There are a few things to keep in mind when controlling temperature.
The first thing to consider is the nature of the thermal system itself. Per volume, the heat capacity of water is approximately 3200 times more than the heat capacity of air. Taking the example of a 5 cu ft freezer with a 5 gallon batch, we have approximately an 8:1 ratio of air to wort. So the heat capacity of the wort is about 400 times the capacity of the air. Cooling the air temperature by 10 degrees and letting the system come to equilibrium will only lower the wort temperature by 0.025 degrees. So if we cycle the freezer based on air temperature, it will cycle many times before the wort temperature falls by 1 degree, depending on the deadband setting of the controller.
Now if we measure the temperature of the wort, either on the insulated surface of the fermenter or using a thermowell, the temperature of the wort will move very slowly, even if the air in the freezer goes below zero, due the differences in thermal mass. And once the temperature of the wort reaches the target temperature, the air will quickly warm up due to the low heat capacity. I am simplifying this a bit, because the inside surfaces of the freezer also have some heat capacity. However, in most cases these are still pretty small in comparison to the heat capacity of the wort.
@Brewer Jim uses the Coke bottle of water as a thermal mass. This is not a bad compromise. It's a smaller thermal mass, so the freezer will cycle more often than taping the probe to the fermenter, but not nearly as often as leaving the probe in open air.
Now, if you are using a glycol system with a cooling coil in a fermenter, the dynamics change significantly. The glycol/water solution has approximately 90% of the heat capacity of water, depending on the concentration and temperature. It is easier to overshoot temperature with this system and placement of the temperature probe is a bit more important.
Hopefully this information is useful without being too full of engineering and physics.