In this paper we present the structure and implementation of a general customizable recommender system designed specifically for nonprofessional users. The recommender system (LSPrec) is based on the Logic Scoring of Preference (LSP) decision method. Our goal is to develop LSPrec as a decision-making aid that can be used without any user preparation. LSPrec users can systematically develop sophisticated decision criteria using only the natural verbal communication and a commonsense logic. The mathematical infrastructure of LSPrec is a graded propositional logic. To develop decision criteria, LSPrec exploits human capability to verbally specify the intensity of four fundamental percepts: truth, importance, simultaneity, and substitutability. LSPrec acts as a guide in a dialog with users, asking them to provide answers to questions about alternatives or objects that they want to evaluate and compare. That includes the necessary suitability attributes, their importance, and logic conditions that must be satisfied. The collected answers are then combined to build the LSP criterion function that is used for evaluation and comparison of multiple objects or alternatives. In a special case, LSPrec can be used for evaluation of a single object. The resulting decision models are saved so that users can visit them multiple times for updating, refining, and multiple evaluations. The results of decision making are recommendations presented to users in both verbal and numeric forms.