- ⁓ build Swimming Ducks
- ⁓ every user-facing learning app needs to be a swimming duck
- ⁓ know whether learners are beginners or experts
- ~ beginner's minds are open, eager and free or preconceived notions
- > It’s easy to lose awareness that we’re talking like an expert.
- ⁓ in learning, avoid tutorial hell
- ~ passive learning is better for beginners, active learning better for experts
- ~ what's effective for a novice is ineffective for an expert and vice versa
- ~ expertise reversal effect can be explained with CLT, ZPD and scaffolding theory
- ~ dp approach﹕ split lesson into a series of ascending steps
- ~ gradually increase the grain size of learning
- ~ facilitate approximate and incomplete runs for complex target skills
- ~ split things in 'black box that just works' and 'detail I'm gonna understand now'
- ~ to do shaping, don't teach final skill, but little parts of it
- ~ discovery learning overstrains the learner's working memory
- ~ experts prefer nuance over compactness
- ~ passive learning by instruction is effective for beginners, but ineffective for advanced learners
- ~ expertise reversal effect can be explained with CLT, ZPD and scaffolding theory
- ~ cognitive apprenticeship works for various fields
- ~ what's effective for a novice is ineffective for an expert and vice versa
- ~ doctors don't improve with experience, even though they try all the time
- ~ learning by exploring is ineffective for beginners, but very effective for advanced learners
- ~ experts will do many things differently from their peers, not all of them relevant
- ~ experts are (too) set in their ways
- ~ self-monitoring requires good mental representations
- ~ use lenticular design to design for both beginners and experts
- ⁓ know whether learners are beginners or experts
- > You press the button, we do the rest
- ⁓ don't expose your inner workings to the user
- ⁓ in apps with complex input patterns, have a simplified mode allowing microlearning on the go (if desired)
- ⁓ a good CMS is the linchpin for learning tools
- ⁓ every user-facing learning app needs to be a swimming duck
- ⁓ optimize knowledge acquisition rate
- > 15 minutes a day can teach you a language
- ~ consistency means higher lows, nothing more
- ~ not item acquisition rate counts, but relevant item acquisition rate
- ~ output leads to deeper language processing than input
- ⁓ prioritize scheduling items that are overdue
- ~ putting something into a (good) SRS is costless
- ⁓ stay in the flow tunnel
- ~ mind not being challenged enough leads to complacency or bias
- ~ core game loop relates to dopamine seeking-reward loop
- ⁓ core game loop is an actual loop you can sketch
- ⁓ define your game loop as an enumerated list
- ~ before designing game loops, imagine the ideal play session
- ~ 90% of skill variation can be predicted by answers about practice planning
- ~ end training session when focus slips
- ⁓ don't forget to define game loop
- ~ good games provide a circle of expertise
- ⁓ core game loop is an actual loop you can sketch
- ~ repetition learning is often boring
- ⁓ sometimes, intentionally show difficult items
- ~ failure rate in digital games goes up to 80%
- ~ use DDA to keep player in the flow state
- ~ to learn to type fast, force exposure speed 10% over comfort
- ~ practice card memorization by using an uncomfortably fast metronome
- ~ plateau after initial speedy learning is normal, and no insurmountable wall
- ~ most people stop practice once a skill is automated and 'good enough'
- ~ there are no limits, only plateaus
- ~ touch-typing is a fascinating example of plateaus in casual practice
- > Generally the solution is not “try harder” but rather “try differently.”
- ~ for everything worth learning, there is a massive slow-down in the middle, where you crawl
- ~ reaction time for a task goes down exponentially at first, then slower
- ~ having hundreds of easily collectible rewards is an underutilized mechanic
- ~ ebb and flow create meta-flow
- ~ addicting games promise permanently that something big could happen
- ~ players stop the game when too frustrated or bored
- ~ players may not only get bored, but also overwhelmed
- ⁓ there is an initial phase where the flow tunnel is hard to hit because of lack of data
- ~ separate simulation room & debriefing room
- ~ experts know the basics of every adjacent field
- ~ learning is boosted by applying new techniques (even if not needed)
- ~ start with identifying ZPD precisely
- ~ solve the same problem twice
- ~ learning-by-doing is effective but not time-efficient
- ~ to improve, go beyond business-as-usual
- ~ push zone is less big than usually assumed
- ~ first, do the thing you're trying to do a couple of times to find your actual bottlenecks
- ~ if learners are only extrinsically motivated, ask questions until you find any intrinsic motivation
- ~ design a puzzle by beginning with the end state, then add obstacles
- ~ specify both what needs to be learned and when
- > Always start from the problem, not the solution
- ~ students should get guided tasks that are slightly harder than what they can accomplish alone
- ~ you need to know the actual knowledge gaps before designing a learning solution
- ⁓ if you are designing a learning experience, first define what excellence looks like
- > Serve businesses, not people. People will do anything to not pay.
