Two of your best people can't be in the same meeting without it curdling. Each has told you, separately, that the other is impossible. You could pick a winner, but whoever loses will quietly disengage, and you'll have traded a visible argument for an invisible one. Mediation is the alternative: a structured way to help the two of them reach their own agreement, with you holding the process rather than imposing the answer.

The quick version

  • Dispute resolution is the broad family of ways to settle a disagreement, from a quiet chat, to mediation, to arbitration, to a courtroom. They sit on a ladder from informal and self-determined to formal and imposed.
  • Mediation is the one where a neutral third party helps the disputants reach their own settlement. The mediator owns the process; the parties own the outcome. Nothing is decided for them.
  • The engine that makes it work is shifting people off positions (what they say they want) and onto interests (why they want it), the central move in Fisher & Ury's Getting to Yes.
  • The trap is leaping to a fix before each side feels heard. Skip that, and the "agreement" you broker collapses the first time the pressure returns.

The idea in depth: a mediator runs the room, not the verdict

The defining feature of mediation is who decides. In The Mediation Process (first published 1986; 4th edition 2014), Christopher W. Moore, a founding partner of the conflict-resolution firm CDR Associates, defines mediation as the intervention of an acceptable, impartial third party who has no authority to make a decision for the parties, but helps them voluntarily reach their own. That last clause is the whole game. An arbitrator or a judge hears both sides and rules. A mediator does the opposite: they keep both sides talking until the two of them can author the answer themselves.

So the move, if you're the manager stepping in, is to consciously drop the instinct to adjudicate. The moment your two reports sense you're scoring the argument, they stop solving the problem and start litigating it to you, performing for the referee instead of negotiating with each other. Say it out loud at the start: "I'm not here to decide who's right. I'm here to help you two work out what you can both live with." It feels like giving up power. It's actually the only way to get an agreement that survives you leaving the room.

It helps to see where mediation sits among the alternatives, because the right tool depends on how much the parties can still self-determine. The field calls this spectrum alternative dispute resolution (ADR), the routes that sit between doing nothing and going to court.

flowchart LR
  A(["Direct talk
parties only"]) --> B(["Mediation
neutral helps them decide"])
  B --> C(["Arbitration
neutral decides for them"])
  C --> D(["Litigation
a court imposes a ruling"])
The dispute-resolution ladder: control over the outcome passes from the parties to a third party as you move right. Leaders Loop

Read left to right, two things rise together: cost and the loss of control. Direct conversation is cheap and keeps the decision with the people who have to live with it; litigation is expensive, slow, and hands the outcome to a stranger in a robe. Mediation is the sweet spot for most workplace conflict, more structured than a chat, far cheaper than the formal end, and crucially, the parties still own the result. The practical reading: don't escalate further up the ladder than you have to. Reaching for arbitration or a formal grievance when a mediated conversation would do is a way of buying certainty with a relationship.

Positions vs interests: the move that actually unsticks people

Here is the mechanism underneath every successful mediation. People arrive stating positions, fixed, incompatible demands. "I need the corner desk." "I'm not changing the deadline." Positions clash by design; if both could be satisfied at once, there'd be no dispute. The breakthrough comes from going beneath them to the interests: the needs, fears and motives that produced the position. This is the heart of Getting to Yes (Roger Fisher and William Ury, first edition 1981; Bruce Patton was added as co-author from the 1991 edition), the foundational text from the Harvard Negotiation Project. Its counsel is plain: separate the people from the problem, and focus on interests, not positions.

The book's best-known illustration makes it concrete, a story the authors credit to the management thinker Mary Parker Follett. Two people quarrel over a library window, one wanting it open, one wanting it shut, deadlocked between two positions. Someone asks why, one wants fresh air, the other wants to avoid a draught, and opens a window in the next room. Fresh air, no draught. Neither stated position was met; both interests were. So the mediator's core question isn't "what do you want?" but "why does that matter to you?" Asked with genuine curiosity, repeatedly, it converts a tug-of-war into a shared puzzle. The desk dispute is rarely about the desk; it's about feeling valued, or needing quiet to concentrate, and those can often both be met without anyone winning the desk.

A position is what someone demands. An interest is why they demand it. Mediation lives in the gap between the two.

An honest limitation. Interest-based mediation assumes the parties are negotiating in reasonable faith and have roughly comparable power. Where one side is bullying, harassing or holding all the leverage, "let's explore everyone's interests" can become a polite cover for coercion, and pushing two people to "work it out" may quietly relegitimise misconduct that should go to a formal process instead. Some disputes belong with HR, a grievance procedure, or a lawyer, not a mediator. The skill isn't only running a mediation; it's recognising the disputes that aren't yours to mediate.

The stages: structure is what makes neutrality possible

Mediation looks improvised when it's done well, but it follows a shape. Moore's work breaks the process into a sequence of stages: the mediator first does the unglamorous groundwork, establishing they're acceptable to both sides, agreeing how the conversation will run, and setting ground rules, before any substance is discussed. Only then do the parties tell their stories, surface the real issues, generate options, and test them against each other until something holds. Practitioners often compress this into a handful of steps, but the order matters more than the count.

flowchart TD
  A(["1 · Set the stage
neutrality, ground rules,
'I won't decide for you'"]) --> B(["2 · Hear each side
uninterrupted, both feel heard"])
  B --> C(["3 · Find the interests
ask 'why does that matter?'"])
  C --> D(["4 · Generate options
widen before you narrow"])
  D --> E(["5 · Agree + write it down
specific, who-does-what-by-when"])
A working sequence for an informal mediation. The early stages feel slow; skipping them is why later ones fail. Leaders Loop

The temptation is always to fast-forward to the options stage, because that's where the resolution lives. Resist it. The reason the early stages exist is that people can't problem-solve while they still feel unheard, they keep circling back to re-litigate their grievance. Letting each side speak fully and uninterrupted, while the other genuinely listens, is not a courtesy; it's the thing that frees them to move on. So the move is to spend longer than feels comfortable on stage two, and to actively protect it: "Let him finish, you'll get the same uninterrupted time." Only once both sides have visibly relaxed does brainstorming options stand a chance.

