Since October 2010, the Chemistry-Biology Combined Major Program, an international course taught in English at Osaka University, has been teaching small classes (no more than 20 in size). We present data from the Force Concept Inventory (FCI) given to first-year classical mechanics students (N = 47 students over three years) pre and post score, for a class that predominantly uses interactive engagement, such as MasteringPhysics. Our findings show a G-factor improved score of about ~0.18, which is marginally about the average of a traditional-based course. Furthermore, we analyze in detail a set of six questions from the FCI, involving the identification of forces acting on a body. We find that student answers tend to cluster about "polarizing choices"--a pair of choices containing the correct choice and a wrong choice, with the latter corresponding to a superset of forces in the former. Our results are suggestive that students have a good idea of the right set of forces acting on a given system, but the inclusion of extra force(s) brings about confusion; something that may be explained by misleading ontological categorization of forces. In an appendix A we also comment on possible correlations between the pre/post score and the level of English ability on entry to the course.