Fishbowl

Goal

The Fishbowl addresses a specific elicitation problem: in a large group, the useful knowledge is spread across people of different roles and standing, and a plenary meeting does not draw it out. Too many people speak at once, or nobody speaks, and the assumptions each camp takes for granted stay tacit. The Fishbowl restricts the floor to a small circle that holds a real conversation, and it gives everyone else an active task, to observe and to record. It is this asymmetry, a few people talk while the rest listen with a brief, that lets first-hand reasoning be heard and buried disagreements surface.

The Fishbowl is one of the collaborative games, and the result it aims for is the one BABOK attaches to that family: bringing hidden assumptions and differences of viewpoint to the surface. Where the affinity map structures a body of ideas already produced and the Product Box frames an offering, the Fishbowl works on the perspectives of two groups that do not listen to each other in ordinary discussion.

The deliverable is an observation log: the hidden assumptions and divergent perspectives the outer circle captured during the conversation, each tied to an implication, a requirement to write, a risk to test or an action to take. To that log is added the harvest, run in session, where the facilitator and the group together determine the decisions and actions that follow.

Usage

When to use it

• Large group, a single topic: a plenary would sentence the room to silence, a few speak while the rest listen.
• Hidden assumptions to surface: the aim is to expose divergent perspectives.
• Roles forced to cooperate: seating them at the centre lets the rest of the room hear the friction first-hand.
• Two groups that do not listen to each other: business and IT finally hear each other set out their reasoning.
• Early elicitation: gathering the lived experience of those involved before writing a single requirement.
• Reserved or lower-status voices: the open chair offers a low-threshold way into the conversation.

When not to use it

• A decision must be made and owned: the Fishbowl exposes perspectives without arbitrating them, run a workshop with a decision rule.
• Public candour is unrealistic (a steeply hierarchical room, a sensitive or confidential subject): a public inner circle is the wrong container, prefer the individual interview, an anonymous survey or document analysis.
• Group too small for two circles: with no audience to observe, the arrangement adds nothing, hold a plain facilitated discussion.

Running a Fishbowl

The Fishbowl shares the frame common to the collaborative games, a purpose of its own, a timed sequence, a neutral facilitator and a harvest, and it works that frame into a geometry all its own: two concentric circles and a rule that governs the move from one to the other.

The geometry: two circles

The inner circle brings together a small group, often called "the fish", who hold the actual conversation. The facilitation literature converges on three to seven seats: below that, the group dynamic does not take hold; above it, speaking time thins out and the conversation reverts to a plenary. These people are chosen for their first-hand experience of the topic. The outer circle gathers everyone else, seated around the first, with an active brief: observe the conversation and record what is said in it, without stepping in while it runs. This division of labour is the instrument itself: it lets a large group engage a topic through a conversation that stays, for its part, human-sized.

The open chair

Two variants are documented, and the choice between them is made in preparation. In the open Fishbowl, one inner-circle chair is left empty. Any observer may take it at any moment to contribute, and when they do, an inner participant voluntarily gives up their seat, so that the number of chairs stays constant. This is the low-threshold rotation mechanism that keeps the conversation fed with fresh voices without the facilitator having to orchestrate turns. In the closed Fishbowl, every inner seat is filled: the initial group talks for a set time, then the inner circle is swapped out as a block for a new group drawn from the outer circle. The open format gives fluidity and surprise; the closed format lets you cleanly compare what successive groups say on the same question.

The facilitator

The Fishbowl gains a clear benefit from a neutral facilitator, and the role is first of all one of execution. They frame the topic and phrase the single question the inner circle has to answer. They seat the fish, state the rules aloud and enforce them throughout. Two of these rules decide the outcome, and the facilitation method states them plainly: the fish talk to each other and not to the audience, and no one makes speeches. The facilitator opens the conversation, holds it to its time, watches the empty chair and manages entries and exits, balances speaking time, then runs the harvest. Without that discipline, the format drifts into its commonest failure: the panel of experts, where each person addresses the audience in monologues and the peer conversation, which is the product sought, never takes place.

The timed sequence

BABOK describes for all collaborative games a sequence of three generally bounded steps, opening, exploration and closing. A fully specified variant, the User Experience Fishbowl from Liberating Structures, gives this arc a reproducible form, a starting template to adjust:

• Introduce the shared challenge and invite the outer circle to listen (2 min).
• The inner circle, three to seven people who have lived the topic, converse naturally about what they did, concrete examples, successes and failures rather than opinions (10 to 25 min).
• The outer circle, in small groups, formulate observations and questions (4 min).
• Questions and answers: the outer circle asks, the inner circle answers (10 to 25 min).
• Closing debrief: "what seems possible now?" (10 to 15 min).

The whole runs to between thirty-five and seventy minutes. Two features of this template make the result. The inner circle speaks of lived experiences, concrete and dated, not of general opinions: it is the concrete that surfaces assumptions, opinion covers them over. And each phase is time-bounded, which stops the conversation from devouring the session and forces the closing, where the harvest happens, to exist.