- ~ form language islands by drilling topic-specific sentences of your own creation
- ~ sentence production is not significantly better than just memorizing words
- ~ sentence learning connects words, grammar and stories
- ~ when learning vocab, form sentences
- ~ grammar rule knowledge doesn't help with speaking, but with writing
- ~ use cloze sentences for grammar instead of memorizing rules
- ~ test learners ability to correctly apply (or not apply) a given formulaic expression
- ~ learn how to make complex sentences early on
- ⁓ to practice grammar, extract all interesting grammar concepts from a natural sentence
- ~ make multiple sentence cards for the same word
- ~ mnemonic stories for vocab aid recall even without remembering the whole story
- ~ sentence learning connects words, grammar and stories
- ~ sentence production is not significantly better than just memorizing words
- ~ superior memory is explained by spatial learning
- ⁓ my brain embodies my bike's speed and brake capabilities and adapts my world model
- ~ humans are very good at learning spaces
- ~ VR can enhance spatial ability
- ~ VR allows spatial thinking, which is human default
- ⁓ I remember visual ideas like Ollie the Octopus so well
- ~ spatial learning can enhance vocabulary acquisition
- ~ examples of user empowerment﹕ command line, spread sheet, the internet
- > technology must never be accepted as part of the natural order of things
- ~ material can be learned during testing
- > language is simply a means of telling stories
- ~ language acquisition only happens when input is attended
- ~ self-monitoring is required for expertise
- ~ language is only acquired if it's consciously noticed
- ~ ask learners to extract the "rule" after exercises
- ~ when learning a complex, intertwined skill, boost learning by rotating which part you pay attention to per run
- ~ sensemaking﹕ people notice outcomes, then reconstruct what must have happened
- ~ focussing on meaning and form at the same time is impossible
- ~ telling your body what to do doesn't work
- ~ paying special attention to open-class words indicates to better lang learning
- ~ learn by focusing attention on one component of the skill at a time
- ~ good stories are simple, unexpected, concrete, credentialed, emotional
- ~ speakers adapt their language to negotiate meaning together
- ~ making your own dictionary enables metacognition, input negotiation and exposure in varied contexts
- ~ simple speech acts allow vast inferences thanks to context
- ⁓ you can practice language by extracting speech acts from text
- ~ willingness to communicate in target language requires base level of confidence
- > I speak Spanish to God, Italian to women, French to men, and German to my horse.
- ~ making input more comprehensible by negotiation does not lead to better acquisition
- > Create your own jargon
- ⁓ consider Kommunikationssituationskontext
- ⁓ point of language learning isn't talking, it's about asking interesting questions and understanding the answer
- ~ ask﹕ "in how many languages do you live﹖"
- ~ goal of lang learning is literacy
- ~ immigrants use the new lang in 3 important contexts﹕ job hunting, standard social situations, admin and nav
- ~ write a diary in your target language
- ~ stories in games can be generated by dev, or by player
- > it is dialog that constructs linguistic knowledge
- ~ learn a language faster by not speaking your native language
- ~ learn a language to communicate with native speakers
- ~ language acquisition only happens when input is attended
- > When you’re ill or injured, the world becomes one of can’ts
- > The UNIX command line is the quintessential end-user programming environment.
- > At this level of the development of self-regulation, players begin to develop a mental model of the game that allows them to move from the concrete experiences of the serious game to more abstract notions of the game
- > academia is struggling at doing science nowadays also because you’re stuck doing all kinds of other things with your time rather than actual research
- > My work on this was an extremely long-drawn-out affair, and it was only after some twenty years of it that I reached some degree of understanding of my fantasies
- > Learn three new things before you come back to us
- > The blade itself incites to deeds of violence
- ~ before starting a project, define commander’s intent
- ~ you can't use reels to teach complex content
- ~ humans may make mistakes by acting upon an item that is somehow similar to the actual target item
- > My working habits are simple - long periods of thinking, short periods of writing.
