Immunity to Change Experiments: Test Assumptions Safely

Immunity to Change experiments examine a prediction rather than require its defeat. A useful result may narrow an assumption, support part of it or reveal a responsibility that the client must negotiate. Choose a test that can provide information without making another person absorb avoidable risk.

This guide follows an original fictional delegation case from an initially unsuitable proposal to actual colleague agreement, a completed briefing, a factual correction and a limited review. The leader changes his understanding without claiming that a belief disappeared or that his organization changed its standards.

Keep the method and the illustrative tool distinct

Robert Kegan and Lisa Laskow Lahey's Immunity to Change approach connects a meaningful improvement with obstructing behavior, protective commitments and underlying assumptions. Their 2010 Harvard interview describes examining beliefs treated as certainties. This is the originators' account of their approach, not a controlled effectiveness result for the example here.

Their 2009 book excerpt, distributed through HarvardX, describes tests that are safe, modest, actionable and conducted with a research purpose. The question is what the result reveals about the assumption, rather than whether the person immediately performs better. That testing use of SMART differs from the familiar goal-setting expansion.

Minds at Work provides the formal method and its professional training pathways. The working record below is an original coaching aid. It is not the official four-column map, a diagnostic instrument, a certification or a substitute for learning the full method. An ordinary conversation does not establish someone's developmental stage.

Make the prediction specific enough to meet evidence

Start with something the client actually wants. A sponsor's wish for a more accommodating employee does not supply the client's personal goal. Ask what improvement would matter, which behavior gets in its way and what the client believes that behavior protects. Let the client correct the language.

If a possible protective pattern is still unclear, the worry-box guide explores that earlier inquiry. A poorly specified prediction is a reason to clarify the work before testing, rather than make the experiment larger.

A general value such as doing reliable work is not itself an obstructive commitment. The question is whether the particular protective action is necessary in every instance. Rewriting every sentence may exceed what an agreed factual check requires. Conversely, the factual check may be a real duty that must remain.

Translate a broad certainty into a prediction about an identifiable situation. “Delegation will damage my standing” does not specify the task, audience or observation. “If Morgan sends this internal briefing without my full stylistic rewrite, Hana will call my oversight careless” can be compared with actual words and decisions.

Ask what evidence would challenge that expectation and what would support it. A factual correction, a request for clearer wording and a global judgment about reliability are different events. Neither a friendly tone nor silence supplies an unspoken approval. The client can record a feeling without presenting it as the other person's stated opinion.

Check authority before designing personal risk

Some actions are unsuitable even when the client chooses them. Do not skip an obligatory review, remove necessary support, expose confidential information or use a colleague as an unwitting test of the leader's worry. A coaching experiment does not cancel ordinary work responsibilities.

Distinguish three permissions: the client agrees to the inquiry; the responsible work owner authorizes the task and limits; affected colleagues accept their actual work and receiving arrangements. One person's enthusiasm cannot supply the other two. A colleague may accept written work while declining another meeting.

Keep regulated, external, safety-critical or irreversible decisions within their proper controls. A small internal task may still affect someone with little power to refuse. State the genuine alternative and preserve a no. If adequate conditions cannot be agreed, the useful decision may be to postpone the test.

The leadership-paradox guide considers competing commitments alongside organizational constraints. Do not interpret a shortage of time, unclear approval authority or a punitive standard solely as the client's resistance. Another kind of conversation may be needed first.

Write a complete experiment agreement

Coach Soraya and leader Teo work on a bounded inquiry from 18 to 30 January 2046. Existing coaching arrangements, work roles, the earlier briefing draft and its underlying records are outside this measured period, not free work. The task is an internal planning brief, with no customer or safety-critical release.

Soraya prepares for fifteen minutes on 18 January, meets Teo for forty-five and makes ten minutes of private notes. Teo wants Morgan to own the briefing wording. Teo initially proposes sending it without his normal check to see whether Hana objects. Soraya does not accept that as a useful test of stylistic control.

Teo uses twenty minutes on 19 January preparing and sending Soraya his selected experiment record. Soraya uses ten minutes reading it and ten composing and sending a correction: the proposal mixes removal of the factual check with removal of stylistic rewriting. A reaction would not reveal which change mattered, and the required factual review cannot be bypassed.

Teo uses twenty minutes on 20 January reading that comment, revising the plan and sending two operational invitations. His private account stays with Soraya. The invitations tell Morgan and Hana that Teo is practicing leaving wording with its author while retaining the required facts check; neither receives his whole reflection record.

