Reflective Equilibrium

Reflective equilibrium justifies a value position by mutual adjustment: you hold considered judgments about particular cases, general principles, and relevant background theories at once, and when a case and a principle conflict you revise whichever has less credibility on reflection, iterating until the set coheres. It is the de facto method of normative philosophy and it has three failure modes that bite in a bounded session. This skill runs it honestly. It leads with that caveat, then forces the one thing a bare run omits and that makes a run worth reading at all: an explicit revision ledger that records which commitment gave way and why, instead of a quiet declaration that "coherence was reached."

Before you run this: the evidence caveat

Reflective equilibrium is tier C (conceptually plausible, philosophically central, empirically untested as a procedure). It has been the working method of normative philosophy since Rawls (1971), but no controlled study shows that performing it improves judgments, and three failure modes bite hardest in a single bounded session. (1) There is no externally checkable termination test: "equilibrium" cannot be distinguished from "I stopped looking for conflicts." (2) Garbage in, equilibrium out: Brandt (1979) warned the method can be "no more than a reshuffling of moral prejudices," and Singer (2005) sharpened this with evolutionary debunking, an objection that bites harder for an agent whose considered judgments are training-distribution intuitions. (3) A license to rationalize: because principles may be revised to fit cases, a motivated reasoner can quietly demote the principle that forbids a preferred outcome and report the result as a principled equilibrium. Even on its own idealized terms convergence is rare: Freivogel's (2023) simulations found different starting points reach identical equilibria only about 27 percent of the time. The revision ledger and the explicit honesty about these traps are the only thing that makes a run worth reading.

When to Use

• The user asks for reflective equilibrium by name and you should run it honestly rather than refuse.
• A stated value, policy, or principle collides with a strong judgment about a concrete case, and the user wants the conflict adjudicated bidirectionally (the case may revise the principle, not only the reverse).
• You can produce, and the user wants, an explicit revision ledger showing which commitment gave way, not just an assertion that the set now coheres.

When NOT to Use