The Diagnosis-Intervention Cycle

Definition

Schwarz breaks every facilitator intervention into six parallel steps split across two phases. The private diagnosis phase: (1) observe behavior, (2) infer meaning, (3) decide whether, how, and why to intervene. The public intervention phase mirrors it exactly: (4) describe the observable behavior you privately noted and test whether the group agrees with it, (5) share the inference you privately made and test whether the group agrees with it, and (6) help the group decide whether and how to redesign its behavior. The two halves are deliberately parallel — every private judgment gets surfaced and checked with the group before it is acted on.

In the Book

Chapter 8 walks through an extended case (Exhibit 8.1) of three executives arguing about a performance-management redesign, where the conversation visibly drifts off-topic. The facilitator privately tracks the drift (steps one through three), then makes it public: first restating exactly what was said and checking whether Ronald and Andrew agree that's an accurate description (step four), then naming the private inference — "I don't see the relationship between your comments and the conversation about the forced distribution" — and checking whether the group sees it differently (step five), then asking the group directly whether it wants to return to the original topic or continue on the new one (step six). Schwarz shows explicitly how each step embeds a specific ground rule: step four uses specific examples, step five tests an inference, step six jointly designs next steps, and every step's "test whether they see it differently" clause is the ground rule of combining advocacy with inquiry. Chapter 7 adds that intervention types can target different diagnostic frames — ground rules for a narrow single behavior, the group effectiveness model for a broader structural issue, or a group's theory-in-use for the deepest, most consequential kind of intervention.

Why It Matters

The cycle solves a specific credibility problem: any observer who intervenes based on a private inference risks imposing an interpretation the other party never agreed to, which breeds resentment rather than change. By forcing the observation and the inference each to be stated separately and checked before acting on them, the cycle makes the intervener's reasoning auditable in real time rather than asserted from authority — the group can correct a wrong inference before it becomes the basis for a decision. The pattern — separate what you saw from what you concluded, and check both before proposing a change — applies anywhere feedback is delivered, from performance reviews to incident postmortems to technical code review.