Long story short: Preferences orderings are over more abstract outcomes instead of money. It's still a story of maximizing expected utility.
Agents are usually assumed to have preferences over outcomes. This is assuming something like consequentialism. In financial models, the outcomes are dollar amounts and so there is a natural preference ordering. In something like a political model, outcomes are policies. So, people have indirect preferences over political candidates based on beliefs and seek to influence the outcome by voting. This is probably a bad model of voting, as most people don't honestly believe their vote matters I'm guessing.
Beyond voting to influence current decision-making, here is a good paper about voting as communicating: Piketty (2000)