Insights into learning, science, and society. For teachers, students and the science-curious.
I’m a cognitive scientist specializing in learning and instruction. I focus on how to develop expert reasoning skills and what the heck this beautiful, fascinating thing called science is.
You’ll also find videos on social science research methods and bigger picture stuff, like how social systems produce scientific knowledge and why that knowledge sometimes goes ignored or misapplied.
I have a PhD in learning sciences and technology design from Stanford and a law degree from Cornell.
You can read more at www.benjaminkeep.com and sign up to my email newsletter there.
I want you to know that I plan to take study techniques courses to improve my grades and my time studying, learning affectively in the long term, I have reviewed several videos and platforms and found 2 that convinced me because of their quality, these are cajun koi academy that has a course called study quest and justing sung with his icantudy course, the latter is expensive in relation to dollars and is in English, but offers discount scholarships to people with poor economic situations. I tell you that recently I am going to have a new little brother and with this the economic situation of my parents will be affected, I already showed them the courses that I want to take but my parents did not say yes or no but for the moment it is under discussion, besides the fact that we do not have a credit card and my country has a shortage of dollars and I do not know if my parents will accept. I would like you to tell me your opinion on this matter with my current situation, and give me some advice since I am about to finish my first semester in medicine and I will have a month off, but the second semester will be more difficult and that is why I wanted to take these courses.
this may sound unrelated but can you please make a video about what reading books actually does to brain that makes reading so important. I also wonder how reading improves problem solving skills, critical thinking etc.
Wow. Solving thousands of problems in physics without learning physics sounds a lot like the real life analog to overfitting, a concept in machine learning. In machine learning it is common practice to separate the data out into multiple sets. A training, validation and test set. They're trained with the training set, of course, and during that time, their progress is assessed with the validation set. What this does is it prevents the model from essentially memorising the data points which wouldn't be really useful in the real world. But the validation set also isn't perfect because it is used to inform the ML practitioner's decisions on how to train things, and eventually they might accidentally land on a model that miraculously understands the underlying pattern... but only for the training and validation sets. So you have to test that as well. Chances are because of the numerous iterations and tweaks from the practitioner, the models that seem the best are likely to have a positive bonus purely from chance. They go through that much for an extra few percentages of accuracy. Compare this to the education system which holds practice tests as bibles. "Guy's it's simple, when we give them practice tests, they do better!" There were literal companies coming in to give a basic talks on ways to increase your grades at my school. And it boiled down to taking practice tests. That was it... Teaching to the test. Perhaps that's why the education system is so wanting - because they inform themselves based on the tests and only the tests. No question of true performance or understanding. This doesn't seem changeable because on which metrics do we evaluate a better education? It's like a misaligned AGI, the government. Can't tell it nothing except what will get it closer to it's predefined immutable goal. I know it may be somewhat a stretch to say that they're teaching to the test, since the practice tests are actually different data-points, so to speak, to the real tests - but call them training papers and it's the same thing. The environment is the same, so you're not testing conceptual understanding necessarily, so much so as you are testing paper doing. And certainly papers have their patterns.
This is a great use of logic for problem solving. But this does not limit using this concept to hard science. I teach people how to drive a truck, to get a Commercial Driver License. The toughest thing to learn is how to back a semi-truck properly. I tell my students the first thing to do is to imagine where the back wheels will be when they are done making an adjustment in their backing problem. Then how to get them there. For me, that is the point of Dr Keep's video - imagine the solution before solving the problem. (I was completely distracted by the Go (wei-chi) board on the stand!)
