Prior work in chemistry education research has viewed how learning unfolds in the actual college classroom setting as a “black box” needing investigation. The Caspari group’s research tackles this question with analysis of in-the-moment learning at the heart of every project. Our research investigates how student groups learn in-the-moment of their interactions with each other and their instructors and how this learning is impacted by instructor facilitation, class design, and power. Across three distinct areas—1) in-the-moment STEM learning, 2) instructor facilitation, and 3) the impact of educational system aspects on each other—we develop mechanistic frameworks and models of learning and instruction. While these three areas are distinct, the third area is made possible by additional analysis of data, components, and results from the first and second areas. To allow for this synthesis, all three research areas rely on sociocultural theory and related methodology. The main data collection methods are classroom video recording and stimulated recall interviews, and the main analysis methods are formal discourse analysis and inductive and deductive coding. Many of our studies are conducted within Learning Assistant (LA)-supported STEM courses, where undergraduate LAs facilitate in-the-moment learning during small group discussions while concurrently engaging in a pedagogy course and instructional team meetings. Through our focus on LAs as key instructors, our research has direct implications on instructional practice: The mechanistic frameworks and models that result from our research are being used in several institutions for training in LA pedagogy courses that bring together LAs from many LA-supported STEM classes. LAs then apply what they learn in these pedagogy courses to the STEM classes in which they teach, improving learning for thousands of students.

Many chemistry education researchers study how students learn by analyzing written assignments or problem-solving interviews from students after their in-class learning. However, post-hoc assessments are only proxies for student learning and do not always accurately capture in-the-moment learning as it occurs in the classroom. How to track student in-the-moment learning is an important open question in the field.

To address this question, the Caspari research group develops frameworks for studying in-the-moment learning. We draw on sociocultural and pragmatist theories to identify what underlying student needs and questions drive student-student interactions and, in response to these underlying questions, what pieces of knowledge students use to advance their problem solving. Through analysis of LA-facilitated student-student interactions, we have developed a framework that we have recently started calling Continuity-Discourse change Framework for In-the-Moment Learning (CDF-IML) (Karch et al., 2024). CDF-IML provides a way to compare interactions across many student groups despite differences in problems, classroom context, and instructional methods—an analysis that was previously impossible. CDF-IML uses underlying student questions and pieces of knowledge identified to describe whether students’ discussion is continuous from question to question (continuity) and at which points their discussion changes (discourse change) (Fig. 1). Both continuity and discourse change are necessary for learning to occur because students only learn if they connect new ideas (discourse change) to previous ones (continuity). Without discourse change, students would not learn because there would not be anything new; they would only continue discussing ideas they had already explored. Without continuity, students would also not learn because, although they might discuss new ideas, those ideas would not become meaningful without connections to their prior experiences.

Using CDF-IML, researchers can compare different types of in-the-moment learning across different contexts. For instructor practice, CDF-IML provides a basic mechanistic framework with which instructors can follow student learning, identify how students are learning or not, and adjust instruction accordingly. My group trains LAs to analyze LA-student interactions with CDF-IML as part of LA training that is discussed in more depth in research area 2.

Example publication: Karch, J. M., Maggiore, N. M., Pierre-Louis, J., Strange, D., Dini, V., Caspari-Gnann, I.* (2024). Making In-the-Moment Learning Visible: A framework to identify and compare various ways of learning through continuity and discourse change. Science Education, 108(5), 1292-1328.  https://doi.org/10.1002/sce.21874

Instructor facilitation is one important way to impact student in-the-moment learning. This facilitation has been characterized as either authoritative, focused on the single perspective of correctness, or dialogic, focused on multiple perspectives including those of the students. Previous research has shown that most undergraduate STEM classes are taught using primarily authoritative discussions, prioritizing a short path towards correct answers over student thinking. Discourse that is primarily authoritative inhibits the process of students developing into collaborative, free-thinking scientists. Thus, an important lingering question in the field is how to navigate the tension between teaching correct science and focusing on student thinking.

To address this question, the Caspari research group studies how instructors—including LAs, TAs, and faculty—facilitate student-student interactions in STEM classes. For LAs, we found that, instead of facilitating purely authoritatively or purely dialogically, LAs facilitate on an authoritative-to-dialogic spectrum including very and moderately authoritative and dialogic facilitation (Fig. 2) (Carlos et al., 2023).

Very authoritative facilitation focuses on the perspective of correctness embodied by the LA. Moderately authoritative facilitation focuses on the LA perspective without claiming correctness. It guides students along the LA’s thinking, like very authoritative facilitation, but it provides students the opportunity to take more agency in deciding what makes sense to them. Moderately dialogic facilitation relies on the perspective of the students in the group, but with the LA encouraging consideration of another outside perspective such as the reasoning of another student group or an overlooked answer choice. We found that moderately dialogic facilitation gives students space for their own sensemaking while advancing their learning, especially when their perspective alone does not lead to further progress. Very dialogic facilitation focuses on the perspective of the students in the group.

