They stand for the weights of those columns.
Start with the null hypothesis. In your case,
Ho: A1=A2
From this (and statistics [N-1*K-1]=3-1*4-1=6), you will see that you need 1 rows in your contrast:
A1=A2
Each of these gets converted to a contrast by making them equal to 0.
A1-A2=0
Now you need to build this contrasts. Start by building each part, the contrast for A1 and the contrast for A2. Each of these contrasts can be built from A1B1 and A1B2 and A2B1 and A2B2, respectively. In turn each of theses contrasts are the average of the subject specific contrasts.
Here is a more complicated example of how to build contrast from the simplest elements.
Here is an example of how to construct any contrast:
This is for a design with 18 subjects in group 1, 9 subjects in group
2, 2 group terms and 2 conditions: Start with the simpliest element,
single subject in a single condition, build its contrast, repeat for
all subjects and conditions, and then combine the ones you want.
S1G1C1=[1 zeros(1,26) 1 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0]
S1G1C2=[1 zeros(1,26) 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0]
....
Now average your G1C1 and by summing and dividing by the number of
subjects, you'd get
G1C1=[ones(1,18)/18 zeros(1,9) 1 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0]
and
G1C2=[ones(1,18)/18 zeros(1,9) 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0]
and
G2C1=[zeros(1,18) ones(1,9)/9 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0]
and
G2C2=[zeros(1,18) ones(1,9)/9 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0]
Now subtract G1C1-G1C2 AND G2C2-G2C1
G1C1-G1C2=[zeros(1,27) 0 0 1 -1 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 0]
and
G2C1-G2C2=[zeros(1,27) 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0]
Now subtract these two:
Interaction contrast=[zeros(1,27) 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 -1
1 0 0 0 0 0]
Hope this helps.