Towards using educational data mining to improve the assessment of student understanding of ecology

Background/Question/Methods

In recent years, educational data mining methods have been effective for assessing student understanding within a range of domains, including mathematics and a range of sciences. They have been successful at modeling students’ developing process understanding and skill at scientific inquiry. In this talk, I synthesize past work on modeling science learning in other domains towards understanding how educational data mining may be useful to improve the assessment of student understanding of ecology.

Results/Conclusions

Educational data mining methods have been used to assess three aspects of science learning relevant to this goal: development of understanding of scientific processes, development of scientific inquiry skill, and the degree to which students can learn domain facts by leveraging these two other types of learning. Our group has been involved in all three. In work in the Genetics Tutor (with Albert Corbett and Arnon Hershkovitz), we have successfully predicted which students will develop robust learning that prepares them for future learning, using evidence from both students’ patterns of learning and their meta-cognitive behaviors. In work in Science ASSISTments (with Michael Sao Pedro and Janice Gobert), we have successfully inferred students’ inquiry skill in terms of effectively collecting data, and have used these models to show that students can transfer these skills between domains. In work in Virtual Performance Assessments (with Jody Clarke-Midura), we have successfully predicted whether students will correctly solve science puzzles and be able to design causal explanations to justify their claims, based on students’ inquiry behaviors within the learning environment. I will conclude with a discussion of potentials for these methods to positively influence practice in assessment and supporting learning in the domain of student understanding of ecology.