- ⁓ high acuity, low occurrence are hard to learn naturally, yet important
- ~ don't demand global solutions to local problems
- ~ a given learning need may be a skill gap vs. a knowledge gap
- ~ procedural learning works differently than declarative learning
- ~ prioritize doing over knowing
- ~ get declarative knowledge out of the simulation
- ⁓ mnemonics are useless associations, but at least you build fundamental fact knowledge
- ~ complex behaviors often involve cognitive and motor skills
- ~ sophia is not phronesis
- ~ procedural and declarative knowledge is useless without perceptual competence
- ~ small volume of practice exercises may work for declarative or procedural, but not perceptual
- ~ expertise cannot be trained in the classroom
- ⁓ understanding and memorizing are not the same thing
- ~ SR may lead to overfitting, because you learn only this answer to this question
- ~ knowledge acquisition isn't the end goal, but a requirement for skill acqusition
- ~ strong declarative knowledge may not influence the procedural, and vice-versa
- ~ memory for things is different from memory for words
- ~ perceptual learning is neither declarative nor procedural
- ⁓ ideal initial interval for practicing a chord switch is definitely not 1 day
- ~ proceduralization is transforming declarative knowledge into procedural knowledge
- ~ language learning happens when learner realize a gap in their knowledge
- ~ procedural learning works differently than declarative learning
- ~ measure skill along dimensions of (in)competent x (un)conscious
- ~ seize the moment when something pops in your head that you can't express in target lang
- ~ language production demonstrates gaps
- ~ language production stands and falls with whether or not you get your point across
- ~ ask﹕ what is the user doing right before the intended habit
- ~ meta cognition like "I have a gap here" breeds motivation
- ~ language production demonstrates gaps
- ⁓ think in trials
- ⁓ tracking stuff like 'trailing success rate' or 'streak correct' doesn't make sense if SR is counteracting
- ~ keep real skill and virtual skill apart
- ⁓ rate SR by asking user for their confidence
- ~ unconscious incompetence is one of the biggest problems in learning (motivation)
- ~ area9 approach﹕ exercise eval + "bin sicher" "denke, ich weiß", "unsicher", "keine Ahnung"
- ~ seize the moment when something pops in your head that you can't express in target lang
- ~ apps cannot replace complex meatspace interactions
- ~ learn a language with another language so you don't translate to your native lang
- ⁓ don't trust that my learning experience is the same as that of others
- ~ you cannot judge how a (video game) experience will be by imagining it
- ~ complex simulations are a black box of the designer's world view
- ~ be aware that a learner may have this or that goal, and that he may change it
- ⁓ language learning tool design is subject to paradigm bias
- ~ the best teachers adapt their style to their students
- ⁓ beware believing in the Omniscent Designer
- ~ software creators overlook how bad their software is
- ~ find what works, not what's popular
- ⁓ other people seem to be way more consistent with duolingo than me
- ~ people want duolingo but more topic freedom
- > When you make assumptions about your users, you run the risk of being wrong
- ~ if you can design one minute of fun gameplay you can stretch it indefinitely
- ~ in games, always give meaningful choices
- ~ in game dev, give multiple puzzles at once
- ~ in gameplay, a change is as good as a rest
- ~ avoid degenerate strategies
- ~ make a game replayable by adding meaningful, dilemmatic choices with unpredictable outcomes
- ~ if the player is about to reach a goal, setup another one
- ⁓ ensure variability in item scheduling
- ⁓ to make games come alive, introduce instability
- ~ layer different reinforcement schedules
- ~ you should know the dominant strat in your game as gamedev
- ~ mechanics are made satisfying by anticipation
- ~ emotions are triggered by change
- ~ use twitch decisions as an easy way to generate flow in games
- ~ experts do not decide between options, they see only one
- ⁓ single-player video game mechanics can ultimately only provide limited variability
- ~ follow the fun
- ⁓ in games, allow for (occasional) high-risk, high-reward
- ⁓ games can be an endless loop or story-based
- ~ games can be finite or infinite
- ~ to get your core gameplay, remove until the game is trivial and uninteresting, then reverse the last cut
- ~ in games, always give meaningful choices
- > In academia there is no difference between academia and the real world; in the real world there is.
- > If you think Psychological Science was bad in the 2010-2015 era, you can’t imagine how bad it was back in 1999
- ~ psychology is mostly good at predicting one-shot situations
- ~ control groups are hard to do for longitudinal studies
- > Science is neat only in theory, rarely in practice
- ~ in user testing, the unpredictability of real life is desirable
- ~ brain research is not sufficiently developed to guide learning experience creation
- ~ if learning things that children also learn, start where they start
- ~ pretotype by minimizing time to date, dollars to data and distance to data
- ~ books don't work
- ⁓ if books are just superstructure slapped onto content, that explains why they work semi-well
- ⁓ the retained value of books for me is based almost exclusively on singular, out-of-context quotes
- > I have read so many books . . . And yet, like most autodidacts, I am never quite sure of what I have gained from them.
