## Highlights *Tetris* players are better today because their environment enables it. Video hosting allows detailed demonstrations of the best play to be broadcast widely. Online forums transform informal conversations into permanent reservoirs of knowledge. Livestreaming encourages extensive practice, including near-instantaneous feedback from an audience increasingly knowledgeable about the top techniques. ([View Highlight](https://read.readwise.io/read/01hy3ww5331asfz00y01c8kenz)) The Three Factors for Getting Better at Anything As the story of *Tetris* illustrates, improvement depends on more than just talent or tenacity. There are three factors that determine how much we learn: 1. **See.** Most of what we know comes from other people. The ease of learning from others determines, to a large extent, how quickly we can improve. 2. **Do.** Mastery requires practice. But not just any practice will do. Our brains are fantastic effort-saving machines, which can be both a tremendous advantage and a curse. 3. **Feedback.** Progress requires iterative adjustment. Not just the red stroke of a teacher’s pen, but contact with the reality we’re trying to influence. ([View Highlight](https://read.readwise.io/read/01hy3wx5qrvc764qhsdbre23wj)) Contrary to our species’ presumed superiority, the researchers found that children and apes performed similarly on puzzles involving spatial, quantitative and causal reasoning, with the chimps even edging out the children on some tasks. The clear exception was in social learning, where the toddlers were easily able to solve a problem when given a demonstration, but virtually none of the apes could. ([View Highlight](https://read.readwise.io/read/01hy3wzx8e8knyesjs8feev1j4)) As historian and chemist Lawrence Principe writes, “Alchemy’s primary sources present a forbidding tangle of intentional secrecy, bizarre language, obscure ideas, and strange imagery. The alchemists did not make it easy for others to understand what they were doing.” ([View Highlight](https://read.readwise.io/read/01hy3x2fm2rxrd8dx9bnf6sc1h)) They omitted, swapped, or added unnecessary steps to confuse unskilled readers. While this certainly had the intended effect of limiting knowledge to a privileged few, it also prevented the accumulation of reliable knowledge. ([View Highlight](https://read.readwise.io/read/01hy3x4x17r88esjp52tm50wny)) Knowledge is not evenly distributed. Despite the promise of the internet age, the majority of the world’s knowledge isn’t written down and freely available. Instead, it’s locked away in the minds of experts, many of whom would struggle to articulate what they know. Knowledge is often not inside the head of any individual at all, but embodied in practices spread over groups. ([View Highlight](https://read.readwise.io/read/01hy3x760ew14knwwpebc693j2)) - 💭 is this still th case with ai being able to resurface what is relevant? bruh it got answered instantly lmaoo the world of book knowledge becomes increasingly accessible, but the tacit understanding of unspoken practice remains within cloistered communities of experts. Access to the environments where knowledge resides is often a bigger hurdle to mastery than learning itself. ([View Highlight](https://read.readwise.io/read/01hy3xayeqj424zhg0gyyjve9c)) - 💭 so then creators try to productize their information and services, which does make it more accessible technically, as long as you do have the money while we’re fantastic imitators, many parts of a skill cannot be emulated. How the arm moves in a tennis serve or how the wrist flicks in a brushstroke can certainly be observed. But everyone’s musculature is unique, so these observations are only approximations of how you must perform the skill yourself. Perceptual skills like discriminating patterns in an X-ray or predicting the path of a golf ball rolling down a green have large tacit components that can’t be easily communicated, even with a patient teacher. Hands-on practice is essential to master the aspects of skills that can’t be taught in a book. ([View Highlight](https://read.readwise.io/read/01hy3xdjrgt9ejr91hck06cc4h)) - 💭 reminds me of the ecological approach where we have to understand each person's unique affordances and how they learn through constraints led approach first then helps with procedural knowing A couple hundred books and several hundred academic papers later, I’ve found satisfying explanations for some of my earlier puzzles. But, in keeping with all quests of curiosity, new questions have multiplied in their place. This book is, in many ways, an effort to make sense of what I’ve found. ([View Highlight](https://read.readwise.io/read/01hy3xmmy27bbfy2hsff3d9qym)) - 💭 book is a result of pure curiosity Each chapter is expressed as a simple maxim. My hope is that long after the details of the research explored in each chapter have faded, these rules of thumb will serve as both a reminder and a useful—if imperfect—summary of the key principles. ([View Highlight](https://read.readwise.io/read/01hy3xpx0e1mv52db77kn6rz9h)) - 💭 reminds me of feel good productivity. distill research into simple advice you can deploy as needed ### New highlights added May 18, 2024 at 12:38 PM In their study of problem solving, Simon and Newell observed a number of generic problem-solving strategies that people applied to a diverse set of problems. They argued that people use these strategies as a fallback when more specific methods are unavailable. Simon and Newell called these weak methods, to contrast with the strong methods of guaranteed algorithms or domain-specific heuristics that drastically cut down the problem-solving search. These weak methods include generate-and-test, means-ends