessment practices, the existing methods exhibit the following technical deficiencies:

[0004] 1. The evaluation granularity is coarse, lacking effective tracking of the problem-level learning process. Existing methods mostly evaluate at the project as a whole or at a stage (such as mid-term review or final defense), using the Rubric scale to score the overall results. The process information that best reflects students' ability development—what questions students raised in the project, who raised the questions, the difficulty of the questions, and the quality of the solutions—is scattered in teachers' personal records and not systematically incorporated into the scoring system. The evaluation granularity remains at the "project level" or "stage level," failing to reach the "problem level," making it difficult for the evaluation results to truly reflect students' actual ability levels.

[0005] 2. Evaluation indicators rely on multi-parameter linear weighting, which is highly subjective and difficult to standardize. Existing methods typically weight and summarize multiple evaluation dimensions such as technical skills, teamwork, and communication to obtain the final score. The setting of weights depends on the teacher's subjective judgment, and the weight configuration varies greatly between different courses and different teachers, making it difficult to compare the evaluation results of the same student in different courses. The large number of parameters and the complex parameter tuning process make it difficult to ensure the uniformity and fairness of evaluation standards in large-scale teaching.

[0006] 3. Lack of effective incentive mechanisms for high-quality learning outcomes. In the traditional linear weighted model, the score contribution of high-quality tasks and low-quality tasks is only linearly proportional, failing to reflect the premium effect of "challenging highly difficult problems." Students can accumulate points by completing a large number of simple tasks without delving into core challenges, leading to a bias in evaluation orientation and hindering the development of students' ability to solve complex problems.

[0007] 4. Weak anti-fraud and anti-cheating mechanisms. Traditional methods have constant marginal returns, allowing students to artificially boost their total score by deliberately increasing the quantity of low-quality output. With the widespread adoption of Large Language Models (LLM), students can further embellish their project outcomes using AI-generated content. Existing evaluation methods rely on final deliverables, making it difficult to identify situations where "the process is authentic but the outcome is not independently completed," and lacking a credibility constraint mechanism covering the entire process from problem identification to solution. Summary of the Invention

[0008] Purpose of the invention: To overcome the shortcomings of existing technologies, this invention provides a quality-driven and credible-constrained formative assessment method and system for project-based learning (PBL). It proposes a novel approach to formative assessment based on problem tracking: using "problems" as the most basic and core assessment unit in PBL; in the PBL process, whether problems are discovered by students or raised by teachers, the entire process of proposing, researching, solving, and reflecting on each problem serves as observable evidence of student ability development; by tracking the complete lifecycle of all problems students encounter in course projects, multi-dimensional information such as the quantity, quality, effectiveness of solutions, and credibility of proposers is integrated into a comprehensive score, achieving accurate, objective, and quantifiable formative assessment of students' professional ability growth in project-based learning.

[0009] Technical Solution: To achieve the above objectives, this invention provides a quality-driven and credible-constraint formative assessment method for project-based learning, comprising the following steps:

[0010] S1: Teachers determine the evaluation framework based on the course teaching objectives, announce the evaluation criteria for the quality weight of questions, clarify the Rubric scale for the four-stage scoring dimensions, set the initial value of the half-saturation parameter, and announce the evaluation rules to students.

[0011] S2: Throughout the implementation of project-based learning, continuously collect problem data from each student and score each problem in four stages according to the Rubric scale, record the original scoring data, and form a problem-level process evidence base.

[0012] S3: Calculate the effective value of each issue based on the process evidence base;

[0013] S4: Calculate and obtain the student's comprehensive score based on the effective value of the question obtained in step S3;

[0014] S5: Outputs the final score for each student, generates an evaluation report with detailed problem-level breakdowns, allows teachers and students to view the score composition, and enables teachers to provide targeted improvement feedback based on the evaluation report.