- ~ written words can only remind you of what you already know
- > Make sure the early questions in a mnemonic essay are trivial﹕ it helps many users realize they aren’t paying enough attention as they read
- ~ you learn by pressing every button
- ~ learn what's relevant
- > Just-in-case learning sucks compared to just-in-time learning.
- ⁓ you can't google what you don't know
- ~ relevancy and context are the core of a learning tool
- > Learning without doing is wasted
- ~ there must be a product
- ~ make your growing knowledge something concrete
- ~ use self-directed writing to learn the parts of the second language you are truly interested in
- > Qui docet discit
- > When I don't have skin in the game, I am usually dumb
- ~ don't ask 'how to take notes', ask 'how to think better'
- ~ successful learning media places the learner in an active, creating role
- ~ learning should ideally be embedded instead of isolated
- ~ do something, get stuck, then learn something
- > Successful digital games involve players in a way that leaves them believing they have something emotionally and personally at stake
- > Just-in-case learning sucks compared to just-in-time learning.
- ~ friction slows you down, leading to better encoding
- ~ sometimes, difficulty is desirable
- ⁓ you want a pressure cooker
- ~ to learn something, put it into an obsession-worthy mission
- > intense practice of very simple activities leads to very surprising and dramatic rewards
- ~ the perfect practice is the one with the most mistakes per minute
- ~ computer can be used as tool or as tutor
- ⁓ you want a failure machine
- > Experiment with doing it clearly wrong
- ~ promise learners that they won't feel stupid
- > I missed more than 9000 shots in my career
- ⁓ simulations are a pretty good failure machine, but not a perfect one
- > failure has been achieved, thank God
- ~ make intentional mistakes
- > top skaters fall more often during their training sessions
- > Don’t raise the pressure, lower the wall.
- ~ never practice struggle
- ~ a language app must be more useful than reading a book
- ~ to make progress, increase intensity
- ~ there is no speed limit
- ~ successful language learners do not depend on a specific method
- > Your first 10,000 photographs are your worst
- ~ you want noiseless feedback
- > simplify while learning
- ~ redundancy enhances learning
- ~ the more you process an item, the better you remember it
- ~ to learn, break homeostasis
- ~ beware minor inconveniences
- ~ DP happens outside the comfort zone
- ~ less helpful software may generate experts faster
- ~ you can't think without writing
- ~ interleaved practice is desirable difficulty
- ~ faster acquisition means poorer retention
- ~ visualizing stuff as a drawing can make understanding trivial
- ~ spaced practice is desirable difficulty
- ~ problem-based learning is based on ill-structured problems
- > We don't rise to the level of our expectations, we fall to the level of our training
- ~ transfer being rare and difficult is key to learning as a whole
- > Grand masters literally see a different board.
- ~ experts use intuition, not deliberation
- ~ receptive skills are easier to learn than active skills
- ~ amateur representations are concrete, expert abstractions are abstract
- ~ mental simulations improve skills
- ⁓ mental currency conversion is just like chess﹕ if you're good at it, you just see
- ~ amateurs don't understand just how much better expert's mental representations are
- ~ mental models are tacit knowledge
- ~ upon entering a situation, experts create a list of expectancies, plausible goals, a list of cues, and an action script
- ~ experts in any given field have insane field-related memory
- ~ better musicians have more detailed representations of music
- ~ automatic pick-up of patterns is characteristic for expertise
- ⁓ expertise level of geoguessr pros is incredible
- ~ since chunking is individual, you can't abstract the relation between number string length and remembering
- ~ systems are made of layers moving at different speeds
- ~ chunk by literally cutting the lesson into pieces
- > analogy is how we think
- ~ use Free Recall with a Paper to go beyond atomic recall
- ~ repetition does not work for complex learning materials
- ~ essay writing is retrieval practice
- ~ retrieval should include synthesis, reorganization, comparison, application and context variation
- ~ spacing effect in cued- and free recall may work differently
- ~ learning is structuring of information
- ~ to learn from lectures, compress and recompress
- ⁓ learning an alphabet demonstrates that atomic learning is not always desirable
- ~ retrieval practice can be more than just flashcards
- ~ simple flashcard-based retrieval practice is one of the less effective recall practices
- ⁓ SR usually ignores when something needs to be memorized
- ~ native speakers use way more pragmatic markers than L2 learners
- ~ don't bother with a curriculum, just simulate the real thing closely without serious consequences
- > So as developers of computer games, we should design our games to be as close to the actual end performance that we seek to develop
- ~ if you want to pass a law school exam, practice taking law school exams
- ~ from day one, speak your target lang
- > Lesen lernt man nur durch Lesen, Sprechen durch Sprechen, und Flüssigkeit kann in einer Sprache nur erreicht werden, wenn man sie auch trainiert.
