In this paper the proble of expert judgment (EJ) in Probabilistic Safety Assessment is recognised as a true knowledge problem. A critical analysis of traditional EJ methodologies leads to highlight some common weak points of theirs. The suggestion is made to reformulate the expert opinion issue in a Knowledge Engineering setting in order to reach a sound basis for comparison, integratio and justification of expert knowledge and relevant uncertainty. The proposed approach is described in its goals, and requirements for its development are detailed.