[0015] Furthermore, the problem quality weighting criteria in step S1 include:

[0016] Excellent quality, with a quality weight of 1.0 to 1.2, applicable scenarios: complex problems involving core algorithm design, architecture refactoring, and cross-module integration;

[0017] High quality, with a quality weight of 0.8 to 1.0, applicable scenarios: involving the implementation of key functions, performance optimization, and the repair of important defects;

[0018] General quality, with a quality weight of 0.5 to 0.7, applicable scenarios: function implementation, routine defect repair, and documentation supplementation;

[0019] Low quality, with a quality weight of 0.2 to 0.4, suitable for: format adjustment, simple configuration modification, and repetitive tasks.

[0020] Furthermore, the four-stage scoring dimensions in step S1 include understanding, research, problem-solving, and reflection.

[0021] Furthermore, the evaluation rules in step S1 include the evaluation criteria for quality weights, the setting principles for credibility coefficients (1.0 for teacher questions and 0.9 for student questions), and the mathematical form of the scoring function, to ensure that the evaluation process is transparent and predictable, and to guide students to focus on the quality of problem-solving rather than the quantity.

[0022] Furthermore, the problem data in step S2 includes a description of the problem content, the identity of the problem proposer (teacher / student), and a four-stage performance record of the entire problem handling process (the answering process, intermediate products, and final results of each stage of understanding, research, solving, and reflection), ensuring that the process data is complete and traceable.

[0023] Furthermore, the calculation of the effective value in step S3 includes:

[0024] A1: Calculation of Comprehensive Problem Score: For each problem i, take the arithmetic mean of the four-stage scores (understanding, research, solution, reflection), and then normalize it to the [0,1] interval to obtain the comprehensive problem score. ;

[0025] A2: Question Quality Weighting Assessment: Teachers assess the quality weighting based on factors such as question difficulty, knowledge depth, and innovativeness. ;

[0026] A3: Credibility rating of the question submitter: Based on the credibility coefficient of the question submitter. (Teacher 1.0, Student 0.9) reflects the difference in the authority of the source of the problem;

[0027] A4: Effective Value Calculation and Square Enlargement: Problem-by-Problem Effective Value Calculation Then ask The square operation amplifies the contribution of high-quality problems and suppresses the contribution of low-quality problems, thus achieving quality-driven growth.

[0028] Furthermore, the calculation of the student's overall score in step S4 includes:

[0029] B1: Sum of squares of values ​​in a comprehensive problem: The effective values ​​of all problems are summed up by squares to form the comprehensive problem value of the student.

[0030] B2: Substitute into the scoring function: The comprehensive value is mapped to the interval [60, 100] using a power-law saturation function, with a baseline score of 60 guaranteeing a passing grade and an upper limit of 100 capping the score. The parameter is for half-saturation.

[0031] B3: Truncating: If the score is > 100, it is truncated to 100; if the score is < 60, it is truncated to 60, to ensure that the score falls within a reasonable range.

[0032] Furthermore, in step S5, teachers rely on the evaluation report to assess students' actual learning engagement and knowledge mastery, identify students' weaknesses and strengths, and provide targeted feedback for improvement, thereby shifting from outcome-based assessment to a comprehensive and refined formative evaluation.

[0033] This invention also provides a quality-driven and credibility-constrained formative assessment system for project-based learning, comprising:

[0034] The course preparation and evaluation framework initialization module allows teachers to determine the evaluation framework based on the course teaching objectives, publish the evaluation criteria for question quality weights, clarify the Rubric scale for the four-stage scoring dimensions, set the initial values ​​of the half-saturation parameters, and announce the evaluation rules to students.

[0035] The student process data collection module continuously collects problem data from each student throughout the entire process of project-based learning. It also scores each problem in four stages according to the Rubric scale, records the original score data, and forms a problem-level process evidence library.

[0036] The per-question scoring calculation module calculates the effective value of each question based on the process evidence base.

[0037] The comprehensive score calculation module calculates and obtains the student's comprehensive score based on the effective value of the questions obtained.

[0038] The results output and teaching feedback module outputs the final score for each student and generates an evaluation report with detailed problem-level breakdowns. It allows teachers and students to view the score composition, and teachers can provide targeted improvement feedback based on the evaluation report.