The harvest

The conversation produces nothing until the outer circle's observations are consolidated. The harvest is done in session, while the group is present and remembers what a remark meant: the facilitator and the participants go through what was recorded together and name the decisions and actions that follow. An assumption brought to light becomes a requirement to develop or a point to validate against data; a log filed in a drawer and used three days later returns the words but has lost the reasoning that carried them.

What makes a Fishbowl fail

The obscure objective

A Fishbowl whose purpose or expected result no one can see rightly comes across as unproductive theatre. The question put to the inner circle is framed before the session.

The panel in disguise

Without a facilitator enforcing "talk to each other, no speeches", the inner circle addresses the audience and the peer conversation disappears.

The skipped harvest

Outer-circle observations that are neither consolidated nor carried into actions evaporate: the session has produced nothing.

The false consensus

A lively conversation feels conclusive; the insights it produces remain hypotheses to validate. BABOK notes that a group's participation can create excessive confidence in the conclusions reached.

The room under watch

Speaking candidly in front of an audience presupposes a safety that not every room offers; in a steeply hierarchical setting, the inner circle self-censors and the hidden assumptions stay hidden.

AI considerations

AI works mainly upstream and downstream of the conversation, never in its place. Upstream, it shapes the preparation: the exact phrasing of the single question, the timing of the steps, the instructions read aloud, provided the stakes themselves have been settled by someone who knows the subject. During the session, an automatic transcription, announced and consented to, frees the outer circle from taking notes on the fly and returns it to its real task, to observe. Downstream, it groups the harvested observations into candidate themes, brings logs together when the same Fishbowl has been replayed with several groups and drafts a first pass of requirements or questions from the transcript, which the analyst checks. For a remote Fishbowl, live captioning lowers the barrier to the open chair.

The limit is intrinsic to the technique. Spotting a hidden assumption is an act of human interpretation: a transcript records what was said, it does not capture what was assumed without being stated, which is precisely what the Fishbowl exists to bring out. No model reads the psychological safety of a room or the non-verbal signal of a fish who self-censors the moment their manager's name is spoken. The observer's judgment, which recognises the implicit as it goes by, cannot be delegated.

Recording itself changes what it measures. If participants know a model is listening, they say less, and the technique's value collapses. A Fishbowl transcript is also personal data, often sensitive: participants' consent and Swiss data-protection expectations govern any recording, and pasting the transcript into a consumer service whose terms allow training on inputs breaks that consent. Anonymise before processing, or process on infrastructure covered by the consent obtained.

Examples

The Fishbowl's deliverable is an observation log. A cantonal administration has run a three-month pilot of a new online counter for residents. Before writing the requirements for the full roll-out, the business-analysis team convenes a Fishbowl: in the inner circle, the counter agent, the caseworker and the IT integrator who ran the pilot; in the outer circle, the business-analysis team and two neighbouring offices, tasked with catching the assumptions the pilot team no longer notices.

The log is the product of the session. Each line ties a remark heard in the inner circle to the tacit assumption it reveals, then to the implication for requirements.

The right-hand column is what separates a pleasant conversation from a usable result: without it, the log stays a list of remarks. Filling it in with the group, in session and while everyone still remembers what a remark meant, is precisely the work of the harvest.

Cost

Tools

The basic material is that of a meeting room: two circles of chairs, one inside the other, a chair left empty at the centre for the open format and a flip chart or a board to carry the harvest. The physical arrangement does part of the work on its own, since it makes plain and unambiguous who has the floor and who listens. No licence or platform is required.

Remotely, a Fishbowl is held over videoconference, provided you impose verbally what the chairs imposed by themselves: who speaks and who observes. The open chair is replaced by an explicit protocol, a queue signalled in the chat or a raised hand the facilitator arbitrates, and the outer circle records its observations in a shared document. A captioning or transcription tool, announced and consented to, returns the outer circle to its observing role by relieving it of note-taking.

For the harvest, a collaborative whiteboard (Miro, Mural, FigJam) serves to group the observations into themes, in session or just after, and its assisted-clustering features speed up the slowest step. On a workstation locked to the office suite, a shared document or Microsoft Whiteboard fills the same role. These tools belong to the after-conversation; the heart of the Fishbowl remains the arrangement and the talk.

Sources

• IIBA, A Guide to the Business Analysis Body of Knowledge (BABOK Guide) v3, §10.10 Collaborative Games: provenance and knowledge area, facilitator role, three-step sequence, strengths and limitations.
• User Experience Fishbowl, Liberating Structures (Henri Lipmanowicz and Keith McCandless): a complete, reproducible run template, with the rules "talk to each other, no speeches". It is one documented variant of the Fishbowl.
• Kaner, S. et al., Facilitator's Guide to Participatory Decision-Making (Jossey-Bass, 3rd ed.): a foundational reference on large-group participatory formats and on the facilitator's role.