- ⁓ Arabic Number Practice has no prompt to actually remember numbers from memory; that's a problem
- ~ avoid tell and test
- ~ if you want to pass a law school exam, practice taking law school exams
- ⁓ serious games are usually not letting the player 'do the thing'
- ~ doctors don't have more time, so convert normal work into learning experience
- > If you wish to develop a skill or impart knowledge, create a game that closely matches the end state you wish to achieve
- ~ the tape doesn't lie
- ⁓ curriculum is useful
- > Do work that fully reflects the final work you wish to produce.
- ~ VR reduces difficulty of transfer
- > Let reality be the teacher
- ~ to truly understand something, you must observe it in situ
- ~ do the real thing
- > So as developers of computer games, we should design our games to be as close to the actual end performance that we seek to develop
- ⁓ the input hardware of old Arcades is very interesting & inspiring
- ~ experts set goals before training
- > What we see when we look at a page depends on what we have in mind, and it’s usually just a fraction of what’s there.
- ~ lifeguards need to be rotated every 15 minutes
- ~ TBLT is agon
- ⁓ when building learning tools, consciously decide how much delay feedback should have
- ⁓ consciously balance how much information to give to the player
- ~ when teaching, give information over compliments
- ~ penalties are discouraging even if meaningless
- ~ without a teacher, learning is much less efficient
- ~ never praise perfection
- ~ give praise only when deserved, not always
- ~ attribute (failure) to technique so you know you can change it
- ~ in learning, praise and criticism are overrated
- ~ remediation = Sanierung = fixing of common learning bugs
- ~ give feedback immediately
- ~ far transfer is rare
- > John Carmack ain't no React.js expert
- ⁓ DP may not allow learning transfer
- ~ measuring (far) transfer is extremely hard
- ~ playing Civ does not improve history knowledge but players understand history narratives faster and better
- ~ transfer from serious games to real world is limited as best
- ~ you can't train memory, you can train your memory for a specific category of entities
- ~ metacognitive engagement with a game is necessary precondition for transfer
- ~ to measure learning outcome, use LTEM
- > Excelling at chess has long been considered a symbol of more general intelligence. That is an incorrect assumption in my view
- > The likelihood of positive transfer is a function of the similarity between the trained task and the transfer task
- ~ learning machines have low retention and transfer
- ~ common elements theory explains near transfer
- ~ failure of learning transfer is often due to surface dissimilarity
- ~ show the goal structure
- ~ give a compelling reason to learn what you're teaching before you teach
- ~ long-term commitment massively boosts learning success
- ~ 10k hours is just the amount of practice your competition has put in
- ~ self-instruction means lack of feedback, lack of interaction with others, and perseverance problems
- ~ consistent practice beats sporadic bursts
- ~ motivate people by moving the goal away from "pass the exam"
- ⁓ ask learner for reason to learn language in lang app
- ~ first, commit 20 hours
- ⁓ know whether your app should affect short-term or long-term memory
- ~ long-term commitment massively boosts learning success
- ~ insist to only design learning experience when you know what is to be done with the knowledge
- ~ learning for mastery means asking "what meaningful thing can I do with what I learned"
- ~ memorization is needed for knowledge that needs to be immediately available, and to build advanced knowledge
- ~ know why you want to learn
- ⁓ measure return on learning
- ~ ask what would happen if your software works perfectly
- ~ companies want hard numbers about the benefit of learning initiatives
- ~ optimize for mastery, not certificates
- ~ superior expertise may mean effectiveness, efficiency or outcome quality
- > Tools for thought must be developed while doing serious work
- ~ plans to solve a problem can be rendered as a dependency graph
- ~ first, write down a clear problem definition
- ~ to learn something new, look at the whole thing first
- ~ give a compelling reason to learn what you're teaching before you teach
- ~ there is no single theory explaining SLA
- ⁓ a language learning app must be enjoyable to use for 1000+ hours
- ~ to build expertise, you need enthusiasm, not skill
- > the best techniques don’t matter if you’re not using the language enough
- ~ 20-40 hours spread over a month are enough for amazing results
- ⁓ you can aim for time investment or outcome
- ~ beginners try to minimize study time, masters try to maximize study time
- ⁓ to learn a language, commit to get 1000 hours of input within a time span
- ~ best design solves a problem and is cool while doing it
- ~ language learning is measured in minutes, not months
