A Measurement Study On Learning Behavior In Programming: An Eye-tracking Approach

Programming course is the foundation of information technology education and it's hard for both teachers and students.Traditional written tests and machine tests can provide students with a gradual evaluation of grammatical knowledge,but it is difficult to provide formative evaluations in programming skills such as reading programs and debugging errors.Conducting the measurement of program learning behavior can help teachers to master the details of the student programming process,carry out personalized teaching guidance,and provide a basis for the development of intelligent learning diagnostic tools.This thesis is oriented to the teaching reform needs of the C programming language course.It takes the typical test activities such as error finding with code reading,error correction with debugging as the measurement object,and uses both eye movement measurement and human-computer interaction measurement to carry out student cognitive activities measurement and analyze programming behavior.This thesis proposes a task-driven description model of student programming learning behavior for the eye-tracking points and human-computer interaction events in the process of code reading and debugging.Through interactive decision-making events(running code,etc.),the learning behavior is divided into multiple sub-sequences.Based on the action sequence data,an individual-oriented and group-oriented analysis method is proposed to analyze the students' programming behavior details and their attention characteristics.Using the above measurement techniques and solutions,the author's research team conducted two batches of programming behavior measurement experiments at the School of Telecommunications of Huazhong University of Science and Technology,respectively measuring 15 undergraduate first-year students and 13 master students in code reading and debugging.The measurement analysis results show that the measurement research method proposed in this paper is helpful to interpret the details of the students' movements during the testing process,and provides a digital basis for evaluating students' learning status and proposing learning improvement suggestions.The experimental measurement results also observe the influence of programming habits and skills on the efficiency of learning activities.Students who spend less time reading the test code for the first time often need to invest more time in the later stage of solving the problem.