Discovering Various Non-Newtonian Misconceptions with a Hierarchical Bayesian IRT Model

We analyze FCI data at the level of individual distractors using our exploratory IRT-based methodology(1) and discover 15 coherent, non-orthogonal student misconceptions, including some not previously identified in the PER literature. These misconceptions range from the expected (i.e., motion implies force) to the surprising (i.e., circular and linear impetus are distinct; accelerated masses travel on straight paths). All appear robust to sampling variability within our multi-school dataset, and most are present in both pre- and post-instruction data. We are also developing simple spreadsheet-based methods to analyze new FCI responses with the aims of (a) empowering physics teachers to monitor and adapt to the amount of each misconception harbored by their classes and (b) facilitating pre-post assessment of the effectiveness of instruction. Our code will be available on GitHub in the coming months.