We introduce a unified model of strategic voting in plurality elections, where uncertainty about poll scores is represented by convex sets of probability distributions over score vectors and deviations are evaluated via lower and upper expected utility gains. Focusing on belief functions, this framework subsumes both probabilistic and neighborhood-based models of uncertainty. In this setting, we establish general convergence conditions for iterative plurality strategic process and show how the model captures more realistic sources of uncertainty about voters, such as subsets of voters with uncertain ballots or incomplete preferences.