- ⁓ 1 hour of actual practice can feel very long
- > after two replies from the bot I was like "booring"
- > If it requires great energy of mind to create a system, it requires even greater not to become the slave of the creation
- ~ the Palm Pilot succeeded because it did exactly four things well
- ~ IQ is an initial advantage in skill acquisition which disappears quickly and completely
- ⁓ decide whether to build a narrow or broad learning tool
- ~ use a perceptual exposure playlist to improve skill without active practice
- ⁓ you can avoid interference or attack interference
- ~ games introduce voluntary obstacles
- ⁓ fight cue interference by varying how learning items look
- ~ elaborative interrogation helps with learning confusing or complex fact relationships
- ~ cognacy helps with learning
- ⁓ avoid interference in language learning by not learning opposites or thematic groups at the same time
- ~ structuring new learning into old learning minimizes interference
- ⁓ force interfering learning items into context to help encoding
- ~ when you are not thinking 'what button do I press' anymore, you've become fluent in a game's non-diegetic language
- ~ create a variety of memory traces
- ~ memorize poetry by painting and memorizing the emotions of every chunk
- ⁓ ensure encoding is done in such a way that the knowledge can be transferred
- ~ start at transfer effects, then design learning intervention backwards
- ~ depth of information processing predicts long-term storage
- ~ schema induction implies teaching isn't about 'covering content' but engineering schema building
- ~ a good encoding method expands the number of applicable situations where the memory can be triggered
- ~ distinctiveness of items on initial presentation boosts retention, even when distinction is silly
- ⁓ with flashcards, I feel like I don't remember anything until it's suddenly easy
- ~ find one venue to master
- ~ at the beginning, get any listening input, you're not understanding anything anyways
- ⁓ simply repeating a target language sentence out loud may be good beginner practice
- ⁓ to automatically learn important language parts, pick random target lang content, then ask﹕ what would you need to understand this﹖
- ⁓ acquire a language by taking in (barely) comprehensible input
- ⁓ use scratching for context- and level-independent, general language learning
- ⁓ superbeginner videos are possibly the ideal CI
- ~ use audiolingual language teaching to build pronunciation and listening skills in large class rooms
- ~ consume your target language casually while doing another activity like running or jigsaw puzzling
- ~ master one piece of content
- ~ use crosstalk for language exchange with beginner speakers
- ~ we acquire a language by understanding messages
- ⁓ use walking audio method to gain a first feel of listening to a language
- ⁓ just reading target lang out loud and recording is probably good practice
- ~ learn a language by immersing yourself in content
- ~ beginner lang learners won't understand random input any more than a 2 year old understands you talking about inflation
- ~ repeat consuming the same target language material over and over, without strict schedule, until saturated
- ~ at the beginning, get any listening input, you're not understanding anything anyways
- ~ in SR, review within 24h is essential
- ~ use extreme learners to understand normal learning
- ~ use call-and-response as a transition activity in the classroom
- ~ VR memory palaces work better than desktop palaces
- ⁓ most educational games suck
- ⁓ think of item scheduling systems as generating a queue
- ~ use Task-Based Learning to teach communication skills
- ~ there are cultures where everyone is expected to sing, is taught to sing, and can sing
- ~ introduce speaking exercises after about 120 hours of TPR instruction
- ~ adults (and classrooms in general) are usually not allowed a silent period
- ~ when students start speaking under TPR instruction varies wildy
- ~ students are ready to speak after 10h of TPR
- ~ in the beginning of lang learning, learners are reluctant to produce language
- ~ methods of cheating silent period﹕ memorized expressions, and L1 grammar
- ~ in the beginning of lang learning, learners are reluctant to produce language
- ~ students are ready to speak after 10h of TPR
- ~ eventual target of TPR should be around 500 commands
- ~ can't have a dinner part without guests
- > Dude, suckin’ at something is the first step towards being sorta good at something
- ~ learn through pain and adversity
- ~ improving technique beats trying harder
- ~ there is a threshold of needed word knowledge for incidental vocab learning
- ~ checkpoints like towers or bases give structure to open world games
- ~ learning variables are harder to control with complex learning material, therefore most researchers use simplified, itemized material
- ~ value of even massive online learning datasets is limited because you're missing the why
