Design, rather than medium, ultimately predicts learning outcomes, but the game-based learning literature has had difficulty successfully linking game design decisions to learning behaviours and outcomes. The current research investigates how explicit game design strategies can promote productive negativity (i.e. learning from failure), which has been identified as an important mechanism in both gaming and learning. We performed a randomized controlled trial with undergraduate biology students to investigate how game design might facilitate misconception resolution about random molecular behaviour through productive negativity. Students engaged with either a computer-based interactive simulation (n=20) or serious game (n=20) for 30 minutes, while their computer screens were recorded and click-stream data collected. We described in detail the theoretical framework underpinning our serious game and simulation using the Activity Theory Model of Serious Games (ATMSG); qualitatively coded and analysed video recordings of gameplay; and visually overlaid this data with the ATMSG models to draw conclusions about how game-design decisions influence learning-related behaviours. We found that the serious game resulted in significantly more productively negative experiences, while the interactive simulation allowed for greater exploratory or experimental behaviours. Based on our analyses of the qualitative gameplay data, we were able to recommend three game design strategies to enhance the occurrence of desired game-flow loops (e.g. productive negativity) with respect to an ATMSG framework: 1) including additional game mechanics on the primary game-flow axis of the ATMSG framework (i.e. mandatory interactions) limits the exploratory nature of the application; 2) integrating two or more primary-axis mechanics in a game-flow loop increases the frequency of interaction with this loop; and 3) gameplay loops that involve mechanics that fall off the primary-axis (i.e. nonmandatory mechanics) occur less frequently than those which involve primary-axis (i.e. mandatory) mechanics. This study is one of the first to successfully make direct comparisons between students' interactions in a game and a non-game application to provide concrete and actionable serious game design recommendations.