In statistics this type of data is called interval data - data where differences ("intervals") make sense but not necessarily other operations. This is the data-science way to say what Matthew Daly is saying in his answer.
Dates and times are a standard example used in statistics classes. For example, you can find the difference between two years:
This means that if you like, you can rearrange the equation to say things like
1999 + 22 years = 2021.
2021 - 22 years = 1999.
but similar expressions make no sense, for example
22 years - 2021 ??
1999 + 2021 ??
Other common examples of interval data, like temperatures, have the same issue but the issues are hidden by a subtle trick in the way we speak. For example, you can find the difference in temperature between an object that is 75° and another that is 60°:
But the "15°" is a different kind of thing; it is an interval, not the temperature of an object.
If I tell you to add this 15° (interval) to the 75° (a temperature), that works just fine.
But if I instead tell you I have two objects, where one is 75° (a temperature) and the other is 60° (a temperature), and then ask you to find the sum of their temperatures, you should feel some cognitive issue here -- can you really add two temperatures? It's never a useful or meaningful thing to do. You can't really add those two temperatures together in any meaningful way, just like you can't add "noon" plus "2 o'clock." This is interval data.
As far as I know, times and dates are the best example of interval data where this issue manifests clearly and does not have any linguistic tricks to obscure the issue.
So -- for interval data, you have two types of quantities: "measurements" and "intervals" (in Matthew Daly's answer he refers to these as two spaces). You can do the following things:
- Measurement - Measurement = Interval
- Measurement $\pm$ Interval = Measurement
But other operations don't make sense.