It is defining an interpretation, and I personally don't like that way. Allow me to explain it.
Definition:
An interpretation is a pair ($\mathbf{M}$, $j$), with $j$ being the assignment and $\mathbf{M}$ being the model. j takes elements like functions, relations, and constants from a language $\mathcal{L}$ and "maps" them into elements in the universe of the interpretation (or domain of), and they're usually denoted as: $c^{\mathbf{M}}$, $F^{\mathbf{M}}$, $R^{\mathbf{M}}$ (unimaginatively). I explained in the comment what they exactly mean by turning it into a function that takes the tuples and maps them into elements in the domain of the model. An interpretation may also be called a structure.
Examples:
($i$): The language of group theory has $e$ as a constant which will be assigned to $e^{\mathbf{G}}$, and $\circ$ which is some arbitrary binary operation (predicate constant), which will be assigned as $\circ^{\mathbf{G}}$, and at the end we have the universe (domain) $\mathcal{G}$, and our interpretation looks like this: $\mathbf{G}=(\mathcal{G},\, e^{\mathbf{G}},\,\circ^{\mathbf{G}})$. Now, how is this useful? Fill in the gaps! ($\mathbb{R}$, $\cdot$, $1$) for example. Now you've multiplicative group.
($ii$): Take the language of graphs which has $R$ as its predicate, and so will will have $R^{\mathbf{Gr}}$. Define a domain $\mathcal{R}$, and you're done. The interpretation $\mathbf{Gr}$ = ($\mathcal{R}$, $R^{\mathbf{Gr}}$). You can replace $R^{\mathbf{Gr}}$ with the Cartesian relation for example, and leave the reals as the domain: ($\mathbb{R}$, $\times$), and you've a Cartesian graph.
The comment:
They define $I$ to be an interpretation (a model) which takes anything from $D^n$ to $D$ which is the domain of discourse over an $n$-arty input. Takes anything from constants, predicates, and functions, and puts them into the model with their own assignments in the universe (or domain) of the model. I personally don't like the way that they're teaching it, but that's the answer for your question. In this case, "$I$" is supposed to be understood, in a meta-semantics sense, as addition.
Note that loosely speaking.. this is not exactly correct, and plays around with some words in a misleading way. That's why I suggest seeing the explanation that I wrote.
Lastly an interpretation is of a language, and a model is of a theory. Switching them around is "fine" because they're the "same structure" but it is not recommended, and should only be left for purposes of intuition.