It sounds like this procedure is widely applicable even to abstract problems such as marketing or art. Trying to predict before executing, e.g. running an A/B test for marketing or painting for art. It sounds like you'd be creating a habit of gauging first whether something is good abstractly - using those abstract ideas to guide your decision making throughout execution phase. I suppose for marketing you'd focus on ideas like "how much of and which of these demographics are affected and how so?", "what might this person say about this brand for this particular webpage?", or "would this audience stay longer or shorter on the site when they first see this?" and art would be about visualising it, asking questions about different parts like "would this make the face look happier?", "would this make the painting more vibrant?", or "would this contrast poorly with that?" I wouldn't know much about some kind of "learning loop" you'd employ. Mr. Keep, I wonder about your thoughts on working through abstract problems that don't have concrete answers as do the perfect worlds of maths and physics. What do you think about the application of mental models in step 2.5 of prediction? Things like inversion? "What could I do to make the potential customer not want to buy at all?" "What simple small thing could I do to ruin this painting?" Or consideration of the extremes? "If X goes to say a billion what does this look like?" And so on...?
You are like the absolute gold standard channel on learning and its sad you're criminally underated!! Keep making these pls!!! (Anyone reading like the video)
This is the genuinely most fascinating channel I have ever come across! Thank you for sharing your gift. If anybody hasn't signed up for the newsletter, it is similarly a goldmine!
Thanks for the video, really helpful stuff, I hope I'll remember to come back to this when the semester starts haha. On a side note, I wanted to ask your perspective on 2 questions about practical implementation of learning science into real life. I'm sure you're busy, so you can answer with as little or as much detail as you like (or maybe make a video if that seems like a good idea!). 1) Do you have a structure or recommendation for going about implementing all the ideas and advice in your videos into our personal, real life pursuits? Some sort of way to identify and adapt relevant theory to whatever thing we want to learn, whether it's physics, social skills, starcraft, stress management, or even something as simple as building an exercise habit? I guess you could call it meta-meta-learning haha. Context: I find myself watching a lot of your videos and feeling very enlightened and captivated, but looking back I haven't actually put as much of it into practice as I'd like. I think it's because it can be difficult to know where to start, which concepts to apply where, and things like this. 2) What's your opinion on how emotional intelligence plays into learning, and how can you practically improve it to make the actual (and hopefully more effective) practice you end up doing in real life adaptable enough to your preferences and the follies of everyday life to be enjoyable and somewhat sustainable? Context: There's a lot of emotional traps that you can fall into while learning that can make it difficult to actually do practice that's more effective. Making practice plans but then procrastinating them or ending up "practicing" in more unhelpful ways because they're easier; feeling demotivated by failure or overwhelm about not knowing what's going wrong or what to do. Things like these have been the other biggest obstacle in translating intriguing and motivating learning theory into actual improvement in my real life. As I've gotten better with EQ skills in other parts of my life, I've found myself (building practice plans that increase my chances of) following through more, but part of me again wishes I had a more structured way of improving on this, especially with respect to learning specifically. Again, thanks for your videos, the things that I actually have implemented like free recall have been really helpful not just in my own studies, but also in tutoring friends and peers. I'm excited to see the videos coming next, and also your course!
I might be a bit slow here, but I don't get your point at all. More precisely, what is "conceptual prediction"? What am I predicting? In what way? What is an example of a prediction? Some kind of guess what the result will be? How would that help in any way? Or does this specifically only have to do with some intuitive understanding of physical systems in physics?
A conceptual prediction is an educated guess about the outcome of a problem before you start any mathematical calculations. It's based on your understanding of the underlying principles and your intuition about how things work. For example, in a physics problem about a falling object, you might predict that "the speed will increase as time passes." Or in a collision problem, you might predict that "momentum will be conserved."
@@kenny-iv5cq Sure, that is what he said. So it is not really trying to guess the outcome, like "I belive the answer will be 5", but rather thinking about the consequences of the assumptions before doing anything else. The final outcome itself might be hard or almost impossible to guess, but general consequences of the assumptions like "the speed will increase as time passes." is of course useful to think about. I just think it was weirldy formulated if that was what he was after. You don't try to guess what the final solution should be, you think about the consequences of the assumptions as much as you can. If in rare cases, this can help you guess the final solution, then fine.
that works btw. i used that approach to learn physics for my exam knowing nothing about physics. its like formal logic actually in a sense that you get factually correct statements and arrange them so that output of the whole thing makes sense
This is awesome. Very applicable also for coding. In code reviews in my school I involuntary throw out ideas of what code does, when I read it and it sometimes annoys the evaluated, because they want to explain their code. I always ask them, to give me some time to engage with their code, before they explain it. And I make rather good progress.
