Functions that select individuals among a population use the Fitness of each individual to compare them.
In the documentation it's stated that this fitness comparation is made lexicographically. So when a multi-objective optimization is set up fitness of objective values will be evaluated following index as an order.
What's the reason to not aggregate weights and values into one single variable, so that all objective variables are taken into account?
$FitnessValue = \sum |weights| * values^{sign(weights)}$