Ben Whately: Memrise, memory and learning
Ben Whately, COO of Memrise comes by the podcast to talk about Memrise, including the science behind and future initiatives such as premium memberships!
June 13, 2014 · 27 min
EP 15: Ben Whately: Memrise, memory and learning
0:00 / 26:52
Guest
Ben Whately
Related Episodes
See all →
On this episode of the actual fluency podcast, I'm joined by Ben Watley, Chief operating officer of Memrise. He and I talk all things Memrise, including the science behind their systems, and also what the future holds for the Memrise website.
Welcome to the actual fluency podcast, the podcast that gives you the tools and inspiration to learn languages faster and more efficiently. And now your host, Kris Broholm. All right, today I'm extremely happy to be joined by Mr. Ben Watley, Chief Operating Officer of Memrise, who apart from managing the daily operations in London, also speaks Mandarin Chinese. Ben, thank you so much for taking the time out of your busy schedule to be on the podcast. How are you doing today?
Very good, very good. pleasure to chat with you.
Fantastic. Before we get into the, the, the core of the episode, that is the, uh, talking about Memrise, uh, can you give the listeners a brief introduction to yourself? Like how did you come to stay in China and learn Chinese? And what inspired you to kind of move into the teaching facility or improving language learning in general?
Yeah, so, so I, I studied experimental psychology at Oxford, and one of the, um, and my main kind of area of interest there was memory and doing a lot of computational modeling of memory systems. Ooh. And, um, not very good computational models, I hasten to add, as computational models tend not to be. Um, in my opinion.
But one of the things that interested me most there was, was language acquisition. And the kind of core scientific finding, which I could, um, very tritely sum up as being that human brains are really, really good at learning languages. Yes. And I found that particularly interesting, because my experience had been that my particular human brain was actually quite rubbish at learning languages. And I, I kind of liked this, this tension between the, the, um, scientific fact that my brain was good at learning languages, and, uh, the 20 years of experience that I'd had, well, not quite 20 at that point, but, uh, suddenly 10 years of school experience that I'd had, which led me to believe my brain was rubbish at it. Right.
So, I decided, um, on leaving university, rather than getting a job, or any of those slightly painful ideas. Um, I went away to be, continue being a student in rural China. Um, and I wanted to see if I could learn Chinese, because I'd been so rubbish at learning languages, that I thought, I might as well pick a hard one and see if now, with my increased insight into the way my brain worked, maybe I'll be able to do a bit better. And it was it was a pretty, pretty harsh experience. It was hard. I found it quite hard. Um, if I'd been able to speak enough Chinese to buy a plane ticket home, I probably would have done. I sure would have. But it, it, it, it, it, it, it definitely took me some time to get going. None of my teachers spoke any English, so it was, it was, it was very much a, a teach-yourself kind of a situation. Right.
Um, one of the things that I did while, while being there, I had a lot of time on my hands, as you can imagine, in, in, in rural China. Um, there were, there were not many, uh, English-speaking people around, so I really had very little choice but to learn Chinese. Sure. Um, and one of the things I did was make a, start building a, a big kind of library of visual mnemonics, uh, for Chinese characters. And my idea was that by getting really strong images for each of the sort of building block characters, which are called the radicals, and there are about 200 of them, which can be recombined to make all of the other characters. So although you've got, like, depending on how arcane you want to go, between 10,000 and 80,000 Chinese characters. Um, the, they're all made up of these 200 core building blocks. So if you, if you learn all those, my theory was, then, um, with a sort of simple mnemonic story, you should be able to put together all of the more complex characters. Which, uh, which is more or less true. Um, so that was how I taught myself to, to read, and it, and it kind of worked pretty well. I, um, although I wasn't great at the drawings themselves, as we, as we admitted, my art skills were relatively weak. Um, but they were, they were good enough to, to do the job.
