Flipped Interaction

Before preparing the answer, let the model ask for the missing information.

use
  • Research
levels
simple · medium · hard
links
3 related methods
license
CC BY 4.0 · View the card’s open source

What is it?

Flipped Interaction hands the turn of asking questions to the model. The user states the goal and the limits; the model gathers the information needed to complete the task. The questions are not the goal; they are the way to reach enough information.

When does it help?

When you roughly know what you want but do not know which details are needed for a correct brief.

Examples

The situations and responses below are fictional teaching examples; they are not results from a model, tool or benchmark that was actually run.

Simple

Situation

A workshop announcement is to be written; basic information is missing.

Prompt

Prompt

We are going to prepare a workshop announcement. Ask for the missing required fields one at a time: topic, date/time, how to take part. Do not ask again for information I have given. When the three fields are complete, write a two-sentence draft; do not invent an unknown link.
What you know right now: The topic is watercolor. The date is 12 October, the time 14:00. How to take part is not decided yet.

Sample output

Model: “Which way of taking part should we state?” User: “They should reply to the announcement.” Model: “The watercolor workshop is on 12 October at 14:00. To take part, you can reply to this announcement.”

What did we get?

The single necessary question was tied to a usable draft.

Medium

Situation

For a one-week study plan, the time available and the priorities are unknown.

Prompt

Prompt

Before you make a plan for me, collect the missing information. My goal is to learn Python lists; I am a beginner. Ask at most three questions, one at a time. Stop when the time, available days and learning criterion are sufficient; summarize the answers and produce a one-week plan.

Sample output

Model: “How many minutes a day can you spend studying?”

User: “20 minutes.”

Model: “Which days are you available?”

User: “Monday, Wednesday, Friday.”

Model: “By the end of the week, what do you want to be able to do without help?”

User: “Write code that filters a list.”

Model: “That is enough information: three days, 20 minutes a day; goal: filtering a list without help. Monday lists/indexing, Wednesday conditional filtering, Friday an unassisted exercise on a new list. Each session 20 minutes.” No further questions are asked.

What did we get?

The plan was built on the capacity given, not on assumed free time.

Hard

Situation

A team will decide on a research question; the participants do not yet know every detail.

Prompt

Prompt

Goal: Complete a small team's user research brief. What is known: 5 interviews can be done; the product is a note-taking app; there is 1 week in total.
First ask what the decision will change and wait for my real answer. Do not fill in a field that is unknown at the start as if it had an answer.
Then ask only about gaps that change the design; at most 4 questions in total. If the user says I don't know, briefly show the options and their effects; do not write as if a decision had been made.
Sufficient-information criterion: target behavior, participant criterion, interview length, decision owner. If anything remains missing, mark the assumptions separately. At the end, present the brief draft for confirmation; do not ask new questions once the information is sufficient, and do not start a new survey or recruitment before approval.

Sample output

Model: “Which decision will the result of the research change?”

User: “We'll decide which problem blocking people from saving their first note to fix first.”

Model: “Whom should we include in the interviews?”

User: “People who didn't save their first note on the day they signed up.”

Model: “How many minutes can each interview last?”

User: “15 minutes.”

Model: “Who will make the priority decision?”

User: “Ayşe, the product lead.”

Model: “That is enough information. Brief presented for confirmation: within one week, 15 minutes each with five people who meet the stated criterion; the aim is to understand the obstacles to saving the first note, and Ayşe will set the priority of fixes.” It stops at four questions; no interviews are started.

What did we get?

The number of questions stayed limited; unknowns did not turn into false certainty.

Where should you stop?

Interviewing again for a job where all the information has been given is unnecessary. Question Refinement improves the wording of the question; Flipped Interaction collects missing inputs from the user. Eliciting preferences focuses specifically on values and trade-offs.

Sources

  • A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT (new tab) — White, Jules; Fu, Quchen; Hays, Sam; Sandborn, Michael; Olea, Carlos; Gilbert, Henry; Elnashar, Ashraf; Spencer-Smith, Jesse; Schmidt, Douglas C.. 2023-02-21. Defines the Flipped Interaction pattern, in which the model reaches a specific goal by asking the user questions. Evidence level: relevant body sections of the original paper.

Two white lattice bridges reaching toward each other stand with open ends on either side of a small amber piece in the middle.
How this image was made

Original illustration made with Google Gemini · 1024 × 572

Generate an image: Create an original museum-grade computational sculpture photographed as a physically present installation, horizontal 16:9. Deep anthracite void, mineral porcelain whites, restrained ice-blue and amber accents, fine particles only where structurally meaningful, tactile surfaces, subtle volumetric illumination, believable depth, exceptional edge detail and intentional negative space. An incomplete white mineral bridge stops at a precise missing joint; a small amber matching piece approaches from the opposite side while finished sections remain intact. The gap is specific and local, not a collapsed entire bridge. Keep the image sophisticated and legible at mobile size: one dominant mechanism, a clear silhouette, no decorative overload. No text, letters, numbers, typography, captions, symbols, watermark, logo, fake chart, screenshot, interface, identifiable people or artist signature. Do not imitate any specific existing artwork.

Adapt the prompts to your own situation. In an example that needs a tool or a separate call, copying the text alone does not set up that way of working.