The authoritative-to-dialogic facilitation spectrum is now being used at several institutions for professional development of LAs or other instructors. The Caspari group leads a collaborative project with Profs. Eleanor Close and Alice Olmstead at Texas State University and Andrea Van Duzor at Chicago State University that has developed activities and a video-guided website (https://sites.google.com/tufts.edu/la-facilitation-pratices) to support use of the authoritative-to-dialogic facilitation spectrum in LA training. Currently, we are studying how this training impacts LAs’ facilitation practices through pre- and post-training comparisons.

Example publication: Carlos, C. M. L., Maggiore, N. M., Dini, V., Caspari-Gnann, I.* (2023). Characterizing facilitation practices of learning assistants: An authoritative-to-dialogic spectrum. International Journal of STEM Education, 10, Article 38. https://doi.org/10.1186/s40594-023-00429-4

In-the-moment learning and its facilitation do not occur in a vacuum, they are impacted by other aspects of educational systems such as course design. Educational systems can be understood in terms of different scales (e.g., society, classroom, interaction) and core elements (power, instruction, learning). For example, course design is situated at the classroom scale and within the instruction element. Most research has its primary focus on one scale and one core element, thus there is little understanding of how these scales and core elements impact each other. To address this gap, we study the impact of power, course design, instructor facilitation, and in-the-moment learning on each other. To date, our research in this area has produced important insight into the complexities of how educational systems impact student in-the-moment learning, including: a model that explains how class design impacts instructor facilitation and, in turn, how instructor facilitation impacts in-the-moment learning; and empirical evidence for the impact of different chemistry problem designs on in-the-moment learning.

In addition to its immediate impact on course design, this project also pioneers new methodology to allow combined analysis of intersecting scales and core elements of educational systems. First, we combined different data strands that give complementary insights into learning as it occurs in the classroom: LA-student interaction videos, whole-class video recordings, instructor slides, researcher field notes, and semi-structured stimulated recall interviews with professors, LAs, and students. Second, we analyzed the data for each aspect of the educational system with an appropriate analytical framework. Third, we investigated quantitative correlations between different core elements and scales and proposed models to explain how and why these correlations might exist. We anticipate that this novel type of analysis, using a combination of data strands with multiple different analytical frameworks to investigate correlations, could also be used by others to further advance understanding of educational systems.

In our study of class design and its impact on LA-student interactions, we considered curriculum, course guidelines, and instructor/LA/student roles in the classroom holistically. In a parallel study, we looked more specifically at how chemistry problem design impacts student in-the-moment learning. In this study, we used design research methodology to compare the impact of different problem designs in two semesters of a “Mechanistic Reasoning in Organic Chemistry” class for graduate and upper-level undergraduate students (Eckhard et al., 2026). We compared video-recorded whole class discussions of single-case and case-comparison tasks for proposing complex mechanisms (Fig. 4). We found that single-case tasks led to more instructor agency while case-comparisons led to more student agency. In addition, we found that students who solved case-comparison tasks made their reasoning more explicit when speaking, expressing more interconnected ideas and providing specific evidence to back up their reasoning.

This work is the first direct comparison of these alternate problem designs. Comparing in-the-moment student learning across two different semesters, we directly demonstrated that use of these different problem types produced differences in student in-the-moment learning. This basic research provides direct insight for improving student learning in upper-level organic courses. Our study provides instructors with a set of problems that promoted students to engage in mechanistic reasoning in an agentic and coherent manner. 

Example publications: Eckhard, J., Scheck, R. A., Caspari-Gnann, I.* (2026). Enhancing students’ agency and coherence in organic chemistry through transformed problem design. Chemistry Education Research and Practice, 27, 423-460. https://doi.org/10.1039/D5RP00268K

Maggiore, N. M., Karch, J. M., Dini, V., & Caspari-Gnann, I.* (2026). Why do they do what they do? A model that describes and connects the drivers of learning assistant facilitation practices. Science Education, 110, 947-973. https://doi.org/10.1002/sce.70048

Maggiore, N. M., Powers, K. P., Lwanga, K. L., & Caspari-Gnann, I.* (2024). The impact of learning assistant facilitation practices on student in-the-moment learning. International Journal of STEM Education, 11, 46. https://doi.org/10.1186/s40594-024-00506-2

We gratefully acknowledge our funding sources, the National Science Foundation awards #2440934, #2417138, #2000603, The Camille and Henry Dreyfus Foundation TC-26-009, and Tufts Springboard. Any opinions, findings, conclusions, and recommendations expressed in this material are those of the authors and do not necessarily reflect the views of our funders.