How to Ask for Feedback

A located request comes back with the work’s current place, the gap to the goal, and the next change. Those three only arrive if the request named the situation, the skill being built, a real self-evaluation, and one or two questions.

Four channels

Self-evaluation is the first response, not the fallback. Reliability tracks the skill being judged, so early on this channel is the weakest of the four. Once the skill is there it is the strongest, because an external cycle turns over in days or weeks and a self-generated one closes inside the session. Developing that self-pass is the long-term goal, not a backup for when nobody is free.

The other three channels fill the gaps a session cannot close.

Channel	What it is reliable for	When it earns its place
Self-evaluation	Tracks current skill: low before the skill is developed, highest overall once it is	First response in every situation
Peers	Varies with expertise	Fast, regular, niche questions, and the accountability that keeps a loop running
Experienced reviewer	Complex or persistent problems; the reply only counts if it is acted on	When the session cannot close the gap
General-purpose model	Standard technique questions; weaker on nuance	Before spending someone’s time

Peers vary with expertise; they are fast and regular. An experienced reviewer is spent on a complex or persistent problem, and the reply only counts if it is acted on — reviewers disagree with each other and miss things. A general-purpose model is the thing to try for a standard technique question before spending someone’s time.

Who is asked is the smaller half. The request is the skill, and a Skills Audit is the structured self-pass that comes before one is sent: the skill rated at its current and target level, with the next action set from the gap.

What a request has to carry

Even the best channel returns what it was given. A request that works holds two things in balance: enough context that the situation is understood, and enough focus that the reply lands on what matters.

Four combinations, and only one of them is the aim. Broad focus with little information comes back as general comment on everything, hard to act on. Narrow focus with little information comes back precise and aimed at the wrong thing. Broad focus with plenty of information comes back thorough and unfocused — more context is not the lesson. Narrow focus with enough information comes back as targeted improvement on the skill the goal names. Naming the goal is what gives the reviewer something to evaluate against.

An unstructured request is not a neutral one

A vague request returns vague guidance, and the risk is real without being absolute. More than a third of feedback interventions leave performance worse than before. The harm concentrates where the reply is about the person rather than the work. Feedback aimed at the self is the weakest kind. A request answered badly can cost more than it returns, and the reply that lands on worth rather than work is the one to expect after an unlocated ask. The structure is not manners. It is what keeps the reply on the task.

Five parts of a request

The house form is five letters. Each letter is something the request does.

C, context. Where the work sits, what is coming — an exam, a deadline — and the circumstances that shape what a good answer looks like.

L, learning goal. The specific skill or process being built. A goal about how the learning happens outlasts a goal about what gets produced: getting better at asking questions, at a mapping technique, at reflecting on what works. Naming that goal is the feed-up the rest of the reply sits against.

E, evaluate the work. This is the longest and most important step, and it costs more than a quick ask. On a question with an obvious answer the write-up is not worth doing — a search of existing notes is the cheaper move. Four prompts do the work: what was tried; what went well and what was hard; how that compared with what was expected; where the most important improvement looks to sit. The third prompt is load-bearing. A reply is only usable against a stated expectation, because the gap between what was predicted and what happened is the thing that changes how the work goes. “I don’t know” is not an answer: if it were obvious there would be no request. How the work is evaluated tells the reviewer about Metacognition: The Control Layer — the monitoring layer the reviewer is reading when they read the reflection — and is often more diagnostic than the work itself. It also forces engagement with the problem before the problem is handed over.

A, ask targeted. One or two questions. Where many areas feel uncertain, the high-yield ones are the ones that ship, because improving many things at once is not possible. The second question can test the self-assessment itself: is this read right, or is something missing.

R, review and follow up. Absorb before responding. Clarify what is unclear. Stay on the work when a point lands hard.

Two requests, same work:

How is my exam prep going?

History exam in three months, prepared with interleaved retrieval. Re-reading, summary pages, textbook questions, and a friend's quiz. Big-picture ideas hold; dates do not. Recall is inconsistent — roughly seven in ten come back, and nearly all the loss is detail. Hypothesis: re-reading is eating the time the other methods need. How to improve recall of fine detail without losing the big picture? Is that read right, or is something missing?

The first returns a comment. The second returns a plan.

After the reply

Feedback improves performance only when it is used. The loop is not closed until a next attempt uses it. Three moves: read it objectively, especially where it disagrees; build a plan with priorities and specific experiments; test, review, and return only when the experiments are used up. Receive, analyse, implement, produce new work, receive again.

The plan a reply produces is a hypothesis. One week of that hypothesis, in the same exam-prep case: re-reading dropped, thirty minutes a day of high-yield cards, summary pages replaced by maps. Inside the week, thirty minutes was not enough and the maps were too hard to build. Time went to forty-five minutes; the maps became a hybrid of short linear notes and small maps. The revision is what “test and review” means. A heavy reply in that same case was handled as material for a plan rather than as a verdict, because the comments were about the methods and not about the ability.

Marginal Gains is how big the next experiment should be: the improvement small enough to be almost certain, then stacked. Kolbs Experiential Cycle is how that experiment gets reflected on — what happened, how it went, what rule that suggests, what changes next. Three or four experiments before returning on a complex problem; weeks of independent work on a simple one. Those are orders of magnitude, not law.

If the last reply has not changed what happens in the next attempt, the loop is open and another request will not close it. Run the experiment instead.

A located request comes back with something that can be acted on in the next session. A reply that is general praise, or “keep going,” is a report on the request.

Giving a reply pays the giver. It forces critical evaluation and forces reflection on the methods already in use; other people’s work widens the approach. Four traits of a reply worth giving: specific with examples, strengths before suggestions, on the work rather than the person, and carrying concrete next steps.

The skill being built is the self-evaluation. The external request is scaffolding for it. The endpoint is an ask that checks a read rather than sourcing one. That monitoring and adjustment is Self-Regulation — the parent loop this sits inside.

Open Questions

What standard of self-evaluation is good enough before escalating.

Sources

• Hattie, J. & Timperley, H. 2007. “The Power of Feedback.” Review of Educational Research 77(1), 81–112. The three questions a reply must answer (where going / how going / where next), and self-level feedback as the weakest of four levels.
• Kluger, A. N. & DeNisi, A. 1996. “The effects of feedback interventions on performance.” Psychological Bulletin 119(2), 254–284. More than a third of feedback interventions reduced performance.
• Wisniewski, B., Zierer, K. & Hattie, J. 2020. “The Power of Feedback Revisited.” Frontiers in Psychology 10:3087. Revised effect is medium and highly variable; a minority of effects remain negative.
• Butler, D. L. & Winne, P. H. 1995. “Feedback and self-regulated learning.” Review of Educational Research 65(3). Internal feedback is the regulator external feedback has to enter.
• Boud, D. 2000. “Sustainable assessment.” Studies in Continuing Education 22(2). Learning to judge your own work is the long-term purpose.
• Boud, D. & Molloy, E. 2013. “Rethinking models of feedback for learning.” Assessment & Evaluation in Higher Education 38(6). The loop is not complete until a next attempt uses the reply.
• Sadler, D. R. 1989. “Formative assessment and the design of instructional systems.” Instructional Science 18. Formative assessment as building evaluative knowledge.
• Seijts, G. H. & Latham, G. P. 2005. “Learning versus performance goals.” Academy of Management Executive 19(1). Learning goals outperform outcome goals on novel and complex work.
• Topping, K. 1998. “Peer assessment between students in colleges and universities.” Review of Educational Research 68(3). Peer assessment as a learning activity for the giver.