analysis, planning, and hill-climbing. ([View Highlight](https://read.readwise.io/read/01hy6fvnbqxpg0x4gbry6pmj3h)) Means-ends analysis works by alternating between a goal, observing a difference between the current state and the goal state, and then finding a suitable method to close the distance. This can repeat in a recursive fashion, as Simon and Newell’s story illustrates. ([View Highlight](https://read.readwise.io/read/01hy6fzx33054ckd3mcj9hq9re)) Planning can be seen as reformulating a problem in a simpler problem space, solving it in the simpler space, and then trying to generalize that approach to the real problem space. For instance, when writing an essay I may start with an outline, which is a kind of simplified version of the essay, only including the main points I want to state and ignoring all the detail. Once I’m satisfied that I’ve solved the problem in the planning space, I can use that to guide my search in the larger space of writing the entire essay. ([View Highlight](https://read.readwise.io/read/01hy6g0na5q73mn1k4fxg2py7s)) - 💭 i didn't see planning from this frame before, as a means of narrowing scope and solving that first Hill-climbing applies this notion to problem solving. Start with a tentative solution to the problem, no matter how lousy, and then make small adjustments in the direction that improves your starting point the most. ([View Highlight](https://read.readwise.io/read/01hy6g24pax3vmsq7bxmpbh541)) Wiles was able to prove Fermat’s Last Theorem, not because he had extensive practice in weak methods, but because he possessed an enormous library of strong methods that drastically reduced the problem space. Yet that same knowledge would probably be of minimal use in helping him fix a car or file his taxes. ([View Highlight](https://read.readwise.io/read/01hy6g5rnnyp84jv06mbjet0cz)) - 💭 weak strategies are inefficient so instead we should just collect a toolkit of niche once we can apply the certain scenarios The best way to discover promising problems is to work with people who are actively pushing at the frontier. The advantage of working at firms, research labs, or in groups of people who are making new contributions is that you can get strong hints as to which aspects of the problem space are ripe for exploration and which are unlikely to immediately yield fruit. ([View Highlight](https://read.readwise.io/read/01hy6gawp28m8eyatwv7qbxwty)) - 💭 competition or collaboration implies plausibility ### New highlights added May 28, 2024 at 12:19 PM if you could somehow suppress means-ends analysis, students would have more cognitive capacity left over to learn from their actions. ([View Highlight](https://read.readwise.io/read/01hz05xa9373x33ycwyt9d5gd4)) - 💭 why solely trying to figure things out on your own is not that effective because it does take a lot of resources A central distinction in cognitive load theory is between intrinsic and extrinsic cognitive load. Intrinsic load refers to the necessary mental effort that accompanies learning. To benefit from a worked example, students need to study it, and this mental engagement requires an unavoidable amount of mental bandwidth. Extrinsic load, in contrast, is all the mental effort that is not directly associated with learning. Means-ends analysis, which requires juggling goals and methods to reach them, is a useful heuristic for solving problems. Yet it may be less useful for learning, because the extra burden it imposes on working memory leaves less space for observing the basic patterns used in solving the problem. ([View Highlight](https://read.readwise.io/read/01hz066e63qyk0wwmr6qne9xsg)) - 💭 introduction to cognitive load theory Studies with eye-tracking software show that students learn more when they can follow the eye movements of experts. It seems we are innately hardwired to follow people’s gazes as a directive about where to pay attention, further reducing our cognitive load when we are presented with a complex scene and aren’t sure what’s important. ([View Highlight](https://read.readwise.io/read/01hz069j6wxmdtf9hbdsccmk6r)) - 💭 does this highlight the importance of a laser pointer when teaching something visual The expertise-reversal effect demonstrates that while problem solving is often ineffective for beginners compared to studying worked examples, this advice flips once students become more advanced. With the patterns of problem solving secured in memory, students benefit more from practice than simply watching. ([View Highlight](https://read.readwise.io/read/01hz06h992gshpm5vxkcpwf33h)) - 💭 continue to put yourself in learning environments where the insights you gain are revelatory and not repetitive. as you understand the fundamentals, then it makes sense for personal application to be more valuable as you can then continue fine tuning it to your own personal needs ### New highlights added May 30, 2024 at 12:36 PM Applications of Cognitive Load Theory ([View Highlight](https://read.readwise.io/read/01hz5ddjj6jzp6dre92zeknaby)) - 💭 seems like strategies are just based on accommodating for limited working memory if I utilizing long-term memory Application #1: Seek Out Worked Examples Whenever you’re facing a new subject of any complexity, look for resources that have a lot of problems with worked-out solutions. In the beginning, these can offer a way to rapidly assimilate the problem-solving patterns. As you progress, you can cover up the answers to use them as practice opportunities. ([View Highlight](https://read.readwise.io/read/01hz5ddp9j6v5sr0qw9w000zr2)) - 💭 this is the imitation part of see, just so there is less solution space potential and less need to maintain during initial working memory Application #2: Reorganize Confusing