I always ask my physics students to imagine the scenario they are working on playing out in their mind's eye as the first step. And to imagine the various parameters have sliders attached to them that one can adjust to get varying results. Really just to get them thinking about causal relationships and the parameter space in some way first, since that is what they are gonna be exploring as they solve the problem anyhow.
thank you so much, the hard solving problem where the prediction depending on time it's really hard to find which steps should to combine to get solution
I think this also ties in with another video of yours about routine expertise and adaptive expertise, I would assume this is probably one of the key steps in achieving adaptive expertise? I have a question though if I learnt some theory (f.e. a mathematical concept & formula), exactly how many problems am I meant to solve, or if posed in other words, how do I know if I solved enough problems? Because when I search online about adaptive expertise, I think I found some information about routine expertise being a building block of adaptive expertise, but do I need to be able to use techniques off the top of my head to achieve adaptive expertise? My perspective has been: "understand the technique, and if you come across a problem that ends up needing the technique, figure out the steps manually (because I understood the why behind the steps) or just look it up". I only solve one problem or so to practice, but I am no longer sure if this perspective is enough to achieve adaptive expertise.
Would be interested what Ben replies with, but I know for myself I teach my students to fully model all relevant variables in the problem every time. No matter if the question asks for them or not. That way you rly start seeing how they are all connected, how they are all related to the scenario, and it makes every problem start to look the same. I've had students tell me after we cover a topic that they are getting bored bc every kinematics or dynamics problem looks the same now, which is a good thing! They start classifying problems based on conceptual framework/familiarity instead of memorizing bespoke 'solutions' for specific variables one at a time that come with asterisks anyhow. If differently framed problems start looking the same in the sense they are easy, that's probably a good place to be such that you can seek out more novel problems.
I would also be interested in what he answers. One thing I know for certain is that I'm definitely not a fan of the very common idea to solve as many problems as you can. I feel like most advice around this question (e.g. on reddit) is to just keep solving problems and turning yourself into a pattern recognition machine. While math and all related subjects that rely heavily on procedural skills DO need practice to get good at, I much prefer the idea from Justin Sung, that if you focus on the highest orders of learning (in most cases that's conceptual understanding), the details will often follow on their own. In this case I like to think of procedural skills (partly) as those "details", and I do in fact find that when I understand a concept well, I'll usually be able to solve it or basically be brushing against the solution but maybe missing some final touches on how to actually get to the solution, at which point there are very few gaps to fill and I don't think I need to solve many problems. When you understand things better, they stick with you much better and longer as well.
By "YouSum Live" 00:02:00 Importance of conceptual predictions before executing solutions 00:02:20 Evaluating answers based on conceptual predictions 00:03:36 Significance of comparing conceptual and numerical solutions 00:04:00 Learning from conflicts between conceptual and mathematical models 00:04:41 Improvement through understanding conceptual predictions 00:05:00 Integration of conceptual predictions and mathematical skills By "YouSum Live"
An example of bad reading. George III was not King of England! Benjamin has never heard of the United Kingdom and misinformed his followers. Victoria was born in England, which is part of the United Kingdom, therefore became Queen of the United Kingdom. "England" is not a short form of United Kingdom. Benjamin should read the Declaration of Independence!
I don't know why but this video was extremely engaging and I am not saying this just as a compliment. I am genuinely curious about why this video is so fun to watch? I think you presented scientific research so good that it feels like all the scientists conducted all these studies essentially for this video to be made.
This framework enabled me to clearly COMMUNICATE a solution that seemed obvious to me but wasn't apparent to my teammates. I've learned that in the real-world, collaborating with others to find a solution is far more useful and satisfying, than solving it alone without any practical application.