It also helps to know which mode each person defaults to under pressure. Kenneth Thomas and Ralph Kilmann's Conflict Mode Instrument (introduced 1974) maps conflict behaviour on two axes, assertiveness and cooperativeness, yielding five modes: competing, accommodating, avoiding, compromising and collaborating. Mediation is, in effect, an attempt to move two people toward collaborating (high on both axes) when their instinct is to compete or avoid. Naming the pattern, gently noticing that one person keeps conceding while the other keeps pushing, can itself shift the room, because most people don't realise they have a default at all.

A worked example

Two designers, Priya and Sam, share a project. (Illustrative scenario; the people and figures are invented for teaching.) Priya says Sam ignores agreed deadlines and "does whatever he wants." Sam says Priya "controls everything" and won't let him near the interesting work. They've stopped reviewing each other's drafts. Their manager, Dana, could simply rule, assign owners, set hard dates, and would buy a fortnight of compliance before the cold war resumed.

Instead Dana mediates. She opens by naming her role: she isn't deciding who's the better designer; she's helping them find a way to work that they can both stand behind. She gives each five uninterrupted minutes. Priya, heard properly for the first time, admits the deadlines matter because she's the one who fields the client's anger when slips happen, her interest is protection from blame, not control. Sam, asked why the "interesting work" matters so much, says he's worried he's stalling and has nothing to show at his review, his interest is visible growth, not chaos. Neither interest is what their position implied.

flowchart TD
  P(["Priya's position:
'Sam ignores deadlines'"]) --> PI(["Interest: not be
blamed by the client"])
  S(["Sam's position:
'Priya controls everything'"]) --> SI(["Interest: visible growth
before his review"])
  PI --> O(["Option that meets both:
Sam owns a feature end-to-end,
with a shared client-facing tracker"])
  SI --> O
Once the interests surface, an option appears that neither position could have produced. Leaders Loop

With the interests on the table, the option almost designs itself. Sam takes full ownership of one feature, real, reviewable work for his appraisal, while both adopt a shared, client-visible tracker so deadline risk is flagged early and Priya is never again the sole shock-absorber. They write it down: who owns what, the tracker goes live Monday, and they'll check in at the end of the sprint. Dana didn't impose a desk-rearrangement of duties; she surfaced two needs the argument had buried, and let the two of them build the fix. That's the difference between a settlement that holds and one that's just a ceasefire.

Frequently asked questions

What's the difference between mediation and arbitration?

It's about who decides. A mediator helps the parties reach their own agreement and has no power to impose one, if they can't agree, there's no outcome. An arbitrator, like a private judge, hears both sides and makes a binding decision for them. Mediation keeps control with the people in the dispute; arbitration hands it to the third party. Mediation usually comes first because it's cheaper, faster, and preserves the relationship.

Can I mediate between my own team members, or do I need someone neutral?

A manager can run an informal mediation, and often should, it's a core leadership skill. The catch is neutrality: if you have a stake in the outcome or visibly favour one person, you can't credibly hold the process, and the parties will sense it. If the conflict involves you, is deeply entrenched, or carries legal or disciplinary weight, bring in a neutral third party, another manager, HR, or a professional mediator.

What if one person just won't engage?

Mediation is voluntary, and a party who refuses to participate in good faith can't be forced into a genuine agreement. First, find out why privately, reluctance is often fear of being ganged up on, or doubt that anything will change. Reaffirm that you won't decide for them and that the process is confidential. If they still won't engage, mediation may not be the right tool, and you may need to move to a more formal route, but go in knowing that's an escalation up the ladder, not a failure of nerve.

Isn't this just "compromise," where everyone half-loses?

No, and the distinction is the whole point. Compromise splits the difference between two positions: each side gives up half of what they asked for. Interest-based mediation tries to satisfy what's underneath the positions, which can sometimes give both sides most of what they actually needed (the open window in the next room). Compromise is the fallback when interests genuinely conflict; it shouldn't be the first move, because it leaves value on the table.

How do I keep my own opinion out of it when I think one side is clearly right?

You probably can't keep it out of your head, but you can keep it out of the room. Your job in mediation is the process, not the verdict, the moment you tip your hand, the "losing" party stops negotiating and starts defending. If you genuinely believe one side is in the wrong on a matter of conduct or policy, that's a sign this is a disciplinary or performance issue, not a dispute between equals, and it belongs in a different process, not a mediation.

Related in the Toolkit

Mediation is negotiation done on behalf of two other people, so the bargaining toolkit underneath it matters, the same logic of interests, BATNA and the zone of agreement drives a mediation as surely as a deal. And because a manager rarely has the formal authority to impose a settlement that sticks, the real lever is influence without authority: getting two reluctant people to move because they choose to, not because you ordered it.

Where to go next