Why is looking inward not enough to know yourself?

What research shows about introspection

Nisbett and Wilson gathered a series of experiments in 1977. People decided, were asked for the reason, and answered at once. The answers were fluent, plausible and often wrong. Nobody noticed the difference. There is a word for inventing a reason like that, konfabulation. We build an explanation that fits us and take it for the real cause.

Pronin, Lin and Ross followed in 2002. We spot bias in other people easily and in ourselves barely at all. Pronin and Kugler described in 2007 what that hangs on: for ourselves what counts is what we think, for others what counts is what they do. West, Meserve and Stanovich checked in 2012 whether being clever protects you. It does not, and in some measures cognitive ability made the blind spot larger.

Trivers supplied the frame for this in 2011. Self-deception has a function. Someone who deceives themselves deceives others more convincingly, because no bad conscience travels along with it. That explains why looking inward feels so reliable and is so unreliable.

I know this blind spot from my own experience, long before I knew the studies. In a video from 2017 I say almost exactly that: by that point I had analysed myself psychologically so often, and I think that is maybe one of the reasons why nothing actually changed in my behaviour. The analysis felt like progress. It changed nothing.

The access you actually have

That sounds like a dead end. It is not. Nisbett and Wilson themselves distinguish two kinds of access.

Access to processes is missing, access to content (what I believe, what I predict) is not, and content can be checked externally.

I cannot reach the process. I can reach the result. I can say: the client decides by the end of October, seventy percent. In four weeks I can lay that sentence next to reality. That is what the Brier score is for in world-modeling. It does not measure how certain I felt, it measures how often my numbers were right. The book is open and free, with no publisher, in two versions.

These studies do not sit in a collection of confirmations. They sit in wager dossier 3, a red team against my own book that collects the strongest counterarguments in their best form. Introspection runs there as line of attack 3 of 6.

What a commitment needs to deliver

A commitment only counts if it is fixed before the check and if a stranger could check it. Three details are enough: what exactly, by when, and how you recognise it was met. Leave one out and the story takes over again.

The lever check on this site asks four questions, and the fourth one is: how will you notice in four weeks that something has moved? It asks for a number or a date before you start. Nothing is stored and nothing is sent, the text stays in the browser. The value is in the sentence you write down beforehand.

• Take a decision that is genuinely coming up this week.
• Write down beforehand what you expect. One sentence, concrete enough for a stranger to check.
• Add a number for how sure you are. Any number will do.
• Set a date for the check and put it in the calendar, not in your head.
• On that date mark one of three results: happened, did not happen, not assessable.
• After three months read all the entries in one go. Your model of the world is in there, not in your memory.

The [[wort-commitment|word commitment]] at school

At the Freie Schule Salzburg the reifestufen include a form of this that I like. A child formulates what it will do, concrete enough to be recognised. A mentor may ask whether the sentence is clear, but may not assign anything. The child decides whether the promise is scored and picks the weight, one to five points. Only then comes the check.

The check has three results: kept, not kept, not assessable. The third one matters most. Sometimes circumstances get in the way, sometimes the conditions were unclear. Without an honest place for that, children learn to give themselves nothing but soft promises.

A word that was not kept is documented honestly and can be repaired, with no minus points.

A hard boundary belongs with it. Word commitments apply only to voluntary, concrete plans, never to academic performance or general behaviour. Otherwise it would be a grade under a different name.

On record

All five papers are cited in full in wager dossier 3 of the World Modeling project, dated 6 July 2026. Paideia handbook, chapter 07, status September 2026.

• Nisbett and Wilson (1977), Psychological Review 84(3), 231-259: no introspective access to your own cognitive processes, but fluent and false explanations.
• Pronin, Lin and Ross (2002), Personality and Social Psychology Bulletin 28(3), 369-381: bias is clear in others and barely visible in oneself.
• Pronin and Kugler (2007), Journal of Experimental Social Psychology 43(4), 565-578: own thoughts count, own behaviour is overlooked.
• West, Meserve and Stanovich (2012), Journal of Personality and Social Psychology 103(3), 506-519: cognitive ability does not shrink the blind spot.
• Trivers (2011), The Folly of Fools, Basic Books: self-deception as a function, because it makes deceiving others easier.
• Paideia handbook, chapter 07: only voluntary, concrete and dated promises are scored, with the result kept, not kept or not assessable.

Sources and links

This note keeps growing

2026-09-03: Deepened with a personal experience from 2017: a lot of self-analysis without behaviour change, years before I knew the research on it.

2026-09-03: Grown: the distinction between access to process and access to content, Pronin and Kugler 2007 and Trivers 2011 alongside it, the five steps of the word commitment with its repair rule and its boundary, the exercise for tomorrow and the lever check as a live example.

2026-09-02: Planted from the steelman dossier for World Modeling.