Empathy Is Not a Concession: Get Their Situation Exactly Right, Discount Nothing
Topic: calibrated questions · the "that's right" moment · deliberate deference · black swans·4 moves · 3 diagrams
The line that most often changes a negotiation isn't your price. It's "let me try to say why this is hard on your side — tell me if I've got it wrong."
"Empathy" sounds like the softer party losing, so a lot of people deliberately go stone-faced at the table. That confuses two separate things. Understanding someone's situation and agreeing to their terms are completely independent. You can describe their difficulty with total accuracy and not move a single term. The difference is this: a person who feels accurately described starts giving you information, and a person who is defending just repeats their position — and the solution you need is sitting in the part they haven't said yet.
MOVE 01
Turn "no" into "then how am I supposed to do that"
Turn their refusal into a problem you solve together
calibrated questionshow / whatautonomy
The principle in one line
A calibrated question starts with "how" or "what," can't be answered yes or no, and hands the problem back to them. It doesn't collide with their position — it asks them to find you a workable path inside their own constraints.
In their words
"How am I supposed to do that?"— Voss, Never Split the Difference (2016)
One "no," two ways to take it
Scripts: four questions for almost any deadlock
All of them: unanswerable with yes or no, and free of accusation
· "How am I supposed to do that?" — the strongest line you have when their condition is one you can't accept. It isn't a refusal; it puts the impossible part where they can see it.
· "What matters most to you here?" — asks for weight, not for demands.
· "If we did it that way, what problems would that create on your side?" — surface the resistance now, instead of meeting it later as "let us think about it."
· "What's been the hardest part to push through so far?" — one question that often explains the entire stall.
In the room
Setting: you're a freelance designer. The client says, "budget is eight thousand, and revisions have to be unlimited — we have a lot of opinions internally."
✗ Give up both conditions at once
"Sure, eight is fine, and we'll be flexible on revisions." You just handed over price and scope in one move and got nothing back — and "unlimited revisions" isn't a deliverable condition at all.
✓ Take it, then hand the problem back
"Eight thousand, noted — and a lot of internal opinions genuinely is a pain. But there's something I need to ask you: how am I supposed to guarantee you have the finished work before the year ends if revisions are unlimited?" (Them: …honestly it doesn't need to be unlimited. It's that my boss might overturn the direction.) —— And there it is: "unlimited revisions" doesn't mean a hundred rounds, it means fear that the direction gets reversed — a process problem, not a count problem. The solution follows: three directions for the boss to choose from, then two rounds of refinement on the chosen one; a new direction is quoted separately. The price never moved and their actual fear got solved.
Why it works · mechanism and evidence
Why it works
Mechanism: a yes/no question splits the conversation into two opposing sides and lets them simply hold their position. A question starting with "how" can only be answered by pulling up their own constraints, and the act of stating them is the act of giving you information. It also leaves the decision in their hands — people have a strong preference for being the one who chooses.
Hard evidence · having a choice is rewarding in itself: Leotti & Delgado (2011, Psychological Science) found that merely anticipating that you will get to choose activates ventral striatum and other corticostriatal reward circuitry — even when the options are worth exactly the same. Handing the decision back isn't just polite; in their head it pays. Strength: peer-reviewed fMRI, small samples, and reverse-inferring function from regions needs care. (cross-ref neuroscience)
Hard evidence · asking more, especially following up, makes people like you: Huang, Yeomans, Brooks, Minson & Gino (2017, JPSP) ran several experiments plus a speed-dating field study and found people who ask more questions — particularly follow-ups on the previous answer — are rated as more likeable, and got more second dates. Multiple studies including field data.
Weaker evidence · the calibrated-question tool itself: it comes from Voss (2016), a practitioner's summary of hostage negotiation, with no controlled trials. What's credible is the mechanism behind it, not the script.
