The Process, according to a policy wonk: identify policy goal; identify obstacles to goal (the problem); explore and become knowledgeable about the nature of the problem (causal factors, moderators, mediators, interactions); explore possible solutions to problem (pros/cons, trade-offs, incentives/disincentives, consequences, impact on other policy goals, etc.); identify the type of data needed to assess effectiveness and desirability of each solution; set up data collection and analysis system; experiment with possible hypotheses/solutions; analyze findings; refine hypotheses; tweak or reject solutions and experiment again with remaining options, ideally in different conditions (as effectiveness may depend on local contexts).
A few general principles: every step of the way acknowledge possibility of being wrong about the nature of the problem and what works; encourage diverse input on goals, problems, and solutions; when possible, engage in small-scale experimentation with solutions (the better to compare relative effectiveness of different proposed solutions and to minimize collateral damage of implemented bad solutions).
While values and emotions necessarily inform policy goals, the Process is pretty cold-blooded. I’m thinking, as a rule, broad moral imperatives, like justice” or “fairness”, make pretty iffy policy goals – partly due to vagueness and partly because moral indignation tends to distort priorities, discourage open discussion, interfere with assessment of trade-offs and increase confirmation bias.