When power between two parties is structurally uneven — boss and report in a complaint against the boss, abuser and survivor, dominant and marginalized — mediation stops being a level table and becomes pressure on the weaker party to be "reasonable." Neutrality there is not neutral; it ratifies the existing power. The peacemaking move is to refuse the role, name why, and point to where real recourse lives.
The two refinements
"Know when to step out" looks simple. It isn't. Two ideas inside it do most of the work — and most failed versions of this move drop one or both.
Refinement 01
Stepping out without redirecting can leave the weaker party more stranded than they were. The move is not "I can't help you." It is "you need a different kind of help than I can give — and here is where it lives."
The destination is part of the sentence. Advocates. Investigators. Authorities with the power to act. Therapists trained in survivor support. Naming where to go takes the refusal from abandonment to redirection.
Refinement 02
Not every power difference is a reason to step out. Situational asymmetry — one party is more articulate, more prepared, angrier on the day — can be managed with caucuses, structured turns, longer pauses, advocates in the room.
Structural asymmetry is different. It is constitutive of the harm being mediated: the two parties' access to safety, livelihood, or autonomy is fundamentally unequal. No technique closes that gap, because the gap is the substance.
The research
Stepping out has a longer scholarly history than most mediators realize. Seven lines of work — across law, workplace research, survivor testimony, humanitarian practice, and the civil rights tradition — converge on the same conclusion.
The foundational case.
Trina Grillo's 1991 Yale Law Journal article The Mediation Alternative: Process Dangers for Women is the canonical text. Grillo argued that mandatory mediation in family law — especially with abuse present — became "the betrayal of mediation's promises," because it forced survivors to negotiate with abusers under a veneer of fairness. Her critique extended to workplace harassment, discrimination, and abuse cases, where "emotional vulnerability, social group responsibility, power imbalances, financial inequality, and discomfort with the mediation process" make the forum itself the problem.
Sources
"What Trina Taught Me" — Minnesota Law Review tribute to Grillo's work and its continuing influence.
"Grillo's Rigorous Path" — Oxford Academic, on the legacy of The Mediation Alternative.
The structural point.
A Queensland law review article puts the principle more sharply: in cases of severe asymmetry, "it is not the parties who are the problem because they are either too powerful or too powerless. Rather, it is the mediation forum itself which is problematic because it does not adequately reflect the power relationships of the parties and can instead be used by the perpetrator as a tool for the further abuse."
Source
QUT Law Review (PDF) — on mediation's structural failure in severely asymmetric cases.
The threshold.
A 2025 Journal of Law piece argues termination should be ultima ratio — used only when "strategic interventions cannot restore equality of conditions" through caucuses, advocates, or structured questioning. But "the presence of violence introduces a distinct and non-negotiable ethical threshold... mediation cannot proceed when fear, intimidation, or abuse deprive a party of genuine autonomy." The rule is not "always step out when power is uneven." The rule is: step out when no intervention you can offer will restore the weaker party's actual capacity to say no.
Source
"Safeguarding Self-Determination" — Journal of Law (Tbilisi State University) on mediation thresholds and termination.
The workplace version.
The Workplace Bullying Institute names the failure modes when mediation is misapplied to bullying or harassment: retribution, absence of trust, coerced settlements, fear of retaliation, "increased threats to personal safety post mediation." Mediation presumes rough symmetry. Bullying is not that.
Source
"Inappropriate Mediation" — Workplace Bullying Institute (PDF).
The survivor's account.
Testimony to the Maryland General Assembly against restorative justice in intimate partner violence cases shows why neutrality gets weaponized. One survivor wrote that she "would have done the process as a victim when I was in the relationship — because many victim-survivors believe that only by 'reforming' the perpetrator will we be safe. But my abuser (and most) would only have participated if he saw legal or societal benefit for himself." The table becomes a tool the dominant party uses.
Source
Maryland General Assembly testimony (PDF) — survivor testimony against restorative justice in IPV cases.
The broader critique of neutrality.
Bernie Mayer and Jackie Font-Guzmán's The Neutrality Trap: "An approach that says we should focus on peace and not justice is not only bound to fail, it is dangerous in these times." The humanitarian aid literature echoes the same caution: "There are deep moral compromises involved in adhering to neutrality, as it often means legitimising the power of those in control."
Sources
Beyond Intractability — Bernie Mayer and Jackie Font-Guzmán on The Neutrality Trap.
ODI Humanitarian Practice Network — "Neutrality: Principle or Tool?"
The civil rights register.
Two adjacent traditions reinforce the move. King's Letter from Birmingham Jail names the white moderate "who prefers a negative peace which is the absence of tension to a positive peace which is the presence of justice." The contemporary tone-policing literature names the mechanism: insisting the marginalized party adjust their delivery before the conversation can continue "prioritizes the comfort of the privileged person... over the oppression of the disadvantaged person." A mediator who keeps trying to "lower the temperature" between unequal parties is doing tone-policing in a chair.
Sources
Letter from Birmingham Jail — Google Arts & Culture archive.
"Tone Policing and the Assertion of Authority" — APA Blog (American Philosophical Association).
How to actually say it
The move is hard not because the principle is unclear, but because the actual sentence is hard to say out loud. Three versions for three contexts — same principle, different register. Each one combines refusal with redirect.
Institutional
"What you're describing isn't a conflict between equals — it's a complaint against authority. That needs HR, legal counsel, or an external investigator, not a mediator. I'd be giving the appearance of fairness to a process that isn't structured for it, and that would help the wrong party."
Personal
"I love you both. I won't sit between you on this, because what's happened isn't something the two of you can negotiate as equals. One of you needs someone in their corner — a therapist, a counselor, an advocate — not someone trying to balance you. I can be a friend to you. I can't be neutral on this."
Minimal
"This isn't a conflict. It's a harm. You don't need a mediator — you need an advocate. Let me help you find one."
Common objections
The move gets confused with two adjacent things it isn't. Worth holding the distinction.
It is not giving up on the relationship. Refusing to mediate between two people is not the same as refusing to care about them. You can stay in relationship with both parties while declining to sit between them in a process you cannot honestly hold. Often the relationship is what enables you to point them toward real recourse rather than just walking away.
It is not a permanent verdict on the parties. Structural asymmetry can change. A boss-and-report dynamic ends when one of them leaves the company. An abuse situation changes when the abused party has reached safety. Stepping out today does not mean stepping out forever — it means stepping out of this conversation, at this moment, when the structure of the room makes mediation a tool of the stronger party.
Why this is the discipline
"Know when to step out" is the move that keeps mediation honest. Without it, the other three moves — setting the frame, slowing the speed, translating — can be used to extract legitimacy for an outcome the structure has already determined. Refusal, when refusal is what's called for, is what makes the rest of the practice trustworthy.
The discipline, in one sentence
A mediator who never refuses is not a more useful mediator. They are a mediator whose presence can be conscripted by whichever party benefits from the appearance of a fair process.
The discipline is to hold both halves at once. The willingness to sit between two people who are roughly even in power and do the work of helping them see each other. And the willingness to refuse, by name and with a redirect, when the structure of the room makes that work a fiction.
The first willingness without the second produces complicity. The second without the first produces withdrawal. Both together, with the judgment to know which is called for, is what mediation actually is.