Addressing Ubiquitous STEM Gender Performance Differences

点击翻译按钮获取中文摘要

This is a wide reaching project that combines strands of a number of NSF priorities and earlier NSF-supported projects, including "Writing to Learn," WIDER (institutional transformation), INCLUDES (broadening participation), Cyber learning, and Big Data Analytics. Using big data analytics, it has been quantitatively established that female students taking introductory STEM lecture courses at 10 major public research universities show evidence of stereotype threat, which lowers their course performance by 5% to 10% and discourages some from continuing as STEM majors. The goal of the project is to continue to refine and employ a digital mentoring/ treatment tool that has undergone pilot stage testing to deliver up to three types of treatments to countervail the presence of stereotype threat. This tool has the potential to be customized for each student. It was developed with early stage support from an earlier NSF/ DUE grant. This tool will use a writing-to-learn tool being developed under an NSF/IUSE grant. The writing tool is building a comprehensive writing system, combining peer review tools with a natural language processing toolkit designed to provide actionable information about student responses to the digital mentoring tool. This project is expected to have a substantial direct impact on female STEM students and has good prospects for encouraging more of them to remain in STEM as their primary college major. If the tool proves to be effective in eliminating gender performance differences in STEM courses, it is likely to be employed in a growing number of other research universities, thereby multiplying its impact. The project itself will work with 5,000 distinct undergraduate women in a randomized treatment design to test the efficacy of basic treatment and to test the improved efficacy of individualizing the treatment for each student. Because all of the interactions with students will take place through the digital mentoring/ treatment tool framework, the project will have complete control over what each student experiences, along with a comprehensive "treatment record" of what each student encountered and did within the system, including the writing they did in response to the intervention prompts. The ability to control and record personalized treatment for every individual allows the treatment to be assessed through randomized trials. This approach has been the key to the development of sequential multiple assignment randomized trials in digital health coaching, for example.