 I must admit there were a couple of times on our train journeys in Hamburg when we experienced confusion and some anxiety, but this did lead to us solving our own problems. So it made me wonder, is it possible that confusion could be good for learning? I had to think very carefully about this. I have a theory of metacognitive red flags. If red flags wave in your mind, then you notice that something's wrong. Either we're going in the wrong direction, or we're stuck, or we've misinterpreted something. Confusion is a kind of red flag. This flag says, I need to stop and think about what I'm doing. If you recognise that you're confused, this is a red flag that makes you think, I don't understand, and that is the key. Rather than letting the confusion overwhelm you as a kind of negative effective state that stops you from thinking, if you can stop, pull back, and ask, what am I confused about? What is it that I don't understand? That's when confusion is really useful. The first red flag is one that signals a lack of progress. I'm stuck. The second red flag should prompt error detection and then correction. I've made a mistake. And the third red flag should trigger a reassessment of the strategies we're using. This isn't working. It doesn't make sense. That's why I encourage students to question themselves and generate their own feedback. And this needs to become a habit of mind. For deep learning, it's important for us to be able to have the skills to critically assess our own performance. We can't always rely on others to tell us how we're doing. Self-regulated learners who employ deep learning strategies work towards having a vision of high quality and what they would like their own work to look like. They can identify their own mistakes. They devise ways to improve and gather resources to help them reach their goals. All learners need to be able to identify how our work differs from what we would like it to be. Think, for example, about a pilot. After many, many hours of training and flying experience, she has a deep understanding of the finer details of her job. She has a concept of what high quality flight should look like. While there are many forms of feedback received in the cockpit, it's the pilot's knowledge and experience that guides her to make refinements, especially when very quick decisions are required. Professor Roy Sadler refers to this as tacit knowledge or a deep understanding of quality and development of competence without explicitly being told by a teacher or authoritative other. Developing this skill in our students is a very important part of deep learning. So ultimately, what we really want is to encourage students to choose deep learning strategies. And the best way to do this is for teachers to model this themselves and then invite students into that process. This means modelling metacognitive habits of mind and doing it in a way that captures the messiness of thinking. One of the drawbacks of using a textbook, for example, is what students see is an idealised version of thinking within the discipline. There are no mistakes, but we know that creativity is not linear and problem-solving is not linear. We encounter many blind alleys and we go off on tangents to explore something and then come back to our original problem. And this includes making mistakes. So often students grow up believing that mistakes are bad, but mistakes can be a very important source of learning. Sometimes students can be guilty of metacognitive blindness if they fail to notice that something is amiss. Or they might commit metacognitive vandalism by taking a destructive action to deal with a roadblock. Or even that they see problems which actually do not exist. Or metacognitive mirages which lead them to reject correct answers. Recognising you're in this situation can be productive if it triggers some corrective action on your part. This is when mistakes can help you to learn and why confusion or puzzlement can be a productive part of deep learning.