So then, this is rather a rambling answer to, um, what was a perfectly innocent question. But the, so, I then came back to London, um, and got a job temporarily. Um, and did that for a couple of years. But all the time, I, I, I was running an investment fund, investing in Chinese equities. And I was still being troubled by the fact that although my Chinese was okay, it wasn't quite it, it wasn't quite good enough. I didn't, I didn't really feel like, um, I should be the China expert investing in a Chinese fund. So I, I thought, I, I felt that I was, I should be seriously fluent, and I was quite good. Um, so I, in my spare time, I was continuing to work on developing the Chinese learning system.
And Ed Cook and Greg Detre, who are both also and also do psychology, at the same time as me, um, were working on a sort of generic system to bring a lot of the findings of the science and, and the art of memory. So Greg's a was a scientist at Princeton, a neuroscientist after leaving Oxford. So he was very much bringing the science of memory, um, and Ed is a grandmaster of memory, and does a lot of competitive memorizing, so he was bringing the, the sort of art of memory side. Fantastic. And so they were building this platform, um, and so I got, I got together with them in 2010 to bring all the content that I'd made for learning Chinese, which I'd been turning into a site, but hadn't got the sophistication of the back-end, um, built. And so we just merged them together. And, and that became the first kind of Memrise courses, the Chinese courses. Fantastic. Um, we, and we very quickly then expanded it out to be vocabulary across all languages, and then expanded other subjects as well. So any, any kind of fact learning. Right.
And that's a great segue into what was going to be my next question about, uh, how Memrise got started, but, but there you go. And of course, I mention it all the time on the podcast because I really enjoy the intuitive design, so easy to use, rewards you with the points, and just the power of spaced out repetition. Yeah. Um, obviously, spaced out repetition, you know, not a new thing by any stretch of the imagination. What do you think sets Memrise apart from, you know, other software, other services that offer similar things?
Yeah, so, um, spaced repetition is sort of fine. Um, although it's a starting point, rather than a, a piece of genius all on its own. Of course. Um, and I think, so one of, one of the things that we do differently is the massive store data that we've got, um, across all the courses, which gives us, so what I mean by that is that, that we have the same central items are used across different courses. So, no matter which course you're learning on, if you're learning a particular Chinese word, we amalgamate all that data, so we understand which words people find easy, which words people find difficult in general. Fantastic. So we're able to, to tweak those numbers, so that the, the algorithm is just more effective for, for everyone.
Um, and those numbers you're talking about is the, the time it takes for a word to be rewatered, as, as the term is on the site. Yeah, yeah, exactly, exactly. So the time it needs until we, until we, and that, uh, when it comes to you for watering, is the point at which we estimate you're going to forget it. Right. Um, and the ideal time to be retested on something is when you are as close as possible to forgetting it, but haven't forgotten it yet. That's the idea of the water.
And so, and this actually was a subject of, uh, Greg Detre's PhD, was looking at the tip of the tongue effect. And finding that if you have a tip of the tongue experience, where you're really straining to remember something, then, um, that is the point where if you satisfy that experience, then that is maximally strengthening to your memory. But conversely, if you fail to satisfy that experience, it actually weakens your memory. Wow. So if you try to remember a word, and you fail to, that actually weakens, that weakens the memory, so you should go and look it up straight away. So that, so the strength of your algorithm is actually in, in, in finding that little moment before, and if you find it the little moment after, it's actually detrimental.
It, it, well, actually, because you then have the, the next presentation, and then we get you to copy type it as well, which does mean that you're doing a little bit of the test, which is designed to head off that weakening effect. Right. So the, the weakening effect really happens if you're out trying to talk to someone in a conversation in Spanish, and you're searching for a word, and you, and you can't get it, and then you move on and forget about it, then your memory for that word has been weakened. Right. So in that situation, you should stop, look it up, make sure that you've actually got it, and, and then your memory will be, will be saved.
That's a good tip. I never, I never actually heard this that, that's how the dynamic of remembering actually is, um, it's a good practical tip. Do you have any other practical tips for people who might not be totally familiar with the platform or anything? How do you approach it? Do you study a lot every day? Do you study a little bit every day? Do you spread it out? Do you do multiple languages?
