Machine Learning Makes Scoring Easier
A fast, easy way to use machine learning models scoring your image data – no expertise or coding required.
The Story of TINA---a PHYSICS TEACHER
Tina is a high school physics teacher in the south of Georgia and has 10 years of teaching experience. She is highly praised and respected by most of her students and their parents because of her professional teaching and profound content knowledge of physics. One important evidence is that most students in her class have great testing scores in physics, one of the most challenging subjects in high school, compared with their peer classes. However three-dimensional (3D) learning goals (i.e., disciplinary core ideas, crosscutting concept, and scientific practices), suggested by the national research council (i.e., NRC) framework and the next generational science standards (i.e., NGSS) challenged Tina’s traditional teaching and assessment practice. To achieve the 3D learning goals, students should learn physics in context, in authentic practice, similar to scientists do in their daily scientific practice. That is, to understand the phenomenon or figure out complex problems, students need to follow the procedure of putting forward a hypothesis, constructing scientific modeling, and collecting evidence to test the hypothesis, then making a valid conclusion. In this procedure, students should construct their scientific modeling based on their prior knowledge and understanding. More importantly, the 3D learning goals put forward new criteria in science assessment practice such as evaluating students’ scientific modeling competence.
To achieve the 3D science learning goals, Tina is concerned about enhancing students’ scientific modeling competence. To enhance students' scientific modeling competence, Tina guided students to construct modeling in scientific practice such as simulation or inquiry learning. Students were excited about learning by doing and had high interest and motivation in modeling construction. However, students’ scientific modeling competence is a complex cognition process, which could not be assessed using traditional multiple-choice assessment practices. Thus, Tina adopted multiple approaches to assess students’ scientific modeling competence (e.g., constructed responses, drawing). For example, she required students to draw the conceptual models and explain their models about how thermal energy is transferred to the water. From students’ drawing and written responses, Tina could have a better understanding of students’ cognition process in solving physics problems.
However, Tina has new problems regarding grading students’ complex responses. She could grade manually, however, not within a timely fashion. More importantly, she can't grade manually when he frequently uses this way to evaluate students’ scientific modeling competence because she needs to give students timely feedback and specific instruction based on the grading. Thus, Tina asks herself: “how could I grade students’ complex responses especially the drawing quickly, in a timely fashion?” “How could I give students personalized feedback?” “How could I interpret the scoring result to help me make better instruction decisions?” Tina feels huge pressure in using drawing and written responses to assess students' scientific modeling competence. She expects to have a robot or machine assistant to help her do grading jobs. She imagines that the grading assistant could automatically score students constructed responses or drawing timely. More importantly, she hopes the assistant could provide him the result report, which is not only the score of students’ modeling but also a specific learning guide. By combing the grading guide and her content knowledge, Tina could provide students specific instruction. With this perception of grading assistant, Tina expects to achieve the three-dimensional learning goals by conducting effective science assessment work, especially scientific modeling competence assessment practice.
When Tina feels frustrated, Lacie, a Ph.D. student in Learning, Design, and Technology program, introduced the automatic analysis of the visualized modelling (AAVM) web portal to her. Lacie told her that this web portal is super user friendly and showed her how to use the web portal grade students' developed image data. Tina was excited to know the AAVM web portal because it is just what she is looking for. She could not wait to try it. She followed the instruction step by step to upload the data file, select the algorithm model(s), and hit the running button to get the result within a minute. After she got the result, she could download the result or save it to her computer. The web portal provided two kinds of results, one is overall score, one is machine-provided instruction guidance. After Tina used this web portal, she frequently used students' developed scientific visualized modeling in her class and had the confidence to make the best instruction decision. More importantly, her students could have timely feedback and personalized instruction, thus having increasingly improved scientific modeling competence and developing higher level scientific skills.
