Scott Ogawa has the floor:

Consider [a] heterogeneous group of people who have different internal temperatures. In the Summer, people who are really hot complain a lot to the building manager, since building is [the] only source of cool. People who are really cold do not complain so much since there is always the “outside option” (literally) as relief. Things switch in the Winter. A complaint-minimizing building manager will jack the heat up in the Winter, and the A/C in the Summer.
I have no data, and it is tough to trust how things “feel” since we are not the best judges of absolute temperature. Nevertheless, I have heard many folks say the temperature inside big buildings always seems negatively correlated with outside temperature, which is extra strange given this costs more than a positive correlation.

Scott’s solution is something like this:  since people differ in their hot/cold preferences you want some variation in the temperature inside.  Most buildings aim for uniformity.  If half of the building is warm and the other half is cool, people will pick their favorite side of the building.  Keeping the mean temperature constant but adding a mean-preserving spread raises overall welfare due to sorting.