I mean, there's, there's, I mean, there's, there's ever, there's the tension between the practical and the ideal. So the, the ideal is to space out your learning as much as possible, do a little bit at a time. Um, a little and often is ideal, but that's, that's fairly impractical for a lot of people. Um, it's much more practical on the mobile app, and, and we see that people use it much more in little bursts, um, through the day, and, and that's, that's a great way to learn.
It is a, uh, that's just reminded me of another, another study put by one of our, two of our scientific advisers, who found that when you're, and this is about the, uh, contrast between the practical and the most effective. Um, that if you ask people, if you show them, give them a word in a new language, and ask them to guess what it means, before you tell them, Right. then, they, once you tell them what it means, they will actually remember that word better than if you just told them what it means straight up. Wow. Which is cool. That's really cool, yeah.
But they think they'll remember it worse. Okay. And, and this is a, this is a classic example of the kind of tension we have in designing the product, that what's actually best for your memory is what you think is worse for your memory. Oh, no. And, and that means that if we, if we, if we put that feature in and asked everyone to guess at each word, as they were going through, they would just come on, they'd be like, guess, I don't know it, that's what I'm here, I'm here to learn it, you idiot. And they would leave and just, stupid, uh, and, and just shut down and leave, go somewhere else. And the fact that we were implementing a system that's more effective, uh, would never come to fruition, because people wouldn't believe it. Right.
And so a lot of what we're trying to do with the designs is to work out how to give people those benefits at the same time as sticking with what they with their preconceptions of what they should be doing. Right. Um, because the preconceptions are obviously a lot for of motivations, and you've got to work with people's motivations. Of course.
I mean, the whole, uh, to optimize, I guess that exists in other branches of science as well. I mean, the keyboard has been a speculation of many, uh, theories that it's designed in a less suboptimal way so you don't, you don't kill your fingers or joints or something. I, I don't know if that's true or but I know,
it's, it's so, it was so that typewriters didn't jam. Right. So you could press it too fast and so you had to make the typewriter a little less effective in order for it to work better. So that's kind of the analogy to, to what you're saying about, you could implement features that would make it more optimum or ideal, but people just could, wouldn't understand it.
Well, well, um, I, I wouldn't be as unfair on people as that, because it, there is a major factor in the, in the forming of the memory that is emotional and is to do with your, your motivation. And if you're being told to guess at the meaning of words, that is quite an annoying thing to do. Right.
And, so, a better way of doing, really, what, what that's, but the core of it, what that's doing is getting you to think about the word. And so, you're looking at the word, and you're thinking, well, it looks a bit like this word that I know, Right. or that word in German, so maybe it means something like this. So, what, in order to get you to do that, we, the reason that you, um, form a stronger memory is that you're thinking about it. Right. And that is exactly what you, what visual mnemonics make you do, and the process, or any, in fact, any mnemonic, and the process of making a mem is the process of thinking about it. Right. And, and so, that is another way of sort of, of trying to give the positive effect of, um, guessing the meaning, but through a different route. I, looking at the word, what does that, what English word does that sound like? Okay, try and make something that sounds like it, and then try and link that into a funny little phrase that's going to help you remember it. Right.
But you can't always do that though. I mean, uh, I've done a lot of mems and I've also come across a lot of words, I mean, for instance, Russian, come across a lot of words where I just, I just can't think of any words, even in two or three different languages that sounds anything like it. What, what do I do then? Do I just kind of skip it or do I just kind of write down parts of it? What, what would be the ideal there?
Um, interesting, can you give me an example of a word which you can't find anything that sounds like it?
Oh, God, I wish I could. But I, I've definitely seen some, or maybe it's just because I'm lacking creativity, but I, I've definitely come across it.
