The simplest ethical framework is Kant's. From that point of view, which of these two agendas could one always imagine everyone followed, which would not ultimately cause you problems. The problem with asking a question in Kantian terms is that it is very important how the question is phrased, and yours is not clear enough.
If everyone follows the agenda that they favor their own benefit, there will be times when you are vastly outnumbered, and the results will be quite bad for you.
If everyone follows the agenda that they vote opposite to their own benefit, there will be times when you are in a vast majority, and the result would be even worse even more often, as more people will be equally badly hurt.
But since neither option can be made universal, this is probably the wrong way to look at what is actually going on.
What is another way to look at this distinction? Instead of being self-centered and voting for or against your own interest, you could think in terms of voting your own interest, or choosing someone at random who differs from you in some important aspect, in whose interest to vote. I think we could generalize that solution.
If the person in whose interest we should try to be motivated is randomly chosen, this does maintain a bias toward benefitting majorities. But overall, people feel OK with the idea that what is good for more people should happen more often, as long as this cannot become a form of systematic oppression.
Since the population differs a great deal in many dimensions, choosing someone else based on different criteria each time, does not create a systematic bias against the overall majority, and it breaks down correlations between advantages that traditionally lead to a single group benefitting over and over again.
In addition, it would lead to something else that could, over time, become quite positive. In order to understand when you are voting in the interest of some other randomly chosen person, you would have to actually actively empathize with that person, and discern the effects on them. Understanding how different people, about whom you would not ordinarily think, might be affected, would improve your ability to see how benefits to given groups function, and to better make correct determinations about how things might affect your target beneficiary.
This is a bizarre and rigid way to make decisions, but it offers a perspective on what is really required. What we need in a voting system, from a Kantian view, is empathy across divisions. Taking the weirdness out of the formula, we should consider a number of others, very different from ourselves in different ways, as randomly as we are really able, and do what benefits that group, instead of choosing to vote for or against ourselves.