If you merely observe thermometers and which days feel hot, you'll see a link but won't be able to tell which way causality goes. If you set the thermometer to

If you merely observe thermometers and which days feel hot, you'll see a link but won't be able to tell which way causality goes. If you set the thermometer to a high value in a cold room and notice that you don't get any warmer, you'll quickly realize which way causality flows. So I agree with the GP post, I want to see them modeling & testing causality here.