6 Principles of Effective Feedback

A common misperception about expertise is that it is merely experience. But this cannot be the full story as even among experienced veterans there are levels of competence and incompetence. Additionally, in some domains, such as individual stock prediction or long-term political forecasts, intuitive expertise never develops (Kahneman and Klein, 2009). 

Experience needs to be shaped through effective feedback loops in order to be meaningful. Without it, experience is an exercise in futility, and when feedback is given poorly, it can stall progress and hurt performance.

But what does effective feedback look like? To help answer that question, I have identified 6 non-exhaustive principles for creating effective feedback.

1.    Effective feedback is based on process, not outcome

There are two types of feedback: outcome feedback focused on what happened, and process feedback focused on how it happened. Of the two, process feedback is superior.

In her book, Annie Duke (2019) recalls her poker coach interrupting her before she revealed whether she won a hand or not. He didn’t want to know the outcome, only her observations, reasoning, and decisions. The final card was just luck and so had nothing to do with the quality of her decision.

We apply the same principle in the Tactical Decision Games (TDGs) over on Shadowbox Decision Games: Warfighters. Solutions are evaluated on their reasoning, not on their outcome. One reason for this is that a trainee may overlearn from the way an enemy behaved in one particular exercise when what we need is for them to improve their ability to anticipate the enemy intentions, locations, and capabilities across many possible situations. 

Outcomes are emotional, memorable and salient, but the reason for an outcome can be ambiguous. Because of this, trainees with only outcome feedback may struggle to identify what matters, focus on the wrong things, and learn the wrong lessons. In one of our safety projects, we found operators would often know the right answer during routine operations, but had no idea why it was correct leading to an inability to adapt to new situations. On the other hand, effective process feedback highlights what matters and why, making relevant variables as salient as a red hot stove after burning your finger. When the reason for an outcome is salient enough, you only need to learn once. 

2.    Effective feedback focuses on cognition

Great process feedback doesn’t just correct actions, but deepens understanding by shaping their mental models—that is, their causal understanding of how things are related—enabling deep insights into different situations. The insight a trainee has into the causal relationships can be more effective than mere “muscle memory”.

Consider Gary Klein’s friend “Jimmy” who asked for help training his backhand in racquetball. After watching Jimmy lunge, swinging wildly, never putting himself in a position to hit the backhand cleanly, Gary announced they were done playing. Instead, Gary would hit the ball to him, and Jimmy was to observe where the ball went. As Jimmy watched, he began to better understand the movement of the ball. Gary then asked him where he should place himself to hit it, which Jimmy dutifully identified. This improved his confidence, and when Jimmy asked to hit the ball again, Gary consented. This exercise turned out to be what he needed. With his newfound cognitive skill, Jimmy could now hit the ball consistently with his backhand. (Klein, 2011)

Without a good mental model, experience teaches nothing and practice is a practice in bad habits. Without a causal understanding, we can’t identify what is relevant, and will mistake saliency for relevancy which results in practicing the wrong thing. Conversely, with a good causal understanding we can make sense of feedback. Good feedback simultaneously depends on and improves our causal understanding of the world. 

3.    Effective feedback encourages active learning

So far we have assumed the feedback is externally sourced, and in the earlier example Jimmy depended on Gary’s expert feedback to improve. But is that best?

Well, no. A trainee needs to learn to get and interpret feedback on their own. Coaches are not always around, and even if they were, they cannot see, hear and feel everything, nor can they communicate all their tacit knowledge. Schmidt and Wulf (1997) even found that continuous, concurrent feedback helped boost the learning curve during training, but led to minimal transfer to the task situation.

The very idea of feedback can suggest a passive learner, which is misleading. Expertise is fundamentally about increasingly refined situational discrimination by being able to recognize many different types of situations and how to respond to them. Developing such discrimination requires actively seeking and interpreting feedback from the world.

A trainee that relies on external commentary cannot ever develop the cue detection and mental models necessary to be a true expert. While effective coaching may be necessary, at the end of the day expertise has to be taken, not given.

4.    Effective feedback is grounded in reality, not oversimplifications

Reality is messy, and so “[w]e all simplify the world. We chop events into artificial stages. We pretend that simultaneous events are sequential, that dynamic events are static, and that nonlinear processes are linear. We deal separately with factors that are interacting with one another” (Klein, 2011). These simplifications are necessary, but our feedback cannot be based on them.

Consider a simple definition of decision-making as “a choice between options.” If this is decision-making, we can provide feedback on how to improve it. For example, consider the Army’s Military Decision Making Process (MDMP): first you gather information, then generate three courses of action, establish a set of criteria, create a matrix to rate the options, and finally choose the option with the highest score. This is just decision-making 2.0; a choice between options with some additional steps to overcome known failure modes. 

