We present ab initio calculations of effective interaction parameters and for dioxides of actinides from uranium to curium. We first use a self-consistent scheme using and constrained random phase approximation (cRPA). For , and , we find self-consistent values of and leading to values of gap in agreement with experiments. For , the value of is underestimated. For and , we find very low self-consistent values. We compare projected local orbital Wannier functions to maximally localized Wannier functions and find a weak effect of the localization on interaction parameters. We suggest that spin-orbit coupling, and antiferromagnetism, could improve these results partially. We also extend our calculations by treating the bands from oxygen as correlated, as in Seth et al. [Phys. Rev. Lett. 119, 056401 (2017)PRLTAO0031-900710.1103/PhysRevLett.119.056401], and show that the results are rather independent of self-consistency in this approach. Comparing these calculations, our conclusion is that including electron interaction on oxygen orbitals is necessary both to improve the density of states and to compute more meaningful and predictive values of effective interaction parameters.