Boundary: fire off three or four "how am I supposed to"s in a row and it becomes evasion, or an interrogation — one or two well-placed ones per negotiation is plenty. The harder line: this is only honest when the constraint is real. Manufacturing helplessness you don't have is the same class of move as inventing an alternative you don't have. And when the other side is visibly hostile, absorb the emotion with one line of acknowledgement before you ask, or it lands as a challenge.
MOVE 02
You don't want "you're right" — you want "that's right"
"You're right" ends the conversation; "That's right" starts it
tactical empathythe summaryfeeling understood
The principle in one line
"You're right" is almost always the polite form of "please stop talking." "That's right — exactly" means they just heard their own situation described accurately. The first means you won the argument; the second means their guard came down — and only the second keeps them talking.
Two kinds of agreement, opposite destinations
Scripts: a good summary has three parts
The goal isn't to convince them; it's to make them say "that's right"
· Part one · label the emotion: "It sounds like you're caught in the middle on this." Open with "it sounds like" or "it seems like," never "I know you…" — the latter draws the conclusion for them and invites pushback.
· Part two · restate the structure: put their situation back in your own words, especially the stakes they didn't spell out — who's watching them, who they have to answer to. Parroting does nothing; naming the structure does everything.
· Part three · ask to be corrected: "Have I got that right?" It hands control back and gives them room to fix whatever you guessed wrong — and that correction is usually the important part.
· Then shut up. Four seconds. Most people can't sit with silence and keep going, and those sentences are often worth more than the previous half hour.
In the room
Setting: at the renewal meeting your contact opens coldly — "results this year were mediocre. Some people internally want to replace you."
✗ Defend immediately
"But we shipped three modules and the failure rate came down…" The numbers may all be correct, but what you're doing is contradicting them — right at the moment they admitted their own position is awkward. The likely reply: "You're right, but we still have to re-evaluate." Polite, closed, no follow-up.
✓ Get their situation right first
"Thanks for being straight with me. It sounds like you've been absorbing a lot for us internally this year, and you don't have a single result you can hold up that anyone would understand at a glance. Now people are floating a switch, and you're the one who has to weigh what switching costs. Have I got that right?" (A few seconds of silence. Them: …yes, that's it. The new VP only looks at one metric, and I can't present any of yours against it.) —— The real problem was never "mediocre results." It's "no metric that survives in the room." You already had the data; the self-defence route would never have bought you that sentence. What to do next isn't a discount to save the account — it's one page of results restated in the VP's metric.
Why it works · mechanism and evidence
Why it works
Mechanism: being accurately understood is a socially rewarding signal in its own right, and being misunderstood activates negative circuitry that partly overlaps with social pain. A person who is defending only repeats their position; once the defence lifts, they trade information. So "getting them to say that's right" isn't emotional manipulation — it's an observable state change.
Hard evidence · feeling understood and not understood look different in the brain: Morelli, Torre & Eisenberger (2014, Social Cognitive and Affective Neuroscience) compared the two experiences under fMRI. Feeling understood activated the ventral striatum and middle insula (associated with reward and social connection); feeling not understood activated the anterior insula and dorsal anterior cingulate — the same regions that keep showing up in social-rejection research. Strength: a single fMRI study with a small sample, and reverse inference from region to function has well-known limits. (cross-ref psychology / neuroscience)
Moderate evidence · verbatim restating really does change behaviour: in a restaurant field experiment by van Baaren et al. (2003, JESP), servers who repeated the order back word for word received noticeably larger tips than the control group who answered "OK, no problem." One setting only — don't stretch it into "restating always works" — but it shows that being repeated accurately does its work without any persuasive content at all.
Mechanism note · putting a feeling into words weakens it: labelling an emotion cools the other person because the act of converting a feeling into language dampens the emotional response itself (evidence at both the fMRI and behavioural level; covered in the piece on listening). Name the feeling before you discuss the facts — reverse the order and it turns into arguing.