[0039] Beneficial effects: Compared with the existing technology, the core innovation of this invention lies in: reducing the evaluation granularity from the traditional "project level" to the "problem level", naturally integrating the multi-dimensional attributes of the problem through a mathematical model, and achieving unified processing of quantity incentives, quality guidance, credibility constraints and volume fraud suppression under the premise of extremely simple parameters.

[0040] This invention proposes a formative assessment scheme that offers a new research approach to course process assessment. Traditional assessment methods often focus on scoring the results of periodic teaching assessments, relying on simple weighted averages to calculate the final grade. The assessment dimensions are result-oriented, making it difficult to accurately capture students' true learning status. This invention provides refined assessment of various learning problems students exhibit throughout the course implementation process. The assessment granularity is reduced to the problem level, making it more targeted. Furthermore, the power-law saturation model has good interpretability, and the scoring mechanism is clear and transparent. The assessment model provided by this invention can effectively prevent cheating behaviors such as plagiarism and the use of artificial intelligence tools to generate content in bulk. Teachers can comprehensively and objectively assess students' true learning engagement and knowledge mastery level based on multi-dimensional process data, including students' performance throughout the entire process and the quality of their problem-solving. This achieves a shift from result-oriented assessment to a comprehensive, refined formative assessment. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the process framework of the method of the present invention. Detailed Implementation

[0042] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0043] Example 1:

[0044] like Figure 1 As shown, this embodiment provides a quality-driven and credibility-constrained formative assessment method for project-based learning, including the following steps:

[0045] S1: Teachers determine the evaluation framework based on the course teaching objectives, publish the evaluation criteria for question quality weighting, clarify the Rubric scale with four-stage scoring dimensions, and set the half-saturation parameter. The initial value (in this embodiment) The initial value is 4), and the evaluation rules are announced to the students;

[0046] The criteria for assessing the quality of issues include:

[0047] Excellent quality, with a quality weight of 1.0 to 1.2, applicable scenarios: complex problems involving core algorithm design, architecture refactoring, and cross-module integration;

[0048] High quality, with a quality weight of 0.8 to 1.0, applicable scenarios: involving the implementation of key functions, performance optimization, and the repair of important defects;

[0049] General quality, with a quality weight of 0.5 to 0.7, applicable scenarios: function implementation, routine defect repair, and documentation supplementation;

[0050] Low quality, with a quality weight of 0.2 to 0.4, suitable for: format adjustment, simple configuration modification, and repetitive tasks.

[0051] The four-stage assessment dimensions include understanding, research, problem-solving, and reflection. The assessment rules include the criteria for quality weighting, the principles for setting credibility coefficients (1.0 for teacher questions, 0.9 for student questions), and the mathematical form of the scoring function, ensuring that the assessment process is transparent and predictable, and guiding students to focus on the quality of problem-solving rather than the quantity.

[0052] S2: Throughout the implementation of project-based learning, continuously collect problem data from each student and score each problem in four stages according to the Rubric scale, record the original scoring data, and form a problem-level process evidence base.

[0053] The problem data includes a description of the problem content, the identity of the problem proposer (teacher / student), and a four-stage performance record of the entire problem handling process (understanding, research, solving, and reflection on the answering process, intermediate products, and final results) to ensure that the process data is complete and traceable.

[0054] S3: Calculate the effective value of each issue based on the process evidence base;

[0055] The calculation process includes:

[0056] A1: Calculation of Comprehensive Problem Score: For each problem i, the four-stage scores (understanding, research, solution, reflection) are weighted and taken as an arithmetic mean, then normalized to the [0,1] interval to obtain the comprehensive problem score. ;

[0057] A2: Question Quality Weighting Assessment: Teachers assess the quality weighting based on factors such as question difficulty, knowledge depth, and innovativeness. ;

[0058] A3: Credibility rating of the question submitter: Based on the credibility coefficient of the question submitter. (Teacher 1.0, Student 0.9) reflects the difference in the authority of the source of the problem;

[0059] A4: Effective Value Calculation and Square Enlargement: Problem-by-Problem Effective Value Calculation Then ask The square operation amplifies the contribution of high-quality problems and suppresses the contribution of low-quality problems, thus achieving quality-driven development.