- ~ measure learning progress as savings on relearning
- ⁓ vocabulary learning studies are mostly akin to nonsense syllables; no deep encoding
- ~ practicing recognition on the computer is easy, but practicing recall is hard
- ~ savings by learning initiatives are easily measurable in $, employee growth only with more complex analysis
- ~ for companies, savings & growth are the most important learning outcomes
- ~ both procedural and declarative knowledge sit on a huge complexity spectrum
- ~ represent student competence as a set of production rules
- ~ emergent narrative is helped by randomness, AI, prodedural, agency and simulation
- ~ make practice less predictable
- ~ to increase replay value, raise the skill ceiling
- ⁓ interesting emerging actions are key to good games
- ~ emergent story is generated by the interaction during play
- ~ get emergence by having an infinite game space
- ~ know the market value curves of your game (or app)
- ~ we know more than we are able to explain
- ~ ZPD spectrum is a learning model that considers motivation
- ~ use a Skillometer to visualize learner progress
- ~ Western thought considers abstractness more advanced that concreteness
- > Constantly think of applications
- ~ using vocab in real contexts is highly effective
- ~ sharing what you learned means you apply your knowledge
- ⁓ learn sentence templates by repeating 3 variations of sentence template, then produce a variation
- ~ practical experience leads to reflection and generalization
- ~ most important thing when studying a foreign language is interactivity
- ~ learn a language by using it in context of meaningful goals
- ~ everytime you learn a new piece of language, play around with it
- > Constantly think of applications
- ~ stealth learning doesn't work
- ⁓ in games, start with a single resource
- > It tells you how fast you're going, how much fuel is left, and how much you are late. That's all you need to know.
- ~ attempt to invent the solution to a problem from scratch
- ~ personal interest is intrinsic motivation to understand a topic
- ~ minimal pairs utilize brain's capability of picking up complex pronunciation rules
- ~ trying to use many learning apps increases inhibition to start
- ⁓ people remember incredibly obscure details about LoL and CS﹕GO
- ~ benefits of testing over looking-at-it are not short-term visible
- > its just my low-energy learning tool
- ~ random language content can be quite boring
- ~ formulaic expressions have variable and fixed parts
- > Remember that there is no code faster than no code.
- ~ model psychological effects as stimulus and response
- ~ humans construct meaning by interpreting their experiences
- ~ think about how the game space looks
- ⁓ engine strongly influences what kind of game you make
- ~ we don't actually want a forgetting index of less than 3%
- ⁓ the very first SR repetition being judged 'correct' is likely special; means that the user already knows this
- > Education is not the filling of a pail, but the lighting of a fire.
- ~ situational interest is spontaneous, individual interest is stable
- ~ learn the 2000 most common words, then specialize into subject area
- ⁓ if you want to design interesting learning experiences, think about individual interest vs. situational interest
- ~ motivation for self-directed lang learning quickly decays
- ~ when thinking about learner motivation, distinguish individual interest vs. situational interest
- ~ self-efficacy in a skill is persistent
- ~ use vicarious mastery as a motivator
- ~ situational interest is spontaneous, individual interest is stable
- ~ if you have to google it then its going to use 1 of your 4 working memory slots
- ~ you must master the prerequisites before moving on
- > Effect sizes find mastery learning tends to improve student achievement between one-half and one standard deviation.
- > Gentlemen, this is a football
- ~ mastering the fundamentals feels like running on the spot
- ~ first, learn the incredibly basic, fundamental rules of a topic
- ~ conscious skill acquisition is only possible if base knowledge exists
- ~ master the rules before breaking them
- ⁓ find the judo moves in your learning area
- > You have to put in the hours before you can see the shortcuts
- ~ promote an abstract understanding of the problem-solving knowledge
- ~ learn the foundational fields
- ~ legitimate peripheral participation helps mastering the fundamentals
- ~ since skills are built from sub-skills, improvements can propagate
- > If you can’t ride two horses at once you shouldn’t be in the circus.
- ⁓ if you repeat a learning item in multiple choice rapidly, user can correctly guess answer just by how often they see it
- ~ learning game dev dilemma﹕ allowing for agency, but compromising that agency for learning
- ⁓ to really learn something like "what's what in Subsahara Africa", you have to do all of those countries, in a row
- > YouTube doesn’t care if the video you’re watching is funny or not.