Materials ([View Highlight](https://read.readwise.io/read/01hz5dfygp8gd7getmb2e794qj)) Application #3: Use the Power of Pretraining ([View Highlight](https://read.readwise.io/read/01hz5dgewa8y5ffbjtp6m00qma)) Application #4: Introduce Complexity Slowly ([View Highlight](https://read.readwise.io/read/01hz5dhav15fq0mqdj79bknxg4)) Video game designers use this brilliantly when they design tutorial levels that have a few of the game’s features, allowing players to use goal-free exploration to learn the mechanics without laborious instruction. As you progress, introduce new complexities in a steady fashion. ([View Highlight](https://read.readwise.io/read/01hz5dj2gevy1cvw02xy0zy1ej)) ### New highlights added May 31, 2024 at 1:52 PM Such a practice may be fine if the skills and knowledge taught are only useful as a sorting mechanism, to separate students by their natural aptitude, with no intrinsic value. But it’s a terrible mechanism if the aim is to teach all students useful skills. Grading on a curve encourages students to be competitive, where one’s gains equal another’s losses. ([View Highlight](https://read.readwise.io/read/01hz820ps6sgmyj66vgnt8xgs9)) - 💭 when should test taking happen then? These analyses show that mastery learning appears to work in elementary, high school, and college classrooms, with particularly large effects among lower-ability students. This is still shy of Bloom’s original goal, to raise student achievement by as much as one-on-one tutoring does. But given the constraints of delivering material to a large student body, mastery learning provides one of the more promising educational interventions that has been studied to date. ([View Highlight](https://read.readwise.io/read/01hz823wh1yvp7sqp6z2y0kx7v)) - 💭 how can ai be used to facilitate this? William Chase and Herbert Simon replicated and extended de Groot’s work on chess in the early 1970s. They confirmed de Groot’s finding that stronger chess players do not seem to rely more heavily on deeper search. Instead, chess masters seem to intuit better moves. ([View Highlight](https://read.readwise.io/read/01hz82wqsbp20vm60qsnegp0t9)) ### New highlights added June 1, 2024 at 10:39 AM A good protocol is to act like a journalist preparing for a story—focus on gathering facts, establishing a timeline, and walking through the decisions step-by-step. This provides the raw material for asking follow-up questions to investigate why the expert made certain choices. A focus on the facts tends to highlight details of a story that may be obscured when simply asking for the broader lessons from the experience. ([View Highlight](https://read.readwise.io/read/01hz8hkzzmht2bta20ddgcem2a)) - 💭 extracting nuance from understanding their rationale during decisions and behaviors, seeing what they found relevant Since knowledge needed to perform in tough problems is often held diffusely, it’s rarely possible to find a singular expert who knows all the answers. Instead, mapping out a list of useful contacts is often the first step to understanding a problem yourself. ([View Highlight](https://read.readwise.io/read/01hz8hsc2h1x401693nehyp38d)) - 💭 knowledge is diffused but with ai you can easily synthesize and aggregate now? From Seeing to Doing Over the last four chapters, we’ve discussed how people solve problems by exploring a problem space, the importance of managing cognitive load when learning new skills, the self-reinforcing cycle of early mastery experiences, and the tacit nature of expertise. ([View Highlight](https://read.readwise.io/read/01hz8hty3hfxbtxyjg2vw2m83y)) - 💭 summary of chapters One way to resolve this tension is to combine the three components of seeing an example, solving a problem, and getting feedback into a practice loop. By repeatedly cycling through the loop, we ensure that all three ingredients of successful learning are available to us. ([View Highlight](https://read.readwise.io/read/01hz8jgbrh4g1ea8kjev4dy6b9)) - 💭 having access to everything gives us a chance to focus on it when needed? but then hoe do you know when to use what? As you progress in a skill, the practice loop can be made more challenging. Seeing examples can fade away as you increasingly tackle problems using your internal reservoir of knowledge. The problems you choose can increase in complexity, as you can manage extra cognitive load from bigger projects. Finally, self-assessment can play an increasing role over external feedback as you develop refined intuitions as to what counts as excellent work. The practice loop creates an opportunity to optimize the level of difficulty. ([View Highlight](https://read.readwise.io/read/01hz8jmshvts3q8x1nak9shnbf)) Whether the activity is brain training, chess, or programming, the results appear to be the same: training improves the tasks directly practiced, but there’s little evidence of substantial improvement to other topics. ([View Highlight](https://read.readwise.io/read/01hz8k434jp6ry2ntnea2d0fkj)) - 💭 i don't believe this anecdotally since there is a small fusion of meta learning and tacit knowledge gained from learning something new. how does this explain how you learn something new by connecting it to something you already know. instead, its based on how you are perceiving the problem Where it misleads is the suggestion that strengthening on one task will lead to general mental fitness in many unrelated tasks. Perhaps a better metaphor is that the mind is a collection of tools, built out of knowledge. Each tool may be specific, but in total they can add up to sophisticated abilities. ([View Highlight](https://read.readwise.io/read/01hz8kw63khk05mdr5aarae0sn)) - 💭 so then to train the muscle you would have to learn ultralearning?