2026 ASEE Annual Conference & Exposition

Researchers find that differing levels of participation in citizen science have different effects on participants’ STEM self-efficacy. We center two participation levels: contributory projects, which correspond to minimal research participation, typically in the form of data collection, and collaborative/participatory projects, which correspond to an active role in the entire research process, typically including the formation of a research question and data analysis. In this study, part of a larger examination of different participation levels and different participant backgrounds in research, we examine changes in STEM self-efficacy after participating in a contributory style citizen science project in artificial intelligence/machine learning (AI/ML) for plant science. Citizen science participation includes engagement with an emerging machine learning segmentation tool designed to detect plant roots in underground images.

Using an explanatory sequential mixed-methods approach, we examined STEM self-efficacy and interest in research of graduate students who participated in a pilot study for the machine learning (ML) tool. We recruited eight graduate students from a [large public university’s] Department of Computer and Information Science and Engineering (CISE), Department of Electrical and Computer Engineering (ECE); we invited but did not get plant science graduate researchers via email. Participants engaged with ML tools for 90 minutes to annotate plant root images for the segmentation model. We administered pre- and post-surveys to assess changes in self-efficacy and research interest. A power analysis indicated that a minimum of 15 participants is required to achieve a significance level of 0.05 and a power of 0.8. Paired-sample t-test results from eight complete survey responses showed no statistically significant change in self-efficacy scores. However, 78% of participants reported confidence in labeling plant root structures, and two-thirds indicated increased confidence in engaging with AI/ML research. Additionally, two-thirds expressed heightened interest in pursuing a research career following the study. Qualitative data supported these findings, with participants reporting improved understanding of ML labeling tools and citing them as influential in their growing interest in AI/ML research.

Future work will involve semi-structured interviews with prior participants to explore their continued engagement with AI/ML research. These findings will extend current knowledge on how even contributory participatory research can influence students’ confidence and interest in research over time. Further studies will include high school students, professional scientists, and engineers to investigate how citizen science impacts self-efficacy across diverse disciplines and educational levels. We will compare the outcomes of our intermediate or “emerging expert” group to outcomes of participants with little to no experience with AI/ML tools and implementations, and experts in the field of AI/ML.