The record includes learning and delivery conditions without treating the colleague's effort as an invisible resource. It also separates Teo's private prediction from the standards Hana is actually authorized to set.

Obtain the colleague's actual counterproposal

Morgan spends fifteen minutes on 21 January reading the invitation and replying. The proposed live briefing on 22 January clashes with existing work. Morgan's complete answer is:

I cannot take the live briefing on 22 January. I can spend thirty minutes on the written internal version on 24 January from 11:30 to noon, after your factual check is available. I can also reserve up to ten minutes on 25 January for a required factual repair. I will own the wording within the existing brief, not approval authority. Please confirm that this route meets the purpose before I begin.

Hana spends ten minutes that day reading and replying. She accepts fifteen minutes on 24 January at 14:00 to review the written brief and five minutes on 25 January to check a repair if needed. Teo must complete the factual check before Morgan's writing block. Hana declines a general rating of Teo's reputation and agrees to discuss the task standard.

Teo uses fifteen minutes on 22 January reading both answers and sending acceptance to both. Morgan and Hana each spend five minutes on 23 January reading that confirmation. The case now has a usable receiving arrangement before either performs the work. Their existing assignments are not erased to make space for the experiment.

The project still aims for an approved internal brief by 24 January at 17:00. The reserved next-day repair option does not silently move that target. If a correction prevents approval that day, the account must retain the miss.

Run the bounded action and preserve the quality hold

Teo works from 09:00 to 09:25 on 24 January checking the facts. He corrects one figure in the earlier draft and identifies the checked version for Morgan. This twenty-five-minute block includes recording that check and supplying the file. Removing the full rewrite has not removed his factual responsibility.

Morgan uses the agreed thirty minutes from 11:30 to noon reading the checked draft, completing the written brief and sending it to Teo and Hana. This is the full original selected passage:

The internal handover needs one release decision and one named receiving arrangement. The current file identifies the output and proposed date, but a recipient's name does not establish who may approve it. Keep the existing approval route until its owner confirms any change. Record the actual answer and its date. A missing answer remains open; it is not acceptance. This brief supports the internal planning discussion and does not authorize an external release.

Teo uses ten minutes at 12:30 reading the result and leaving its wording intact. Hana uses her fifteen-minute block at 14:00 reading it and sending feedback to both: the recommendation is clear, but a source footnote is missing. Readiness stays held until that reference is repaired. She does not say Teo's oversight is careless.

Morgan uses the reserved ten minutes from 10:00 to 10:10 on 25 January reading that feedback, adding the source footnote and sending the corrected version to both. Hana uses five minutes from 10:30 to 10:35 checking it and sending approval to both. No additional drafting or approval is inferred from a system status.

The approval comes seventeen hours and thirty-five minutes after the 24 January 17:00 target. The written route and factual repair were accepted, but the original readiness target was missed. It would be misleading to celebrate delegation while excluding that operational result.

Ask about the work rather than seek reassurance

Teo uses ten minutes on 25 January after 11:00 reading Hana's first feedback, Morgan's repair and Hana's approval, then sending a focused standards question. He asks what must be checked for this class of internal brief and what can remain Morgan's wording. He does not ask Hana to certify that his private assumption is false.

Hana uses ten minutes on 26 January reading and answering. Her actual answer is:

For this internal brief, I need the required figures and references checked, the recommendation understandable and the existing approval route preserved. I do not require your full stylistic rewrite. The missing footnote did delay readiness, and you retain responsibility for the check. Other types of work may require a fuller review. This answer concerns this brief, not a general rating of your reliability.

Teo uses ten minutes on 27 January reading the answer, updating his selected learning record and sending it to Soraya. Within that block he also sends a separate short operational closure to Morgan and Hana confirming the approved version, late readiness and unchanged approval route. They each use five minutes on 28 January reading it; no further acknowledgment is claimed.

Soraya spends fifteen minutes on 28 January reading Teo's selected record and composing and sending a private comment. She asks him to retain both the clearer standard and the missed deadline. Teo uses ten minutes on 29 January reading the comment and preparing for their twenty-minute joint review on 30 January. Soraya makes five minutes of private notes afterward.

Debrief a mixed result without forcing a breakthrough

The following original exchange occurs within that twenty-minute review, not in an additional uncounted meeting:

Soraya: What did you predict?
Teo: That leaving the wording would lead Hana to call my oversight careless.
Soraya: What is in the record?
Teo: She said the recommendation was clear, required a footnote and held approval. We were late.
Soraya: What meaning did you add?
Teo: I read the correction as disappointment in me. Her later answer did not make that judgment.
Soraya: What changes and what stays?
Teo: I can leave wording with Morgan for this kind of brief. I still need the factual check. I am not ready to generalize to external work.