But as John Schmitt has recounted of his experience with the MDMP, “It made no sense to me.  As a platoon commander, in the middle of a firefight, with mere seconds at my disposal and lives at stake, I was expected to undertake this exercise in ranking and tallying? It seemed patently ridiculous. I had never made a decision that way in my life, and I doubted whether Hannibal, Bonaparte or Rommel ever had either.”

The MDMP is “ridiculous” because it is based on a false simplification that says decision-making is a choice between options. But most real world decision-making looks nothing like that, and is better explained by the RPD model. When feedback is based on a simplification of reality, you get absurdities like the MDMP. 

This is why, whenever possible, experience and feedback need to be as grounded in reality. Our mental models develop through exposure to the actual structure, constraints, and causal relationships of a task environment. When feedback is tied to concrete context and meaningful relationships, it helps trainees refine their internal representations (or sensitivity to the constraints) and make better decisions in the future.

5.    Effective feedback is at least as variable as the situations which will be faced

Experts may not be able to adapt to novel domains where they do not have experience, but they must be able to adapt to novel problems within their domain. Such adaptive expertise requires variations in the type of situations a trainee faces.

Consider the opposite: a basketball player that spends hundreds of hours practicing a free style shot to get as consistent as possible at shooting hoops. Or perhaps a firefighter that fights the same fire day after day. The danger in both cases is that the real world is not consistent. Bodies change, tire, and get injured. Fires vary in their cause, intensity, and how they move through structures. 

Repetitive practice mistakes two very different things; consistency and adaptivity. When someone is adaptive they can seem consistent, but in reality they are making hundreds of micro-changes to ensure consistent outcomes under varying conditions. In many professional sports, the best players in the world are practicing with different kinds of constraints to make sure they can play more adaptively.

Without variety, novel problems seem impossible because our mental models are overfit, stale, and impoverished. We lose the ability to adapt to small changes, and even more so large ones. Whether through repetitive practice, or through dogged adherence to standard procedures, learning plateaus without variety, and we lose the adaptivity necessary for consistently high performance in a world that is very inconsistent.

6.    Effective feedback is about the quality of learning, not the quantity of success or failure.

What’s better for learning: success or failure? Before I answer, it might be worth thinking through how you would answer this question using the above principles. 

Ready?

Whether an outcome is a success or failure is irrelevant to learning. What matters is why an outcome happened and how to deal with similar situations in the future. Remember, process not outcome. Good feedback changes our mental models. Sometimes success sharpens that model, sometimes failure will, but neither has an inherent advantage. 

There is a nuance, though. Situations which are too easy or too hard may pose a challenge for trainees to interpret as the reason for the outcome becomes opaque. If the trainee changes their behavior and they see no difference in results, it is difficult to diagnose what aspects of their performance are necessary, and which are superfluous. 

This could be why some of our best learning happens right at the edge of our abilities. In that zone where we dance between failure and success, we have just enough control to disambiguate what is happening while still pushing ourselves into uncharted territory. But do note that this has nothing to do with the ratio of success to failure, and is instead about the difficulty of extracting signal from the noise.

Motivation, however, is a different story. Too much failure demoralizes, and too much success breeds complacency. Even if the success/failure ratio doesn’t affect learning directly, it may impact it indirectly through motivation. If success or failure gets in the way of a trainees sense of progress, achievement, comradery, fun, empowerment, meaning, and purpose, that is a problem which must be addressed. 

In the end, the goal isn’t to aim for a certain amount of success or failure, but to create conditions where people stay engaged, extract insight from experience, and keep refining how they understand the world.

Conclusion 

Feedback isn’t about reinforcing behavior, but about shaping cognition. When done well, it helps trainees make sense of situations, refine their mental models, and build the capacity to adapt under real-world conditions. The best feedback doesn’t just say whether the outcome was good or bad, but why things unfolded the way they did, and how to handle such situations in the future. 

Of course, this is not the whole picture. Effective feedback also depends on factors like timing, delivery, modality, and the trainee’s mindset. It benefits from spacing, interleaving, cognitive fidelity, and well-designed challenges that push people to the edge of their competence without pushing them over. Additionally, applying this all in practice may not always be possible. 

What I’ve offered here isn’t a comprehensive framework, but a set of core principles to help readers think more clearly about what effective feedback accomplishes. The key is this; feedback works when it helps us to make sense of and adapt to a complicated reality. If you keep that in mind, everything else falls into place.

References

Duke, A. (2019). Thinking in bets: Making smarter decisions when you don’t have all the facts. Portfolio/Penguin.

Kahneman, D., & Klein, G. (2009). Conditions for intuitive expertise: A failure to disagree. American Psychologist, 64(6), 515–526. https://doi.org/10.1037/a0016755

Klein, G. (2011). Streetlights and shadows: Searching for the keys to adaptive decision making. MIT Press.

Wulf, G., & Schmidt, R. A. (1997). Variability of practice and implicit motor learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(4), 987–1006. https://doi.org/10.1037/0278-7393.23.4.987