How one could determine that an interaction term is needed to make the linear model a better fit?

I am studying ISLR the book with applications in R.

​

The thing is, there's a section called "Extensions of the linear regression model" and I got the maths, how to interpret it and everything.

​

But the book stated this as common sense, generally, TV advertising is more likely to be pushed by radio advertising and vice versa.

​

How would one determine that an interaction term is needed in a dataset that is new to you, unfamiliar, with …

applications book datascience everything extensions linear linear model linear regression maths regression studying