Three boundaries. One, "that's right" is not a buying signal: it means the guard dropped; the terms still have to be negotiated. Two, the summary has to actually be accurate: getting it wrong is worse than not trying. Three, the hardest one: using the feeling of being understood to extract terms they wouldn't otherwise give is manipulation. What empathy should produce is a proposal that fits their situation better, not a more precise pressure point.
MOVE 03
Say the worst of it for them, then ask them for help
Say the worst of it first, then ask for help
accusation auditasking for helpcognitive perspective-taking
The principle in one line
The sentence you're most afraid they'll say loses most of its force once you say it first. Then don't rush to defend — hand them the problem. Asking for help doesn't read as weak; it usually reads as more professional.
Scripts: the accusation audit
Use it to open bad news, a high price, or any situation where you're clearly in the wrong
· One, list the three harshest things they might be thinking: "here comes another price rise," "this outfit is too small to handle it," "they were late last time." Write them down.
· Two, open with "you're probably going to think" and say them out loud: "What I'm about to say will probably sound like we're using this as an excuse to charge more." Flat tone, no defence, then pause.
· Three, give the conclusion with no warm-up: bad news always goes first. Listing wins and then mentioning it "by the way" makes people feel handled — and feeling handled damages trust more than the bad news does.
· Four, hand them the problem: "How do I keep this from hurting you on your side?" That's a request for counsel, not an apology.
In the room
Setting: the project is slipping by two weeks, you have to tell the client today, and their annual meeting is three weeks out.
✗ Lead with the wins, mention it in passing
"Good progress these two weeks, the integrations are through, and delivery timing may need a slight adjustment…" They'll snap to attention on "adjustment" and re-evaluate every sentence you just said. The three seconds of awkwardness you saved cost you a permanent discount on your status reports.
✓ Own it first, then ask for help
"This next part is going to sound unprofessional, and you may well wonder whether we took on too much work. Conclusion first: the core module will be two weeks late." (pause) "I can go into why, but I'd rather solve your problem first: the annual meeting is fixed — how do I make sure you have something to demo there?" (Them: actually we only demo three scenarios; the rest can come later.) —— Owning it means they don't have to spend energy exposing you; asking for help moves them out of the judge's seat. The outcome is a re-ordered scope: those three scenarios first, everything else on the new schedule. The delay itself never came up again.
Why it works · mechanism and evidence
Why it works
Two mechanisms. One, when you say it first you set both the framing and the timing, and their experience shifts from "discovering" to "confirming" — and the sense of having been kept in the dark does more damage than the bad news itself. Two, asking for help moves them from "judging you" to "solving it with you," and that shift costs far less than any defence.
Hard evidence · disclosing it yourself does less damage: Williams, Bourgeois & Croyle (1993) found in mock-trial experiments that a party who volunteers damaging information first (the researchers call it stealing thunder) suffers significantly less than one who waits for the other side to reveal it; Dolnik, Case & Williams (2003, Law and Human Behavior) replicated the effect and suggested part of the mechanism is that saying it first changes how serious listeners judge the information to be. Peer-reviewed with independent replication, but largely in courtroom simulations — carrying it into business communication is a mechanism inference.
Hard evidence · people who ask for advice are seen as more competent: Brooks, Gino & Schweitzer (2015, Management Science) showed across a series of experiments that seeking advice raises the other person's estimate of your competence — the opposite of the intuition. The mechanism is that being consulted feels like being validated. The boundary is clear: it only holds when they actually know the subject.
Hard evidence · you want accuracy, not feeling: Galinsky, Maddux, Gilin & White (2008, Psychological Science) compared two instructions in negotiation experiments. The group told to work out cognitively what the other side was thinking achieved better individual and joint outcomes; the group told to feel emotionally what the other side felt gained far less, and in some settings did worse. The "tactical" in tactical empathy is accuracy, not softness.
