Localized arc-filament plasma actuators (LAFPA) have been demonstrated to have the control authority needed to fundamentally alter the development of supersonic turbulent jets and suppress their noise. Fundamentally, these actuators are thought to work via a Joule heating by the plasma, causing locally hot regions of fluid, but the mechanism by which this affects the instabilities of the jet is unknown. A simulation model of the near-nozzle region of an actuated jet is used to investigate this. The model matches the jet Reynolds number of corresponding experiments at The Ohio State University (OSU), which is important because the boundary layer interaction with the plasma actuator is expected to be central to the mechanisms. The model is also limited in that it is two dimensional, so direct correspondence with the experiments is not expected. However, it does include what are thought to be the key features of the flow: a localized rapid heating, a recessed groove in which the plasma sits, a thin boundary layer, and the early development and saturation of instabilities in the shear layers. It also matches some observations from the experiment. With the simulations, we show that the geometry of the groove, which was originally designed to prevent the plasma from being "blown off" by the flow, is of central importance for the actuation. The rapid expansion resulting from the heating causes an ejection of fluid from this cavity perpendicular to the shear layer that forms over it. Numerical experiments using reduced models of the actuator are designed to assess the respective role of the the cavity, the vertical ejection of fluid from the cavity, and the direct effect of thermal actuation on the flow. An actuator placed above the nozzle wall without a cavity is shown to have little effect on the flow. None of the reduced models match the mixing layer spreading or the flow organization caused by the the full actuator.