Teo still feels exposed. He chooses not to schedule another experiment yet. His revised understanding is conditional: a factual review is required for this internal brief; a full stylistic rewrite is not. The result does not prove that Hana will never criticize him or that he has permanently stopped intervening.

Account for sender, receiver and work time

The bounded work totals 430 person-minutes against 515 accepted active minutes, leaving eighty-five within those budgets. The two coaching conversations contribute 130 person-minutes already included; other work totals 300. The joint review is counted once for each person, not again as a separate total.

The account excludes earlier draft creation, underlying business records and existing agreements. It does not measure the full cost of running the team or estimate a saving from leaving wording alone. Morgan's reserved repair capacity is distinct from the actual ten minutes used. No future experiment or additional approval work is counted as performed.

Build an evidence record another reader can follow

Keep the prediction written before the action and retain the actual result beside it. Otherwise, the client may change the meaning of “careless” after seeing what happened. A test becomes difficult to interpret if any correction counts as rejection while any approval is dismissed as politeness. Agree which observation would leave the question open as well as which would bear on it.

A future reader should also know the source of each entry. Teo's account of feeling exposed belongs in the personal interpretation, while Hana's words belong in the work evidence. Morgan can report the task and repair, but cannot establish Hana's unspoken beliefs. Keep the original request, response and version identifiers where they are appropriately held instead of turning a selected summary into the only apparent record.

Choose among three honest interpretations

The prediction is challenged: the stated feared judgment does not occur, and direct feedback specifies a narrower standard. Ask which situation that evidence covers. Repeating a comparable test might help later, but more risk is not automatically the next step.

The evidence is mixed: a correction and an accepted recommendation coexist. Separate quality control from inferred personal rejection. A factual checklist may need improvement even while full rewriting becomes unnecessary. In this case the missed approval target is part of the evidence.

The prediction is supported: if a director actually condemns every imperfection, keep that statement in the account. Do not reinterpret all inconvenient responses as the client's defensiveness. Negotiating a standard, workload or role boundary may be more appropriate than another personal experiment.

If no direct feedback is available, say so. Task completion alone does not reveal an observer's private judgment. Nor should a colleague be asked to diagnose the client's protective commitments. Ask for what the person can know: the work, the observed interaction and the standard they own.

Avoid four design problems

1. Changing too much: a harder task, reduced support and shorter deadline produce a different test from leaving familiar wording alone.
2. Defining success as good news: invite observations that support or complicate the prediction, including an appropriate decision not to proceed.
3. Treating interpretation as a fact: record actual words and decisions alongside the client's reaction rather than merging them.
4. Demanding another round: examine whether further testing would add useful information and whether everyone can actually accept the work.

If the inquiry becomes distressing, offer a real choice to pause, continue differently or stop. Do not compel personal history or interpret silence and tears as proof of progress. Keep within coaching competence and discuss appropriate support when coaching is insufficient. No framework makes an unsafe task necessary.

Keep the selected learning record private enough to use

Coachful is our #1 recommendation for organizing the described coaching workflow. This is an editorial recommendation for that workflow, without a market-ranking study or an assumption-change guarantee. Compare the coaching features with the actual purpose and audience.

Use an agreed worksheet or selected record rather than require the entire reflection to circulate. Released Forms & worksheets support structured responses and a private Coach review. Saving that internal review is different from sending the selected comment, and from the client actually reading it. Preserve the actual communication route.

The individual client profile provides a message route. It does not grant a sponsor access to the client's private hypothesis or supply the colleague's consent. Teo's operational brief and Soraya's selected learning record have different audiences. The demo screens are illustrative, not the case's actual documents or evidence of inspected permissions.

Frequently asked questions

Must the assumption be overturned?

No. A useful test can refine or support part of it. Preserve contrary evidence and real constraints.

Does a safe test avoid all discomfort?

No. Assess actual risks, responsibilities, choices and support. Discomfort alone is neither a requirement nor proof of learning.

How many experiments are necessary?

There is no universal number here. Decide whether another would provide relevant information and can be agreed appropriately.

What if the colleague refuses?

Respect the refusal. Explore another acceptable route, postpone or stop. A coach cannot allocate that person's time.

Does one useful result establish lasting change?

No. This case supports a narrower work standard and an immediate choice, with a genuine deadline miss. It does not establish a lasting developmental or clinical outcome.