If you're tired of studies where you inevitably get deceived, electric shocked, or tricked into developing a sexual attraction to penny jars, you might want to sign up for Brian Wansink's next experiment.

Some people would prefer being shocked to gaining ten pounds.

Generalizing these preferences, when it's possible at all, takes several complicated equations. Do I really want to discount gains more than losses, if I've never consciously thought about it and I don't consciously endorse it?

I think we need to differentiate between the utility function you might use to model your actions, and one you'd make for what you consciously think would be better. Essentially, akrasia is when the two contradict each other.

For the former, making a more complicated model would make perfect sense. For the latter, you'd probably pick something simpler, especially exponential or constant, so it doesn't require constantly changing your mind.