- ⁓ don't memorize without understanding' implies SR should have a 'Don't Understand' button
- ⁓ note whether trials are passive review or active recall
- ~ customization is accessibility
- ~ fiction is even more interesting than learning useful things
- ~ give meaningful error messages
- ~ systems can be modeled by their constraints
- ~ working demo to demo is the key for successful product building
- ~ don't memorize without understanding
- > Surprise is triggered when our schemas fail
- ~ limiting factor of loci building is image-making, which will get faster
- ⁓ instead of continous SR, overlearn, then have "refreshers"
- ~ understanding of words influences understanding of other words, but within limits
- ~ do experimental and playful reps in something you love
- ⁓ uke learning is about chunking
- ~ aim of learning may be increased robustness, resilience or adaptivity
- ~ you can learn the rules of something by mentally constructing a schema from examples
- ⁓ things that evolve why you are not looking are very satisfying
- ⁓ interdependent flashcards open possibilities for leech curb
- ~ use Suggestopedia in small groups, for a stress-free language learning environment
- ~ use learning curves to model how much learning is needed for expertise
- ⁓ production user interfaces can go arbitrarily horrible
- ~ Korean Go masters have an avg. IQ of 93
- ~ goal of learning﹕ be able to convo with a domain expert, just
- ~ to learn, explicitly define feedback mechanisms
- ~ 'use Excel' and 'do nothing' are also your competition
- ⁓ aim for multi-sensory reinforcement
- ~ spend time with people who are great at the thing you want to be great at
- ~ cloze deletion is low-cost alternative to content simplification
- ⁓ problem from teacher perspective when students practice on random websites﹕ no checks, no data
- > ask﹕ what demo can I do next
- ~ creating something from nothing is hard, modifying something is easy
- ⁓ SR does not distinguish between a fatigue lapse & a genuine learning bug
- ~ you can reward the player to do a volume of activity within a timeframe
- ~ video games should inspire learning experiences, because they are optimized by Darwinian market forces
- ~ innovation is based on understanding the problem space better than anyone else
- ⁓ many researchers built algos for custom learning paths based on big learning data
- ~ make learning goals so granular that you can't split them further
- ⁓ to measure complex learning, define list of relevant terms and how they are supposed to be used
- ⁓ watch for inappropriate quantification
- > Rats can't be sped up by any amount
- ⁓ 2 weeks no Tailwind, and skills are gone
- ~ short game loop is about engagement, long game loop is about retention
- ~ when designing coop games, consider giving predefined roles
- ~ building expertise is slower than desired
- ~ consider world narrative
- ⁓ tones exist in every language
- > complexity very, very bad
- ~ some words have their own meaning, some words get their meaning from other words
- ~ there is a dearth of open ITS
- ~ there is no consensus on how to actually best teach languages despite lots of research
- ~ VR in lang learning boosts all kinds of lang learning params, from motivation to reduced anxiety
- ~ definition of learning﹕ same condition, different behavior
- ~ embodied cognition can boost SLA
- ⁓ binary choice scoring in learning apps has implications for both items
- ~ embodied extrinsic learning game boosts alphabet learning
- ~ every existing memory has been encoded somehow at some point
- ~ when changing values in game dev, double or half them
- ⁓ "self-made flashcard" may be almost meaningless, e.g. when practicing letters of an alphabet
- ~ multiplayer works by unevenly distributing resources to players
- ~ trying to make input comprehensible by drilling all rare words is silly
- ~ DDL research is trickling down to practitioners, at best
- ⁓ Chomsky-flavored linguistics seem to assume languagues live in a platonic space
- ⁓ interlanguage hypothesis shows that applied linguistic researchers are obsessed with the idea that lang learning is somehow special
- ~ input isn't intake
- ⁓ use knot systems as memory aid
- ~ teach verb conjugation with loan word verbs
- ⁓ can't build good learning algos without data, can't have learning data without an algo
- > You don’t know a language, you live it. You don’t learn a language, you get used to it.
- ~ monetize by asking users for money
- ~ use Communicative Language Teaching when the main goal is communicative competenc
- ~ learning can be formal or informal
- ~ theoretische Führerscheinprüfung hat 40-50% Durchfallquote
- ~ physical work environments are full of peripheral cues
- ~ TBNA produced lots of pragmatic datasets
- ~ use Bayesian Knowledge Tracing to estimate how likely a skill is learned
- > it is possible to increase effective human memory by an order of magnitude
- ~ scaffolding can be used to stay within ZPD
- ~ scaffolding can be done by reducing DoF or modelling execution
- ~ ask﹕how would 'getting it exactly right' look﹖
- ⁓ early language self-learning is missing delightful successes
- ~ 3 subfields of skill modelling﹕IRT, KST, BKT
- ~ perceptual learning works for tactile stimuli
- ~ magazines and children's books with pictures are best source of talking material for lang exchange
- > Not every task is a chore
- ~ learners may not use any strategies to learn vocab
- ~ making people who muddle through feel smart is a killer feature