AWK - Select columns in a variable not input records

Assume that you have an AWK program file with a complex procedure: example.awk. You want to assign specific columns of a variable, rather than a whole line of the current record.

#!/usr/bin/awk -f

BEGIN {...}
... {...}
...
END {
    ...
    var1 = ... # do something
    print "var1 ... " var1

    var2 = ⬛⬛⬛⬛⬛⬛
    print "var2 ... " var2
}

Note that var1 is independent from the file input. Run this program from a shell:

awk -f ./example.awk ./text.txt
...
var1 ... I like chocolate cake
var2 ... chocolate

As in this example, var2 has the string that is the 3rd field/column. In the UNIX shell, it might be equivalent to

var2=$(cut -d ' ' -f 3 <<< "$var1")

How would you script var2 in AWK?