[0060] S4: Calculate and obtain the student's comprehensive score based on the effective value of the question obtained in step S3;

[0061] The calculation process for a student's overall score includes:

[0062] B1: Sum of squares of values ​​in a comprehensive problem: The effective values ​​of all problems are summed up by squares to form the comprehensive problem value of the student.

[0063] B2: Substitute into the scoring function: The comprehensive value is mapped to the interval [60, 100] using a power-law saturation function, with a baseline score of 60 guaranteeing a passing grade and an upper limit of 100 capping the score. The parameter is for half-saturation.

[0064] B3: Truncating: If the score is > 100, it is truncated to 100; if the score is < 60, it is truncated to 60, to ensure that the score falls within a reasonable range.

[0065] S5: Output the final score for each student and generate an evaluation report containing question-level details (s_i, w_i, τ_i, v_i, v_i² for each question), allowing teachers and students to view the score composition;

[0066] Teachers use evaluation reports to assess students' actual learning engagement and knowledge acquisition, identify their weaknesses and strengths, and provide targeted feedback for improvement, thus shifting from outcome-based assessment to a comprehensive and refined formative evaluation.

[0067] The core parameters involved in this embodiment are defined as follows:

[0068]

[0069] Based on the rating distribution of the first batch of student samples, the k value will be fine-tuned after 1-2 weeks of trial operation (adjustment is only allowed between 3.5 and 5). The k value will be locked and will not be changed during the course to ensure the fairness of the evaluation.

[0070] Based on the above, the present invention is analyzed and summarized as follows:

[0071] I. The objectives of this invention include:

[0072] 1. Design a fine-grained formative assessment method that uses problems as the evaluation unit, thereby reducing the evaluation granularity from the project level to the problem level;

[0073] 2. By using the effective value sum of squares mechanism, a strong incentive effect is achieved for high-quality problems, guiding students to challenge high-difficulty tasks;

[0074] 3. Utilize the inherent properties of hyperbolic saturated functions to mathematically suppress the behavior of obtaining high scores by piling up low-quality problems;

[0075] 4. The model reflects the difference in credibility between teacher and student questions, enhancing the authority of the evaluation results;

[0076] 5. Implement strict normalization of scoring results to a fixed interval of [60, 100] to ensure the comparability of cross-project evaluations;

[0077] 6. Avoid the subjectivity issues of linear weighting of multiple parameters under minimal parameter design.

[0078] II. The target scenarios of this invention are characterized as follows:

[0079] Teaching Scenarios: Suitable for PBL teaching scenarios that emphasize the quality and technical depth of results, such as engineering training, competition courses, and core professional project courses. The core training objective of this type of scenario is to enable students to master the ability to solve complex problems, emphasizing the tackling of highly challenging tasks and the production of high-quality results.

[0080] Student characteristics: Students will need to complete course projects containing multiple technical problems, each with different levels of difficulty (weight), different problem proposers (teachers or students), and different solution scores. Potential strategic behaviors among students may include: choosing easier problems to avoid challenges, using AI tools to generate solutions, and increasing the number of problems solved to boost scores.

[0081] Evaluation requirements: There is a need for an evaluation method that can (1) effectively distinguish the contribution differences between high-quality and low-quality questions; (2) suppress the behavior of brushing up the number of questions from a mathematical perspective; (3) reflect the higher credibility of teachers' questions; (4) output strictly normalized scores; and (5) have parameters that are extremely easy to implement.

[0082] III. This invention uses a power-law saturation model as the core algorithm framework, and has the following algorithmic features:

[0083] Feature 1: Product-coupled structure. Problem scores, quality weights, and proposer credibility are naturally coupled through product, rather than through traditional linear weighting. This structure avoids the subjectivity problems associated with introducing multiple independent weight parameters.

[0084] Feature Two: The Square Amplification Mechanism. By squaring the effective value of each problem, the contribution of high-quality problems exhibits a quadratic increase. For example, under the same scoring conditions, a high-quality problem (weight = 1.0) contributes 4 times more than a general problem (weight = 0.5). This mechanism creates a strong incentive effect for high-quality problems.

