Examining Students' Mathematical Evidence in Claim-Evidence-Reasoning Explanations During Science Inquiry Contexts (Poster 10)

• In Event: Artificial Intelligence–Augmented Assessment Tools to Support Instruction in STEM

Abstract

Introduction
Using mathematics and constructing explanations are practices outlined by the NGSS (2013). However, students struggle with mathematizing in science: determining the mathematical relationships between data (linear; Lai et al., 2016; Shah & Hoeffner, 2002), understanding the components of equations (slope; Nixon et al., 2016, Planinic et al., 2012), applying best-fit lines to data (Casey, 2015), and generating explanations about covariational relationships (McDermott et al., 1987; Sokolowski, 2019). These are critical barriers to high school science (Basson, 2002; Sadler & Tai, 2001), and “plugging and chugging” rote formulas fail to develop deep understanding of the DCIs expected by the NGSS (Brandiet et al., 2018). Thus, students need to be supported in using mathematics in science inquiry contexts so that they can develop deep understanding of phenomena.
Methods
84 students (Table 5) completed an [ITS] virtual lab that involved mathematical modeling (Authors, 2021) to determine how the mass of a sled going down a ramp affects the momentum of the sled when it reaches the end of the ramp (Figure 6, Table 6).
We developed and refined a fine-grained rubric for Claim-Evidence-Reasoning (CER; McNeill et al., 2006) statements that assess the mathematical evidence provided in students’ science explanations in a virtual lab in [ITS] (Authors, 2013; Figure 7; Table 7) and substantiated the rubric elements using prior literature on explanations (Authors et al., 2017; Sokolowski, 2019). Four of the authors independently coded the first 10 sets of responses for each rubric element; 91% agreement was reached, disagreements were discussed, and the agreed-upon codes were used for analyses. Each rater then coded their own assigned portion of the remaining sets of responses.
Results
In the students’ claims (Table 8), most (45%) described a covariational relationship between the variables; 27% described a mathematical relationship; only 18% described both the covariational relationship and the mathematical relationship (the ideal answer), and 10% did not describe either. When asked to substantiate claims with evidence, only 5% included the specific equation of the model they built in the lab, and 7% mentioned the fit of the model to their data. Regarding students’ reasoning statements, only 8% mentioned a scientific theory or concept that explains why their evidence supports their claim, which aligns with earlier CER research (McNeill et al., 2006). Overall, these results indicate that students struggle with understanding how to generate CER statements using mathematical evidence.
Scholarly Significance
The NGSS (2013) emphasize the importance of students using mathematics and constructing explanations to develop deep understanding of science. However, we found that students struggled with using mathematical evidence in the CER responses, which aligns with previous findings about students’ difficulties with mathematizing in science (e.g., McDermott et al., 1987). To be able to assess and support students with these intersecting practices, we must be able to operationalize these practices in a fine-grained and rigorous way. Our present work contributes to this goal and will ultimately inform future development of automated scoring and scaffolds of students’ scientific explanations involving mathematical evidence for [ITS].

Authors

• Amy Elise Adair, Rutgers University
  Presenting Author

• Rachel F. Dickler, Denver Zoo
  Presenting Author

• Janice D Gobert, Rutgers University
  Presenting Author

• Michael Sao Pedro, Apprendis
  Presenting Author

• Joe Olsen, Rutgers University
  Presenting Author

• Jessica A Owens, Rutgers University
  Presenting Author

• Christine Lott, University of Rutgers
  Presenting Author