Directory · F
Not sure this is the right entry? Run the check · New here? Start with the premise
Feedback
2 min read · 340 words
Feedback is information about what your output actually produced. It usually arrives wearing judgement's clothes, which is why most of it gets defended against rather than used.
Is this you?
Check whatever lands. Nothing is sent anywhere.
Feedback is data returning from the environment about the system’s output.
The organism produces output — behavior, work, communication, decisions. The environment responds. That response is feedback: information about what the output produced. Not judgment, though it often arrives wearing judgment’s clothes. Information. The output landed here. It produced this effect. The model was this accurate.
The Criticism entry covers the identity defense system’s response to negative feedback. This entry covers the mechanism itself: the signal returning from the world that tells the system how its output is being received.
Feedback has two sources, and the system processes them differently.
External feedback — the response from other systems and from environmental conditions. This is the data the organism cannot generate internally. The plan looked good in the simulator. The feedback says it didn’t survive contact with reality. The communication felt clear from the transmitting end. The feedback says it arrived distorted. External feedback is the only source of information about how the output actually landed, which the internal model cannot accurately predict.
Internal feedback — the system’s own assessment of its output. The body reports: this movement was efficient or it wasn’t. The alignment gauge reports: this activity produces the meaning signal or it doesn’t. Internal feedback is real-time data about the organism’s own operating state during and after the output.
The operator benefits from both. The system that only processes internal feedback is operating blind — its self-assessment is running through the same filters and biases that shaped the output. The system that only processes external feedback has outsourced its assessment to other systems whose filters are equally biased, just differently.
The balanced read: receive external feedback as data about the output’s effect. Receive internal feedback as data about the operating state. Compare both against the intended result. Adjust. This is the iterative loop that produces improvement — not the single data point but the accumulated pattern across many cycles of output and response.
Information about the gap between intention and effect
You produce output — work, behaviour, communication, decisions — and the environment responds. That response is data about what the output actually did: where it landed, what effect it produced, how accurate your model was. Not a judgement, though it frequently arrives dressed as one.
That distinction is the whole difficulty. The identity defence evaluates incoming assessment for threat before it evaluates it for accuracy, so the sting arrives before any consideration of whether the point is correct — and both valid and invalid feedback feel identical in the first minute.
Which produces two expensive failures. One is defending rather than evaluating, so accurate data gets discarded along with the inaccurate. The other is avoiding the channel entirely: not asking, not recording, not checking — which feels safer and leaves you operating with no information about what your output is actually doing.
The practical move is to make it specific and to ask before it's volunteered. Vague requests produce vague answers — it was great is a social response, not information. A narrow question about one element gets you something usable, and it's far easier for the other person to answer honestly.
What to actually say
Ask narrowly. The broad version gets reassurance, which is pleasant and useless. Naming the specific element gives someone permission to be honest about it.
- Instead of "what did you think?"
Was the second half clear, or did I lose you somewhere?
- Making honesty easy
I'd find one thing that didn't work more useful than five things that did.
- Asking about your effect on someone
When I said that in the meeting, how did it land? I genuinely can't tell.
- Receiving it well
That's worth taking seriously. Let me think about it properly and come back to you.
- When it's accurate and unwelcome
That's fair. I'd rather know.
- Asking for it ongoing rather than once
Could you tell me when you notice this happening, rather than saving it up?
Ask one narrow question
You're testing whether specific requests produce usable answers. They do, reliably.
Real questions
Why does everyone tell me my work is great?
Because you asked a broad question and broad questions get social answers — what did you think is an invitation to reassure, and most people take it, partly out of kindness and partly because honest critique is socially expensive. Narrow it to one element and say explicitly that one thing that didn't work helps more than five that did. People are usually relieved to be given permission.
How do I stop being defensive about feedback?
Not by trying to not feel it — the defence fires before any evaluation happens, and it will fire for accurate feedback as readily as for unfair feedback. What's available is not acting from it for ninety seconds. “That's worth taking seriously, let me come back to you” buys the window where the evaluation can actually run, and costs you nothing if the point turns out to be wrong.
How do I separate useful feedback from someone being unkind?
Strip the delivery and look at what's left. Unkindness lives mostly in tone, audience and generalisation — and underneath it there's frequently still a real claim. Write only the factual assertion and judge that on its own. Two things can be true at once: poor delivery doesn't make a point wrong, and you're entitled to address how it was said separately from whether it was accurate.
Why do I avoid asking for feedback?
Because not knowing feels safer than finding out, and in the short term it is. The cost is that you're operating with no information about what your output actually does, which means errors persist and go on compounding. It's also why people can be genuinely surprised by an assessment that everyone around them had formed years earlier — the channel was open and never used.
Should I act on all feedback?
No, and treating it as instruction rather than information is its own error. It tells you what your output produced in that person, which is real data and not automatically a mandate — their reaction is partly about them. Sort it into true, partly true and false, act on the first two, discard the third without argument. The partly-true pile is usually the biggest and the most worth separating carefully.
How do I give feedback without the other person shutting down?
You can't prevent the defence — it fires in them exactly as it does in you — but you can reduce what it has to defend against. Be specific about behaviour rather than character, give one thing rather than a list, pick a moment without an audience, and say plainly what you want to be different. Then expect the first two minutes to be the alarm talking, and don't judge whether it landed by that.