Boundary: there is real deference and fake deference. Real means stating a constraint that exists; fake means performing helplessness to buy a concession — and once that's seen through, nobody believes any "I can't do that" of yours again. Also, if the other side is a pure zero-sum hardliner, posture doesn't fix a power gap: that's a question of alternatives (the BATNA from the piece on negotiation basics). And don't turn the accusation audit into a self-flagellation session — two or three of the harshest lines is enough.
MOVE 04
What's holding it up is usually what you never thought to ask
What decides it is usually what you never thought to ask
black swansunknown unknownsunstructured moments
The principle in one line
A deal that won't move usually isn't a price problem — there's a constraint at work that you don't even know exists. Voss calls this kind of information a black swan: you don't know that you don't know it, so it isn't on your question list. Finding it isn't about better phrasing; it's about getting them to talk more, in less formal moments.
An information grid: your question list only covers the left half
Checklist: five moves that force black swans out
All of them raise how much they talk and how informal the setting is
· Change the channel. Don't negotiate anything complex over email — a ten-minute phone call gets you half-sentences that would never have appeared in writing.
· Don't hang up when the business is done. "By the way, something I've been curious about — how long does a decision like this usually take on your side?" The three minutes after the agenda ends is when black swans show up most.
· Ask "what's been hardest to push through," not "how's it going." The first asks for structure; the second gets you "still in process."
· Four seconds of silence. Say your piece and don't fill the gap — most people will, and what they fill it with is often what they hadn't planned to say.
· Watch the words they repeat. Someone who keeps mentioning "the annual meeting," "the audit," "our last supplier" is showing you the door in.
In the room
Setting: a deal has been stuck for three weeks and every follow-up gets the same reply — "still going through the process."
✗ More frequency plus a test discount
Two emails a week, then in week three "sign this month and I can take five percent off." You're solving an undiagnosed problem with price — if the blockage isn't money, that five percent is a gift; worse, you've taught them that waiting produces discounts.
✓ Change the channel, ask for structure
Call them: "I'm not chasing you — I just want to understand how to help from my end. What's been the hardest part to push through so far?" (Them: …honestly, we're replacing our ERP at year end. Yours will have to integrate with the new system, and nobody wants to sign for that right now.) —— Three weeks of silence explained, and it has nothing to do with price. The fix is now obvious: an integration note, and a contract split into "the parts that don't depend on the ERP now, the integration module signed once the new system lands." Discounting would not have closed this at any price.
Why it works · mechanism and evidence
Why it works
Mechanism: stalled negotiations usually come down to information asymmetry, and the hardest gap to close isn't "the piece I know I'm missing" — it's "the piece I never realised existed." Since you can't aim a question at something you can't conceive of, the only workable strategy is to raise the volume and informality of the other side's talking: treat hit rate, not questioning technique, as the lever.
Hard evidence · the opening minutes are already deciding it: Curhan & Pentland (2007, Journal of Applied Psychology) used only the conversational dynamics of the first five minutes — vocal variation, turn-taking, activity level, with no content at all — to predict a substantial share of the variance in final outcomes. Small talk isn't a warm-up; it's already transmitting and collecting information. Peer-reviewed, lab negotiation task.
Moderate evidence · a bit of chat first produces better results: Morris, Nadler, Kurtzberg & Thompson (2002) had some participants make a brief, non-business phone call before negotiating by email: that group reached agreement more often and did better, and rated each other as more trustworthy. Strength: lab study, student samples, but consistent with "the narrower the channel, the more gets lost."
Weaker evidence · the "black swan" category itself: it comes from Voss (2016) as a practitioner's framework, not an experimental result. Its value is reminding you that the unknown-unknown quadrant exists, and shifting the action from "ask better" to "get them to say more."
