June 15, 2019
June 15, 2019
October 19, 2019
Computing and Information Technology
Learning how to learn and apply the new knowledge is a vital skill students need to develop. A student’s curiosity in exploring more about a topic lends toward learning that knowledge. In recent years’ text mining has seen increasing focus on investigating sentiment, feelings, behavior, intent, etc. for various purposes, including linguistic understanding, product marketing, and pedagogical improvements. Our project focuses on a relatively unique area, i.e., Curiosity Detection in Text. This “Work in Progress” paper presents preliminary, but promising, results of empirically data mining curiosity in student-produced text data.
Psychologists have wrestled with understanding the nature of curiosity. Recent work by Grossnickle has provided a framework for understanding facets, factors and dimensions of the construct of curiosity that are relevant to the education audience. The key dimensions identified in the framework for curiosity include: focus of curiosity (with four factors of physical, perceptual, social, and epistemic), scope of curiosity (breadth vs. depth), cause of curiosity (diversive vs. specific or interest vs. deprivation), and consistency of curiosity across situational contexts (state vs. trait).
The success of our project will positively impact efforts to assess both curiosity and its impact on educational outcomes. We applied a tool developed by the Right Question Institute, called Question Formulation Technique (QFT), in an Electric Circuits (EC) course to improve students’ ability to formulate questions and deepen their curiosity on EC. The solutions developed will be useful to detect whether curiosity is demonstrated in the results of the QFT exercises, provide analysis on key dimensions of curiosity, and predict associated behaviors of students’.
The QFT student answers from five labs of the EC course are sentences provided in response to a thought-provoking topic, e.g., “Why doesn’t the U.S. adopt SI units”. As per the QFT approach, students respond using a question formulation in incremental steps. The sentences were converted into a token-based numeric format using a text-to-feature conversion and a stop-words list. Each unique word (token) is a feature in the obtained data set. The target class, “Potential for Exploration”, determined by two experts as a measure for student curiosity, has three categories: 1 for Novice, 2 for Intermediate, 3 for Advanced.
The goal of this phase was to identify important words that effectively capture curiosity using data mining. Feature Selection (FS) was performed to reduce the feature space due to data sparsity. Three Wrapper-based approaches (C4.5 decision tree with BestFirst, GreedyStepwise, and Evolutionary search algorithms) and three Filter-based Rankers (ChiSquare, ReliefF, and GainRatio) were used for FS. The C4.5 decision tree was used to classify the “Potential for Exploration”. The wrappers indicated BestFirst and GreedyStepwise generally yielded similar feature sets, while Evolutionary provided much larger feature set. Filters provided less relative similarity, except for a small group of similar words with ChiSquare and GainRatio rankers.
The experts agreed that key features selected were effective representation of students’ curiosity levels. We plan to continue work with: studying curiosity improvement trends, studying curiosity trends across courses, larger data sets, other learners, using linguistic assessment techniques.
Seliya, N., & LeBlanc, H. J., & Hylton, J. B., & Youssfi, Z., & Schweinefuss, M. (2019, June), Data-Driven Investigation of Curiosity in Student Text Responses Paper presented at 2019 ASEE Annual Conference & Exposition , Tampa, Florida. 10.18260/1-2--32577
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 2019 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015