[0085] Feature 3: Hyperbolic Saturation Function. Employing a hyperbolic saturation function, it naturally exhibits diminishing marginal returns: scores increase rapidly when the overall value is low, but decrease sharply once a threshold is exceeded. This characteristic, mathematically speaking, discourages the practice of artificially inflating scores by accumulating low-quality problems.

[0086] Feature 4: Fixed Minimum Passing Score and Normalization. The output formula strictly maps the results to the [60, 100] interval, where 60 is the minimum passing score and 100 is the maximum passing score. All project scores are on the same scale and can be directly compared horizontally. It is important to note that this invention is positioned as "formational assessment," primarily measuring students' input and skill development throughout the project process. It should not necessarily be equated with the final grade. In actual teaching, this formative assessment score is proportionally included in the final grade (e.g., 50% or 60%), with the remaining portion determined by the final exam or final defense. If a student performs extremely poorly in the final defense or exam, their final grade can still be below 60.

[0087] In selecting a formative assessment modeling method, this invention does not employ a data-driven modeling method based on artificial intelligence deep learning, but instead uses an explicit mathematical modeling method based on a nonlinear fitting function (power-law saturation model). This choice is based on the following technical considerations.

[0088] Deep learning is a data-driven implicit modeling approach, and its model performance is highly dependent on the scale and distribution of the training data. In teaching evaluation scenarios, the teaching objectives, assessment methods, and student characteristics of different courses vary significantly. When the evaluation scenario changes, deep learning models are prone to prediction failure due to data distribution shifts. This often requires re-completing the entire process of data annotation, model training, and parameter tuning based on the new scenario dataset. This is not only technically cumbersome and lacks flexibility in adapting to different scenarios, but also places high demands on the quality and quantity of labeled data, which is difficult for most teaching institutions to meet.

[0089] In contrast, the nonlinear fitting function (power-law saturation model) used in this invention has an explicit mathematical structure. The model framework is fixed and requires no reconstruction. It can quickly adapt to the formative assessment needs of different courses and teaching models by adjusting only a few parameters with clear educational semantics, such as the half-saturation parameter k. This solution lowers the technical threshold for scenario iteration, avoids the time and computational costs of repetitive training, and allows teachers to adjust assessment rules according to teaching objectives (e.g., changing the intensity of quality incentives by adjusting the k value). It possesses greater operability, flexibility, and stability in diverse teaching assessment scenarios. Simultaneously, the explicit mathematical structure ensures the interpretability of the assessment process; both teachers and students can clearly understand the scoring mechanism, meeting the transparency requirements of educational assessment.

[0090] The method of the present invention in this embodiment has the following technical advantages:

[0091] 1. Lowering the granularity of evaluation: Using questions as the basic evaluation unit, the granularity of evaluation is lowered from the traditional project / stage level to the question level, and the evaluation results can better reflect the students' actual ability level and process performance.

[0092] 2. High-quality incentives: Through the square mechanism, the contribution of high-quality problems is 4 times that of ordinary problems, effectively guiding students to challenge difficult problems and cultivating their ability to solve complex problems.

[0093] 3. Naturally prevents score manipulation: The squared contribution of low-quality questions decreases sharply. It takes 13 low-quality questions to get the same score as 4 high-quality questions. In actual teaching, score manipulation is not feasible.

[0094] 4. Credibility distinction: Teacher questions are counted at the original value (credibility = 1.0), while student questions are counted at a discount of 0.9, reflecting the higher authority and credibility of teacher questions.

[0095] 5. Strict normalization: The output strictly falls within [60, 100], and all item scores are on the same scale, allowing for direct horizontal comparison.

[0096] 6. Extremely simple parameters: The entire system contains only one adjustable system parameter (half-saturation parameter k=4) and one classification parameter (confidence discount), which does not require complex parameter tuning and is easy to implement in teaching.

[0097] Example 2:

[0098] To implement the method of Example 1, this example provides a quality-driven and credible-constraint formative assessment system for project-based learning, including:

[0099] The course preparation and evaluation framework initialization module allows teachers to determine the evaluation framework based on the course teaching objectives, publish the evaluation criteria for question quality weights, clarify the Rubric scale for the four-stage scoring dimensions, set the initial values ​​of the half-saturation parameters, and announce the evaluation rules to students.