I, I'm pretty sure that you can, you, you can manage it with any words. Um, I mean, they, you get pretty strange, and, and they get, and they, they can get pretty tenuous. And the next person coming along might look at it and think, cocky, this guy's a lunatic. Right. But that doesn't actually matter. Or, all you're trying to do, you're not trying to create the perfect way to remember a word. All you're trying to do is give your brain a way to think about it. Right. And, and so, an a way to start recalling that memory, and start getting that pattern together. So even if it really doesn't sound very much like it at all in the abstract, that's okay. As long as it does enough to start re-forming that memory for you. Right.
And what about, uh, a lot of people obviously create mems on these courses. Um, is there a, is there a weakness in, in using other people's mems over creating your own? And if so, how big is that weakness?
So there's, there's a, there is a tension between that. There are some, particularly say the Chinese courses, you're going to see some of this, with Chinese characters, Yeah. Um, can have, sorry. Yes, please. if you're on the blue one. Okay, don't worry, don't worry. Okay. Yeah, yeah, go there, go there. Yeah, yeah, sure. Anyway, all of the banks. Cool. Okay. That's. Now both are okay. That's. Just make sure you have all the when you no worries. No worries. Um, So, where, where was I? The difference between creating your own mems and using pre-create.
Okay, so, so in the Chinese characters, for example, a lot of those are sort of process, you know, there's a picture that animates into the Chinese character, and say, well, that's pretty easy. I'm going to, and that allows you to think about it as well. So you see, when you're watching those gifs animate back and forth, you're thinking, oh, okay, I see, oh, yeah, I see how that turns into that. And then when you next see the character, you kind of imagine a bit of that happening, and you, and it's quite easy to make your brain to make the step to the picture that it came from. Right. So that, those ones are kind of nailed, you don't need to draw your own animation for every Chinese character.
For, um, other words in other languages, it's going to depend so much on what sounds right for you. Sometimes you'll find one which is totally nailed, like, one of my favorites is the Spanish word, "Polvo", which means a one-night stand. Oh, really? And, uh, there's a great mem where there's a picture of a Volvo car, um, and there's kind of dusty bonnet with handprints, um, on it. And it, and it, and a general sort of bodily imprint on it. And the, the caption is, because he drove a Volvo, it was only a Polvo. Probably, you probably have. And it's, it's basically unforgettable. Right. And how awesome, isn't it? The, the more bizarre and extreme they are, the more you remember them. Yeah, and that's totally crazy.
But on the other hand, um, a lot of the times there may be words like the ones that you mentioned that, don't, don't have a very easy match, where your target shouldn't be to try and make a perfect, absolute killer, this is nailed it, mem. But means something that works for you. So, the idea of, of a kind of community crowd-sourced imagination is that we can pick off as many as possible of the ones that are an absolute, diamond, cast iron mem, so, so that we can help you as much as possible, so that when you get to ones where they are more difficult, you can have ta take the time to do it yourself. Yes. And, and, that, that's the kind of it takes the onus off you. Because it is, it feels like a lot of work to make a mem for every single item. Whereas if you're making a mem for one in 20, then you can invest that time.
Yeah, definitely. Um, Ben, I know the time is, is running out, but I just wanted to ask you, um, one more question, or two. Um, is there a problem, you know, there's a lot of the content on the site is, is user-created. How do you manage the quality control of making sure that what's on your site is, is of high enough quality to, to be there? Do you have any such requirements?
Yeah, we, I mean, it's, it's community edited. Um, we have the central dictionaries, um, where there are little teams of three to five people who work on each of those dictionaries, making sure that there's audio files, that, that the words and definitions are correct, that there are acceptable alternatives, so that if you type the word without, um, or a slightly different translation that's still correct, that will be accepted. Um, and then whenever you create a course yourself on Memrise, you draw from those dictionaries. So you immediately get the tactical, tactical. Right. Um, you can also then edit it, because sometimes you want to have a different primary definition to the one that the, um, the dictionary has, so you can tweak that around. Um, and the, and then there's user feedback to, for the forums on the course, asking the course creators to, there, there would to make any changes, as and when people see them.
