Saying “interaction” suggests that you are looking for the effect of disrupting two genes at the same time. However, the CRISPR experiment is assuming that each gene is disrupted independently. That is to say, the infection rate is low enough that cells should only be receiving at most a single guide, and so only one gene is being disrupted.
I believe I’ve seen at least one paper which used pre-existing mutations that have deactivated genes in combination with an additional gene being knocked out. However, such an approach is going to be reliant on the having sufficient samples with mutations in the specific genes you want to assess for interaction scores.