How does "Notice which data you selected — and which you ignored" work?

Attention is finite, and selection is always happening, usually in the direction of information that confirms existing beliefs. The data you selected feels like "the data" — the rest does not feel excluded, it feels absent. Deliberately searching for the data you did not attend to is the countermeasure to selection bias operating at the bottom of the ladder.

How does "Surface and examine the assumptions you are making" work?

Assumptions are the invisible links in the inferential chain — the beliefs you did not consciously choose but that are required for the interpretation to follow from the data. They are often so obvious to the holder that they are not recognized as assumptions at all. Making them explicit converts them from invisible architecture to examinable hypotheses, which is the minimum condition for checking whether they are warranted.

How does "Test conclusions before acting on them" work?

Conclusions reached via fast ladder-climbing feel certain — not like hypotheses to be tested but like descriptions of reality. Acting on untested conclusions generates the costly errors the ladder model predicts. Treating the conclusion as a hypothesis, even briefly, reactivates the uncertainty that the fast climb suppressed and makes a small amount of verification more likely.

How does "Walk down the ladder together in disagreements" work?

Disagreements at the conclusions level are rarely solvable at that level — two people have climbed different ladders from different starting data. The productive move is to descend together to the data level and find the lowest rung where agreement holds. Disagreements that look irreconcilable at the top of the ladder often resolve quickly once it is clear that the parties selected different data or made different assumptions.

How does "Slow the reflexive loop" work?

Argyris identified a reflexive loop: beliefs at the top of the ladder shape the data selection at the bottom, which generates evidence that confirms the belief, which strengthens the selection filter. Without interruption, the ladder becomes more entrenched over time. Breaking the loop requires deliberately selecting data that the current belief would filter out — which is uncomfortable and requires a specific prompt to do.

Chris Argyris

What is the ladder of inference and how does it help you reason more carefully?

From raw data to action — and how to catch the errors that happen in between

Short answer

Chris Argyris's ladder of inference describes the rapid, largely invisible mental journey from raw observable data to a firmly held belief and action — selecting data, interpreting it, making assumptions, drawing conclusions, and acting, often in seconds. The practice is to slow this climb and check each rung, especially in high-stakes situations where conclusions feel certain but may be built on shaky selections and assumptions.

Chris Argyris introduced the ladder of inference in the 1970s to explain why intelligent people in organizations reach contradictory conclusions from the same events. Peter Senge popularized it in The Fifth Discipline (1990). The ladder has seven rungs: observable data, selected data, interpreted data, assumptions, conclusions, beliefs, and action. The problem is that the climb happens automatically and almost instantly, making the intermediate steps invisible — so we act on conclusions that feel like observed facts. Here are the practices that make the model actionable, with honest evidence.

The practices (6)

Why it works

The first and most foundational error on the ladder is treating interpretations as if they were observations. "She was dismissive" is an interpretation; "she did not respond to the last two emails and cut the meeting short by fifteen minutes" is data. Interpretations can be wrong; data can be checked. Confusing them makes disagreements about interpretations feel like disagreements about facts, which are much harder to resolve.

How to do it

1When you describe a situation, ask: what would a video camera have captured here?
2Write the observable events separately from your interpretation of them.
3Test whether two observers with different priors would see the same "data" or different ones.
4When in conflict with someone, find the lowest rung you both agree on before discussing interpretation.

Evidence

Mechanistic

The distinction between description and evaluation is fundamental in behavioral feedback research and in nonviolent communication frameworks. Reducing evaluative language in conflict situations is associated with reduced defensiveness, consistent with research on the impact of labeled vs. described behavior.

Honest caveat: Perfect separation of observation and interpretation is impossible — all perception is theory-laden at some level. The practice aims for better separation, not pure objectivity.

— Rosenberg (2003), Nonviolent Communication — observation vs evaluation as a foundational step

Common mistake: Believing that because something feels obviously observable, it is — "she was hostile" feels as concrete as "she raised her voice," but the former is a judgment.

Go deeper on this practice →

Practice this with IX Coach

Reading about a practice changes nothing on its own. IX Coach turns these into a guided, adaptive routine — discerning where you are in real time and walking the practice with you, session after session.

Start with IX Coach →