[0100] The student process data collection module continuously collects problem data from each student throughout the entire process of project-based learning. It also scores each problem in four stages according to the Rubric scale, records the original score data, and forms a problem-level process evidence library.

[0101] The per-question scoring calculation module calculates the effective value of each question based on the process evidence base.

[0102] The comprehensive score calculation module calculates and obtains the student's comprehensive score based on the effective value of the questions obtained.

[0103] The results output and teaching feedback module outputs the final score for each student and generates an evaluation report with detailed problem-level breakdowns. It allows teachers and students to view the score composition, and teachers can provide targeted improvement feedback based on the evaluation report.

[0104] Example 3:

[0105] To verify the effectiveness of this invention, a simulation test was conducted in this embodiment. The test simulated the PBL learning situation of a normal class of 40 students in the "Software Engineering Project Training" course, as follows:

[0106] I. Test Design

[0107] The simulated class has 40 students, who are divided into 6 student profiles based on their natural distribution:

[0108]

[0109] II. Calculation of Student Profile Scores for Six Categories

[0110] In this example (k=4), scores are calculated for six types of students:

[0111]

[0112] III. Verification of Class Grade Distribution

[0113] Based on the natural distribution of the above six types of students, the academic performance distribution of a class of 40 students is as follows:

[0114]

[0115] IV. Verification of Anti-Fraud Mechanism

[0116] Test scenario: A student attempts to boost their score by submitting a large number of low-quality questions. Assume there are already 3 low-quality questions (score 60.3), and then 10 more of the same type of questions are added.

[0117] The original score was S = 0.165 (3 questions, each with a mean v² of 0.055), resulting in a score of 60.3. After adding 10 more similar questions, with each new question still having a v² of 0.055, the new S = 0.165 + 10 × 0.055 = 0.715, and the new score = 60 + 40 × 0.715² / (4 + 0.715²) ≈ 64.5. Although the number of questions increased from 3 to 13 (a 333% increase), the score only improved from 60.3 to 64.5 (an approximately 7% increase), making the gains from artificially inflating the score extremely low. Furthermore, if teachers discover through the defense that these questions were not completed independently, the overall score will be adjusted. If the value is lowered to 0.3, then for each question v = 0.3 × 0.3 × 0.9 = 0.081, v² = 0.0066, and for 13 questions S = 0.086, the score is only 60.1, which is almost no effect of boosting the number of views.

[0118] Conclusion: The marginal benefit of low-quality problems decays sharply under the squared mechanism, and the behavior of boosting traffic is not feasible in actual teaching.

[0119] V. Test Conclusion

[0120] After testing six student profiles, verifying class performance distribution, and validating anti-fraud mechanisms, the method of this invention meets the design requirements: the performance distribution conforms to the expected "Excellent (90 points and above) 15-25%, Good (80-89 points) as the core range 40-50%, Average (70-79 points) 20-30%, and Poor (60-69 points) 5-10%"; the scores of each type of student fall within the target range, with Excellent A (Independent) scoring approximately 95 points, Excellent B (Mentored) scoring approximately 90 points, Good Standard scoring 80-85 points, Good Growth scoring 85-89 points, Average students scoring 70-79 points, and Poor students scoring 60-69 points; the anti-fraud mechanism is effective. This method is applicable to various PBL project courses that emphasize the quality of outcomes.

Claims

1. A quality-driven and credibility-constrained formative assessment method for project-based learning, characterized in that, Includes the following steps: S1: Teachers determine the evaluation framework based on the course teaching objectives, announce the evaluation criteria for the quality weight of questions, clarify the Rubric scale for the four-stage scoring dimensions, set the initial value of the half-saturation parameter, and announce the evaluation rules to students. S2: Throughout the implementation of project-based learning, continuously collect problem data from each student and score each problem in four stages according to the Rubric scale, record the original scoring data, and form a problem-level process evidence base. S3: Calculate the effective value of each issue based on the process evidence base; S4: Calculate and obtain the student's comprehensive score based on the effective value of the question obtained in step S3; S5: Outputs the final score for each student, generates an evaluation report with detailed problem-level breakdowns, allows teachers and students to view the score composition, and enables teachers to provide targeted improvement feedback based on the evaluation report.

