The Two-Stage Evaluation (TSE) model is motivated by revisiting Ellsberg's paradox that has been challenging the Savage's Subjective Expected Utility (SEU) model for decades. Besides accommodating preferences described in Ellsberg's paradox, it also describes Machina's paradoxes in more recent decision theory literature. The TSE differs from other existing descriptive models for decision making under ambiguity by treating uncertainties of risk and ambiguity differently in SEU framework, which belongs to the family of source-dependent preferences documented extensively in behavioral economics literature. Beyond its application in static settings, the TSE can also be extended to dynamic setting while maintaining dynamic consistency. Therefore, one can apply the standard "rolling back the decision tree" approach to solve a problem if the preference is described by such a model.