Recall the formal definitions as given on the slides:
* pre∃(Q2) := {s1∈S | ∃s2.(s1,s2)∈R ∧ s2∈Q2}
* pre∀(Q2) := {s1∈S | ∀s2.(s1,s2)∈R -> s2∈Q2}
* suc∃(Q1) := {s2∈S | ∃s1.(s1,s2)∈R ∧ s1∈Q1}
* suc∀(Q1) := {s2∈S | ∀s1.(s1,s2)∈R -> s1∈Q1}
Computing these operations manually can be done by the help of the following two functions that compute the predecessors and successors of a single state only (and therefore we do not have to distinguish ∃ and ∀):
* pre(s2) := {s1∈S1 | (s1,s2)∈R}
* suc(s1) := {s2∈S2 | (s1,s2)∈R}
With the help of these two functions, we can rewrite the original definitions of the predecessor and successor functions as follows:
* pre∃(Q2) := {s1∈S | suc(s1) ⋂ Q2 ≠ {}}
* pre∀(Q2) := {s1∈S | suc(s1) ⊆ Q2}
* suc∃(Q1) := {s2∈S | pre(s2) ⋂ Q1 ≠ {}}
* suc∀(Q1) := {s2∈S | pre(s2) ⊆ Q1}
Now, look at your problem. The formula <>phi means that phi holds in one of the successor states that we consider, and that means that if we have the states Q where phi holds, then we have to compute pre∃(Q).
So you want to compute pre∃({s0,s1}) = {s∈S | suc(s) ⋂ {s0,s1} ≠ {}}. Thus, you need to find the states s where s1 or s0 belongs to suc(s). Looking at your table, I see {s0,s1,s3,s7}.
Also, you want to compute pre∀({s0,s1}) = {s∈S | suc(s) ⊆ {s0,s1}}, i.e., those states s whose set of successors is within {s0,s1}. Looking at your table, I see {s3,s4,s5,s7}.