Reeq : Testing and Mitigating Ethically Inconsistent Suggestions of Large Language Models with Reflective Equilibrium | Synapse