Poster Q14 - Exploring Undergraduate Bioinstrumentation Laboratory Data Analysis with both Mathematical Software and Generative AI

Introduction: Generative AI (Gen AI) tools are becoming more widely available and have increasing functionality. Students are beginning to integrate AI into their current practice and will need to be able to ethically use AI as professionals in their future careers. The act of writing an accurate and well-organized lab report is the foundation of undergraduate science education. However, the time-consuming nature of this process has students searching for quicker alternatives, such as Gen AI. From simple calculation aid to drafting whole lab reports, Gen AI has become a frequent tool for students. Instead of banning AI in an undergraduate instructional laboratory course, an assignment was adapted to walk students through a comparison of analysis by hand, with mathematical software, and with Gen AI. The goal of the updated assignment was for students to become able to critically consider the difference between analyzing by hand, with purpose-built and validated software, and with a generic tool based on a large language model.

Materials and

Methods: This assignment is from a biomedical instrumentation laboratory required for students in the third year of a bioengineering program at a large public university in the Midwestern United States. During the course, students design, build, and test several circuits to measure biosignals. In this laboratory experiment, students built two circuits to measure temperature: one with a silicon band gap sensor and one with a thermistor. After confirming the circuits could measure the room temperature, students placed a cold pack on both temperature sensors and used an oscilloscope to record the voltage change from room temperature to the cold pack temperature, then exported the data as a CSV file using Keysight BenchVue, the software from the oscilloscope manufacturer.

In the post-laboratory assignment, students used equations derived from sensor datasheets to convert voltage to temperature. They then estimate a curve fit by hand using a plot to estimate the time constant and temperature change. From this, they fit an exponential curve to the data. Students then generated the same curve fit for the data twice more, first using MATLAB and again with a Gen AI tool of their choice. Students then compared the resulting equations and the relative ease of obtaining an equation for each method. The post-laboratory reflection asked students if they understood the difference between the two temperature sensors and what they learned from comparing techniques to generate mathematical models. The assignments were analyzed to identify which tools the students used, their method preference, and to evaluate the assignment updates.

Results, Conclusions, and Discussions: All student teams were able to complete the post-lab assignment, deriving curve fit with all three methods: by-hand calculation, MATLAB Curve Fitter App, and the Gen AI tool of their choice. However, the approach and takeaways from the assignment varied. The majority of students (82.3%) utilized a version of ChatGPT. One or two groups used each of Claude AI, Copilot, Curve.Fit or Julius AI.

Each of the three methods received about equal rankings for easiest and most difficult. However, the students ranked complexity differently, with Gen AI and by hand methods creating simpler models, while MATLAB was overwhelmingly chosen as the most complex model. Since each method fits a curve to the same data, all three equations should have the same level of complexity. However, the by-hand and Gen AI options required more user input and finessing to increase the fit of the model than the purpose-built Curve-fitter app in MATLAB.

Student responses also revealed both frustration and interest in the integration of Gen AI in the assignment. Many students expressed frustration with the amount of time it took to use Gen AI for curve fitting. They mentioned not knowing enough about Gen AI to understand how to approach certain issues or finetune their model. For instance, one student noted that the use of Gen AI was “definitely challenging for us and not similar to anything I have done before in my other classes.” Students also mentioned that it was difficult to understand each AI tool’s limitations and found adjusting the AI model “finicky.” On the other hand, many students also had positive opinions, noting that “generative AI is something I have been wanting to explore more, so this was an intriguing part of the lab for me.”

A key piece of student feedback was that “it should not be an assumption that students already know how to use this technology with no training and no additional information from the lecture slides and lecture that was provided.” More resources will be provided to assist students in getting started with both MATLAB and Gen AI in future semesters.

Acknowledgements (Optional):