Boundaries. One, don't turn "listen more" into fishing for leverage: what you dig up should make the proposal fit their situation better, not locate a soft spot to price against. Two, don't dig forever: if three good questions produce nothing new, there probably isn't a black swan — decide on what you know. Three, small talk has to be actual small talk: purposeful faux-warmth is easy to feel, and once it's spotted, every bit of empathy that came before gets reinterpreted as technique.
Your Day 23 Action
Pick one conversation that's stuck: a client who's gone quiet, a raise you need to ask for, a partner you've deadlocked with. Today, prepare two sentences and make one phone call.
1 (5 min): Copy out, word for word, the last thing they said that stung. Next to it write: what were they actually worried about when they said that? Three guesses, at least one of which is "a problem inside their organisation that has nothing to do with me."
2 (10 min): Write a summary using the three parts — label the emotion, restate the structure, "have I got that right?" Read it back: if you were them, would you want to argue or want to add something? If it's argue, you're still reasoning at them. Rewrite.
3 (5 min): Prepare one calibrated question starting with "how" or "what," to which you don't know the answer yourself.
4 (3 min): If any part of this is on you, write down the two harshest things they might be thinking and pick one to say for them.
5 (today): Call, don't email. After the summary, stay silent for four seconds — those four seconds are the real assignment.
Boundary: what all of this should produce is a proposal that fits their situation better. If what you dug up only tells you more precisely where to squeeze, you weren't practising empathy, you were extracting — and extraction works exactly once.
Think It Through
1. I'm not much of a performer. Won't "how am I supposed to do that" just make me sound weak and invite a squeeze?
It depends on two things, neither of which is acting.
First, whether you have an alternative. The same sentence coming from someone with two other clients in play is a request for counsel; coming from someone who can't make payroll without this deal it's a plea. The difference isn't tone — it's whether you know you can walk. The preparation that makes the line stand up happens away from the table (that's the BATNA from the piece on negotiation basics).
Second, whether you stop after saying it. The common failure isn't a soft tone, it's answering your own question: "…of course I know that's difficult, maybe we could think of something else?" That takes the problem back off them and onto you, and wastes the whole move. Ask, then shut up.
One practical adjustment: swap "I" for "we," which turns the posture from asking for help into solving together. Not being a performer is actually an advantage — the point is accuracy, not theatre, and a flat delivery reads as more credible.
2. What if they've read the same book and can see straight through the technique?
Separate what's been seen through. If they've spotted the form — "you're labelling me, aren't you" — that does almost no damage, because what's underneath the form is true. Seeing through an accurate summary doesn't make it less accurate.
What actually collapses is the other case: they notice your empathy and your behaviour don't match. You say "sounds like you're caught in the middle" and then use that information to price against their deadline. What they've seen through then isn't the technique, it's the motive — and once a motive is settled on, everything you say afterwards gets translated into tactics.
So the fix isn't at the script level: only dig for information you intend to use on their behalf. If you're called out to your face, honesty is the best answer: "Yes — I'm genuinely trying to work out where this is stuck on your side, because I don't want to come back with a proposal you can't use. Have I understood it?" Pull the attention back to content; denial is what's awkward.
3. Where exactly is the line between tactical empathy and faking it? "Tactical" already sounds calculating.
Unpack the word calculating. The accuracy of your empathy can be deliberately trained, and that isn't fake. What's fake is a manufactured feeling, and understanding that gets used against the person who gave it to you. The research supports the distinction: in negotiation experiments, people told to work out cognitively what the other side was thinking did better than people told to feel what they felt (Galinsky et al., 2008). Train for accuracy, not for being moved.
A line you can apply on the spot: if they saw your prep notes from today, would they feel understood or studied? "They have to answer to the VP; the real pain is having no metric that survives the room" is understanding — you wrote it in order to give them that metric. "Their boss is on holiday next week, squeeze them to sign before then" is study — you wrote it to use their time pressure.
Simpler self-check: did what you learned change your proposal, or only sharpen your pressure? This works so well in long relationships precisely because it can't be faked for long.