Confronting the system: Can modeling practice facilitate systems-based teaching and learning in college-level biology?

Background/Question/Methods

The ability to think and reason like a scientist is not innate, but develops through iterative practice, reflection, and feedback. Research in the learning sciences has shown that engaging students in core practices that reflect the work of scientists (e.g., modeling, argumentation, and analysis) promotes critical thinking skills and scientific habits of mind. However, most college-level biology classrooms do not reflect the work of practicing biologists. This contrasts with a widespread view that our historical mode of passive information delivery cannot sustain learning in a field growing increasingly complex, multi-variate, and systems-based.

We developed a pedagogical approach for introductory biology that uses system models as a way to teach, learn, and assess students’ understanding in introductory biology.  System models are adapted from Goel’s Structure-Behavior-Function Theory in which structures (model components, represented in boxes) are linked by behaviors (arrows representing relationships between pairs of structures) that collectively describe a system’s function (role or purpose).  A specific objective of our approach was to engage students in iterative construction and revision of a “core” model that would explain relationships among concepts linking genetics and evolution.  Early in the semester, students built simple models linking a few key genetics concepts (e.g., DNA, gene, chromosome), then progressively revised their models to integrate new concepts as they learned them in class (e.g., mutation, allele, fitness).  

Results/Conclusions

In an introductory biology class for majors (n=368 students), we analyzed student models at three times during a 15-week semester: Quiz 2 (week 4), Midterm Exam (week 8), Final Exam (week 15).  We explored trends in both complexity and correctness for students’ models overall, as well as within low-, mid-, and high-performing achievement tritiles (determined by incoming GPA). Students in all 3 tritiles improved in model correctness with each successive assessment (p<.001), but lower-achieving students showed the greatest relative gain, reducing the performance gap from 21% at Quiz2 to 13% at the Final. 

Model complexity was measured as web complexity index (WCI), which measures the proportion of structures with >1 associated behavior.  WCIs equal to zero indicate a linear model; values near 1 indicate high connectivity.  Trends were similar across tritiles with complexity significantly increasing from Quiz2 to Midterm (p<.001), but then significantly decreasing by the Final (p<0.02).  We hypothesize that this pattern may be explained by students’ construction of more parsimonious models later in the semester containing more biologically correct connections and fewer instances of irrelevant information.