So, the way, and, and then what we do is these, so we feature up the top, we're, we're always going through, um, curating and editing the courses to turn up on the browse pages, to make sure that the most accurate ones, or at least the ones that are the best courses are accurate, and those show up first on the browse experience. Fantastic. So it, it, it's definitely still possible to find courses that, um, people have just put up and haven't checked properly. Um, normally someone will have commented in the forums, uh, saying, this doesn't look, this doesn't look right. Right. And, and so you can then see that. So, so really, the judgments there are, looking at, um, courses that a lot of people are using, courses that are high on the, um, on the browse page, those ones will be checked. Fantastic.
Okay, uh, the last thing I had today was that, you know, it always surprised me that Memrise was 100% free to use, because similar sites would always have some kind of premium plan. And then I recently heard you speak about, you would thinking of implementing some premium, uh, premium version. I was wondering if you could give the listeners a little, um, preview maybe into what kind of feature that would be, and also if, Yeah, sure, sure. Will it change the, the usage of the basic feature that we have today?
No, we'll also be, um, extending the basic the, the main site usage. So in, in terms of kind of freemium price models, um, there's there are different ways of having freemium pricing models. There's Sure. a what I would call a kind of bouzou model where although you can use the program, you can't really use it. Yeah. I know what you mean, yeah. Um, and, and you, it's a, it's a very impoverished version, and you really, you need to pay to use it. Right. That's not what we're doing.
There's another version which is more the Skype or Evernote model, where everyone uses a really good product for free, and there are some features that you can pay for if you want. Right. And that's much more what we're going for. Um, and the specific things that we are doing, uh, it's all around the data, um, that we have and the insights that we get from that data.
Um, and using those, we, we're working out different profiles, different clusters of users. It's really, it's really extending what we've done through the research into different, um, how people learn, when they're most likely to forget words. Right. And we're now doing another deeper dive into all the data we've got. We've got much more resources to do that now. We've got a partnership with, um, some scientists in University College London, and Oxford University, and also in-house, we've got someone as well. And so we're going to be building out richer and richer understandings of what type of learner you are, and therefore, specific to you, what is the best thing for you to do right now to improve your learning and, and learn more effectively. Right. And, and I think the way to see this is that this is really like research and development, very deep into understanding how people's brains work more effectively. And passing on to you.
That sounds, that sounds really incredible actually.
Uh, it, it is. And those, and those tools are going to be, they're going to come first to the premium users. And I think that what we'll do is then, once they're, it's gradually move those into the free experience as well. So that the premium users will always be this, like, right at the cutting edge, really taking it to the next level. And then the, the, the main free site will always be improving as well. Right. And, and We'll always be getting a richer feature set there that they'll benefit from.
But surely also just supporting, you know, a lot of people are willing to, if you had a big donate button, for instance, I'm sure a lot of people would chip in, and that will kind of be the same as having a premium option.
Yeah, it, it, it's not far off. Um, we have considered having premium, Memrise Premium. It's exactly the same, but we like you more. Yeah. So acknowledge appreciation. Yeah, and, and really what, what we're in, in a way, this is just first, it's just first access to the best new tools that we're building. Um, and we will, as I say, we'll be releasing better tools to the main site anyway, and continue to improve that.
Fantastic. So what is the, do you have any kind of timeline on that? Is, is that this year or is it, is it
That's this year, yeah.
Wow, okay.
We, we should, I mean, some people are testing. We, we have early versions of it in testing right now. So we, we have a few hundred users on that. Um, we aim to roll it all out by kind of the beginning of August. That kind of time scale.
Fantastic. So that's coming soon. That's coming soon. All right. Ben, thank you so much for, for coming on. I know I took a little bit more of your time than we, we agreed on, but I hope everything went okay.
It was a pleasure.
All right. All right, thank you so much and, uh, I will see you on Memrise. Take care. Bye.