Door-in-the-Face Technique

SKILL.md

---
name: door-in-the-face
description: "Activate when: user asks 'they made a big ask then backed off, should I concede?'; user is designing a negotiation opening and wants to land at a specific price; someone says 'anchor high then retreat' or 'reciprocal concession'; user suspects they are being manipulated by a two-step request pattern.
  Do NOT activate when: there is only a single offer with no prior refusal (use anchoring instead); the user is asking about foot-in-the-door (small ask first, then large ask — the reverse technique)."
---

# Door-in-the-Face Technique

## Overview

Ask for something large (expect refusal), then retreat to the smaller request you actually wanted. The empirically documented result: compliance with the smaller ask is 2-3x higher than asking for it directly (Cialdini et al., 1975). The mechanism is *reciprocal concession* — the target perceives your retreat as a concession and feels social pressure to match it.

Three operations: **recognize** DITF when used on you; **design** it ethically as a proposer; **distinguish** it from pure anchoring (two requests with refusal vs. a single number). Composes with [`reciprocity`](../reciprocity/SKILL.md), [`anchoring`](../anchoring/SKILL.md), [`signaling-games`](../signaling-games/SKILL.md), [`batna-zopa`](../batna-zopa/SKILL.md).

## When to Use

- Designing a negotiation opening where you want to land at a specific price/term
- Designing a fundraising ask where the target gift is moderate but commitment to the cause is high
- Designing a sales process where you want customers in a specific tier
- **Recognizing** that you are being run with DITF by an opposing party
- Someone says: "door-in-the-face," "reciprocal concession," "anchor high then retreat," "they conceded so I should too"

**Not when:** the larger request is so absurd the target reads the opening as bad faith (the technique collapses); the target has no reciprocity norm operating; you are negotiating with a counterparty who will read your retreat as DITF.

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** user has a concrete case → run The Process directly.
- **Coach mode:** user is unfamiliar or has no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line: if you ask for X (large) and get refused, then ask for Y (smaller, what you actually wanted), you get Y more often than if you asked for Y directly — because the target feels they should concede after watching you concede.
2. Check fit — absurd-anchor situations → DITF collapses; point elsewhere.
3. Elicit the real situation.
> **[WAIT — do not advance until user responds]**
4. One question at a time: is this technique being used on me? What's the actual smaller ask? What's my counterparty's BATNA?
> **[WAIT — do not advance until user responds]**
5. Close: the specific opening + retreat sequence, or the specific defense.
> **[WAIT — do not advance until user responds]**

## The Process

### Designing DITF as the proposer

```markdown
## DITF design

- **Goal:** what I actually want the target to do
- **The "real" ask (Y):** specific request, plausibly worth saying yes to in isolation
- **The "large" ask (X):** larger version of Y, in same direction, plausibly defensible
  - X must be 2-4x Y in magnitude — too small and it's not a real anchor; too large and the target reads bad faith
- **Timing of retreat:** how long do I wait between X and Y?
  - Same conversation: classic DITF
  - Days later: weakened (reciprocal-concession pull decays)
- **Framing of retreat:** "I hear that doesn't work; how about Y?" — explicitly acknowledging the concession
```

### Recognizing DITF as the target

```markdown
## DITF defense

- Did the opening offer make sense as a real request, or did it feel like an obvious set-up?
- After my refusal, is the second request structurally close to what they actually wanted from the start?
- The reciprocal-concession pull: "they conceded so I should too." Recognize this is the *technique*, not the merits.
- Counter: evaluate the second request on its own merits. Would I have agreed to Y if asked directly without seeing X first?
- If no: refuse. If yes: agree, but at your own price, not because of the concession dynamic.
```

### Distinguishing DITF from anchoring

- **DITF:** two requests with a refusal between them; mechanism is reciprocal concession.
- **Anchoring:** a single number influences later numbers; mechanism is cognitive bias.
- A single high opening offer is anchoring. An opening offer + retreat after refusal is DITF + anchoring.

*→ Method in Action: [Cialdini, Vincent, Lewis, Catalan, Wheeler & Darby, 1975](examples/cialdini-1975.md)*

## Packs / Applying It Well

Composes with [`reciprocity`](../reciprocity/SKILL.md) (mechanism), [`anchoring`](../anchoring/SKILL.md) (opening number), [`signaling-games`](../signaling-games/SKILL.md) (retreat signals reservation price), [`batna-zopa`](../batna-zopa/SKILL.md) (tactic within ZOPA).

- Opening ask at **upper bound of plausibility** (2-4x the real ask) — not absurd.
- Retreat in the same conversation; effect decays across days.
- Frame retreat explicitly: "I hear that doesn't work; how about Y?"
- Defense: name the structure aloud. Recognition weakens the pull substantially.

*→ Primary sources: [references/sources.md](references/sources.md)*

## Common Rationalizations

**[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.**

| Fake move | Reality |
|---|---|
| [D] "Their retreat means they're compromising in good faith" | Maybe. Or maybe the retreat is the technique. Distinguish by asking whether the original ask was plausibly serious. |
| [D] "They're being unreasonable, I should refuse the second ask too" | Refusing both is right when the original was obviously DITF. But refusing when the first was genuinely large can leave value on the table. Test: evaluate the second on its merits. |
| [D] "If I open with a large ask, they'll think I'm unreasonable" | DITF research shows the opposite — within the plausibility range, larger opening asks produce better outcomes. The constraint is plausibility, not pleasantness. |
| [D] Using DITF on a sophisticated counterparty | If they've read the literature, they recognize the structure. Use anchoring (a single bold offer) instead. |
| [D] Opening with an absurd ask | The technique collapses. The opening must be in the plausibility range to function. |
| *→ Add [O] entries here after each real use — paste the actual failure pattern* | *What went wrong and why* |

## Red Flags

- An opening offer clearly beyond plausibility, followed by a "compromise" to something still aggressive
- A retreat framed in unusually generous language ("I really stretched to make this work")
- Pressure to respond to the retreat quickly
- The "compromised" position is exactly what they wanted from the start, by your read
- You feel a strong pull to concede because they conceded — the diagnostic of the reciprocal-concession mechanism

## Verification

- [ ] Distinguish DITF from pure anchoring (two requests with refusal vs single number)
- [ ] If designing: opening ask is in plausibility range (2-4x the real ask)
- [ ] If defending: original ask evaluated on whether it was genuinely serious
- [ ] Reciprocal-concession pull recognized, not acted on unreflectively
- [ ] Second ask evaluated on its own merits (would I have agreed to Y in isolation?)
---

*Part of **deciqAI Knowledge Skills** — open-source thinking skills that make rigor executable for AI agents. Built by deciqAI · https://deciqai.com · Contributions welcome — see the template at the repo root.*