The Ladder of Inference

Check the explanation before you act on it

When you're certain why a colleague did something, pause long enough to check. You may have good evidence. You may also have combined a few observations with an explanation that the other person has never heard.

The Ladder of Inference gives you a way to examine that explanation. The model comes from Chris Argyris's work; Rick Ross presented it in the 1994 Fifth Discipline Fieldbook. It traces the steps between noticing something and deciding what to do about it.

I would use it before a difficult conversation, particularly when you've started describing someone's character. Write down what happened and what you've concluded. The gap between those two descriptions tells you what to ask about.

At a glance

• The ladder traces seven steps from what you observed to what you decide to do.
• Most misreadings start low on the ladder, with the details you selected and the meaning you added.
• Before acting on a conclusion about a person, separate what you saw from what you've inferred, and ask about the gap.

How a team can get the same message wrong together

I worked with teams in different time zones. A project update would arrive from the other office while its team was asleep. We'd read it, misunderstand the intention and then spend hours discussing our interpretation with each other.

By the time the other team was available, we'd given that interpretation plenty of attention. Then they would explain what they had meant, and it became clear that we'd misunderstood the update.

The practical problem was the delay between receiving the message and being able to check it. In that gap, agreement within our own team made our reading feel increasingly reasonable. We were discussing the same incomplete information.

If you recognize that pattern in your business, make checking the message easier. Agree how an urgent clarification gets raised across time zones. For a consequential update, include the decision, the reason and what the recipient needs to do. If the written exchange is getting harder to understand, speak directly when both sides are available.

Walk through the ladder

The diagram separates seven parts of the process. Use them to unpack a particular incident; you don't need to recite them in every meeting.

The model's reflexive loop describes how those beliefs influence what you notice next time.

The most useful question is often about selection. Which detail have you kept returning to? Which detail have you left out? Two people can begin with the same meeting and remember different parts because they were watching for different things.

See what happens with a late report

Here's a hypothetical example.

A report arrives after the agreed deadline. You remember that this has happened before. You decide the manager treats deadlines casually, conclude that they can't be trusted with important work and start routing assignments around them.

Before you do that, check the dates and the agreement. Was the deadline clear? Did you approve a change in priorities? Was the manager waiting for information you expected someone else to provide? Did they tell you about the delay in a message you missed?

You still need the report on time. The answers help you choose what to address. If a shared system delayed the input, improve that dependency. If the manager accepted the deadline and failed to raise a foreseeable problem, address the missed commitment.

Your initial interpretation might survive those questions. If it does, you'll have a clearer basis for the conversation. If it changes, you've avoided asking the person to defend themselves against a conclusion you hadn't checked.

What research supports, and where the model needs restraint

Raymond Nickerson's 1998 review of confirmation bias describes how people seek and interpret information in ways that favor an existing belief or hypothesis. The review brings together studies across different settings; the ladder is a practical model for examining your reasoning, with its precise sequence serving as a teaching aid.

The useful warning is that feeling certain doesn't remove the need to look for contrary evidence. If you believe a colleague is careless, a new mistake will be easy to notice. A careful piece of work may get much less attention.

I would build a challenge into any consequential judgment about a person: identify evidence that could change your mind and actively look for it. If every possible response becomes another reason to hold the same belief, you're making the belief difficult to test.

Make your reasoning available to the other person

Start with the observation you can both inspect. Identify the document, the date or the words you heard. Then explain the concern you've drawn from it and ask what you're missing.

Give the person enough information to answer. A question about whether they are “committed” is difficult to resolve. A question about why they changed an agreed priority can lead to a useful explanation and a decision.

Listen for information that changes the problem. If they show that you gave two incompatible instructions, acknowledge that and clarify the priority. If they explain a constraint you hadn't considered, decide how to handle it. If their answer leaves the original concern intact, say what still needs to be resolved.

You don't have to abandon judgment to use this model. You have to make the path to your judgment clear enough that someone else can examine it.

Use it when your leadership team disagrees

Ask each person to write down the facts they think matter before debating the recommendation. Then compare what they selected. You may find that the disagreement begins with different information, different meanings or a different expectation about what happens next.

Keep the decision in view. If the team is debating whether to hire, distinguish the workload you've already measured from the demand you're expecting. Ask which assumption would make hiring the wrong choice and how quickly you could detect that.

Record the decision and the assumption that deserves follow-up. Assign someone to check it. That lets you act while giving the team a clear reason to revisit the decision when new information arrives.

Avoid using the ladder as a way to win an argument. Telling a colleague that they are making assumptions while presenting your own view as pure fact defeats the purpose. Start by examining your own reasoning in front of them.

Keep standards and authority clear

Checking an interpretation can become another form of delay if you use it to reopen a settled expectation indefinitely. Once you've heard the relevant information, make the decision you're responsible for.

If someone has repeatedly missed a clear commitment, describe the pattern and agree the response. You can investigate contributing factors and still require better performance. The model helps you choose a fair response to the actual problem.

It also helps to distinguish a disputed fact from a disputed preference. Your team may agree on the evidence and still weigh customer experience, cost and speed differently. At that point, explain the trade-off and who has the authority to decide.

Try it before your next difficult conversation

Try this

Choose a current concern about a colleague. Write your answer to these questions in plain sentences:

1. What did I directly observe?
2. Which parts of the situation have I paid most attention to?
3. What explanation have I added?
4. What information would change that explanation?
5. What do I need to ask the person before I act?

Bring the relevant observation and your question to the conversation. Let the person answer before you explain the full case you've built. Then decide what the work needs and make the next commitment clear.

Sources

• The Ladder of Inference

  Rick Ross, excerpt from The Fifth Discipline Fieldbook, 1994. Reprinted with permission; hosted by EdReports. Attributes the ladder to Chris Argyris.

• Confirmation Bias: A Ubiquitous Phenomenon in Many Guises

  Raymond S. Nickerson, Review of General Psychology, June 1998. Original article hosted by UC San Diego.