The present paper discusses two experimental designs exploring the relationship between risk assessment and emotion recognition. The first uses publicly available data from the Human Connectome Database preprocessed and parcellated according to the Shen-268 atlas. The proposed second design would use novel data to refine and contextualize the results from the first. The paper proposes an integrative approach to models for human empathy relying on frontosubcortical and motor connectivity. It opens with an exploration of the literature on automatic and cognitive models for empathy. Specific attention is devoted to the overlapping functional networks involved and to the evidence that both pathways have emerged from a need for performance under threat. Both designs are intended to reinforce the argument that functional pathways involved in risk assessment offer a link between automatic and cognitive emotion recognition pathways. For both paradigms, connectome-based predictive modelling is used to predict scores on the HCP gambling task and the HCP emotion recognition task from 7T fMRI data. Separate analyses are conducted for whole-brain and frontosubcortical activity. It is hypothesized that significant nodewise correlations can be used to predict behavior for both tasks in all conducted analyses (p ≤ .01). In the conducted design, the model did not significantly predict participant behavior on the gambling task or the emotion recognition task. Potential confounds and future directions for empathy research are discussed.