2. The quality-driven and credibility-constrained formative assessment method for project-based learning according to claim 1, characterized in that, The problem quality weighting criteria in step S1 include: Excellent quality, with a quality weight of 1.0 to 1.2, applicable scenarios: complex problems involving core algorithm design, architecture refactoring, and cross-module integration; High quality, with a quality weight of 0.8 to 1.0, applicable scenarios: involving the implementation of key functions, performance optimization, and the repair of important defects; General quality, with a quality weight of 0.5 to 0.7, applicable scenarios: function implementation, routine defect repair, and documentation supplementation; Low quality, with a quality weight of 0.2 to 0.4, suitable for: format adjustment, simple configuration modification, and repetitive tasks.

3. The quality-driven and credibility-constrained formative assessment method for project-based learning according to claim 1, characterized in that, The four-stage scoring dimensions in step S1 include understanding, research, problem-solving, and reflection.

4. The quality-driven and credibility-constrained formative assessment method for project-based learning according to claim 1, characterized in that, The evaluation rules in step S1 include the evaluation criteria for quality weights, the principles for setting the credibility coefficient, and the mathematical form of the scoring function.

5. The quality-driven and credibility-constrained formative assessment method for project-based learning according to claim 1, characterized in that, The problem data in step S2 includes a description of the problem content, the identity of the person who raised the problem, and a four-stage performance record of the entire problem handling process.

6. The quality-driven and credibility-constrained formative assessment method for project-based learning according to claim 1, characterized in that, The calculation of the effective value in step S3 includes: A1: For each question i, take the arithmetic mean of the four-stage scores, and then normalize it to the [0,1] interval to obtain the comprehensive question score. ; A2: Quality weighting determined by teachers ; A3: Based on the credibility coefficient marked by the issue submitter. This reflects the difference in the authority of the source of the problem; A4: Calculate the effective value on a problem-by-problem basis .

7. The quality-driven and credible-constraint formative assessment method for project-based learning according to claim 6, characterized in that, The calculation of the student's overall score in step S4 includes: B1: Sum of squares of values ​​in a comprehensive problem: The effective values ​​of all problems are summed up by squares to form the comprehensive problem value of the student. B2: Substitute into the scoring function: The comprehensive value is mapped to the interval [60, 100] using a power-law saturation function, with a baseline score of 60 guaranteeing a passing grade and an upper limit of 100 capping the score. The parameter is for half-saturation. B3: Truncating: If the score is > 100, it is truncated to 100; if the score is < 60, it is truncated to 60, to ensure that the score falls within a reasonable range.

8. The quality-driven and credible-constraint formative assessment method for project-based learning according to claim 7, characterized in that, In step S5, teachers rely on the evaluation report to assess students' actual learning engagement and knowledge mastery, identify students' weaknesses and strengths, and provide targeted feedback for improvement, thus realizing the transformation from outcome-based assessment to whole-process, refined formative evaluation.

9. A quality-driven and credibility-constrained formative assessment system for project-based learning, characterized in that, The system for implementing the method according to any one of claims 1 to 8 comprises: The course preparation and evaluation framework initialization module allows teachers to determine the evaluation framework based on the course teaching objectives, publish the evaluation criteria for question quality weights, clarify the Rubric scale for the four-stage scoring dimensions, set the initial values ​​of the half-saturation parameters, and announce the evaluation rules to students. The student process data collection module continuously collects problem data from each student throughout the entire process of project-based learning. It also scores each problem in four stages according to the Rubric scale, records the original score data, and forms a problem-level process evidence library. The per-question scoring calculation module calculates the effective value of each question based on the process evidence base. The comprehensive score calculation module calculates and obtains the student's comprehensive score based on the effective value of the questions obtained. The results output and teaching feedback module outputs the final score for each student and generates an evaluation report with detailed problem-level breakdowns. It allows teachers and students to view the score composition, and teachers can provide targeted improvement feedback based on the evaluation report.

Images