There are many aspects of life that contribute to one's well-being, such as food, housing, income, community... Suppose that each of these can be quantified. People rate these and get a vector in the real space of aspects of life. Then people combine the results into something that may be termed the index of happiness. If we use a linear model, then this means that we project the vector to a one-dimensional hyperspace (line).
Different people assign different weights to the contributing factors and arrive at disparate results... and blame us for whining excessively despite leading such a good life. :( Well, if we value one factor too much, then everything else does not count at all and then standard procedures will require that dominant factor regularised... In the over-arching unhappiness, we eke out a slightly less unhappy life.