Interpreting Interaction Effects

I have ran interaction effects in Binomial logistic regression. I know how to interpret the interaction effect but I am having trouble interpreting the other variables in the model that makes up the interaction effect. Are the other variables held at 0 in the interpretation or are they held at the mean.

For example:
Education is coded as 0= college education, 1= HSD or GED, and 2=No HSD or GED

Homeless is coded as 0= not homeless, 1= homeless.

OR for Interaction term (HomelessXEducation) is 0.70
OR for Homeless is 3.36
OR for Education is 0.93
Well as you mention youself, in such models the effect is nonlinear so they have to be evaluated at certain points decided by the analyst. Therefore you choose for what value of Homeless and what value of education you want an answer. Suppose you want the answer for Homeless = 1 and education 0, then the interaction term is evaluated for 1 * 0 = 0.