Bo Bennett on Language & Mind

HOST: Welcome back to "Language & Mind." I'm Dr. Anya Petrov. Today, we're joined by Bo Bennett, social psychologist and the founder of ScienceBasedLearning.com. Bo, when you call this "science-based," you're making a stronger claim than most apps do. What's the actual psychological mechanism behind spaced repetition that makes it so powerful? GUEST: Well, thanks for having me. The spaced repetition has been a proven technique in science and language learning for quite some time. And by proven, I mean study after study of which techniques work the best for long-term memory and recall. Spaced repetition always comes up on top, and that's the whole idea of instead of just remembering words, and then once you get it right on a flash card, for example, you go to the next one, and once you get it right, then you just put it in a I I know this pile. Well, you may know it for 15 minutes or 5 minutes, but then you're going to forget. We have different little compartments of our memory. Some are it goes anywhere between very short term and very long term. So the ideal the idea behind spaced repetition is that we test each compartment. So we tested after 15 minutes. And if you still know it, great. Let's try you in an hour. We tried in an hour. If you still know it, great. Let's do it in a day, 3 days, 7 days, and then 30 days. If you remember a word like after 30 days, then it's a pretty good sign, probably like 95% that this is a word that's going to stay with you for long term. And again, it's something that's been tested in science and demonstrated time and time again. HOST: That's such a fascinating way to think about it, that we're testing each compartment. Now, your platform uses this SM-2 algorithm to manage that sequence, but it also enforces active recall through translation, speaking, and answering questions. From a cognitive science perspective, is active recall doing the same cognitive work whether I'm translating a paragraph or having a voice conversation, or are those fundamentally different memory processes? GUEST: They're maybe not fundamentally different, but they are different enough where it it it matters. So, when people start learning a language, if they only learn a language by reading, then they're going to be really good at reading. But, when it comes time to speak the language or listen if somebody's talking in the language, they're going to be completely lost. So, it's very important that you have to you practice all of those different aspects. You practice the reading, the translation, the vocabulary, speaking it, answering questions like being interactive, and more. And if you could if you could kind of fill all those buckets, then you're going to be very well rounded, and that's the way to to become fluent and and learn a language. Well, HOST: So we need to work on production, not just recognition. Speaking of those different buckets, your Testing Tenses tool customizes drills for specific difficulties, like the Spanish subjunctive or German cases. As someone who built this across 15 languages, what did that process teach you about how structurally different they really are, and does cognitive science have anything to say about why certain grammatical structures are just harder for English speakers to internalize? GUEST: Well, in terms of the different languages, and I have to be honest here, that I am not absolutely an expert in 15 different languages. Building the app, I didn't need to be an expert in 15 languages. I just needed to understand the concept of how language tenses differ in the different languages in order to put this together. HOST: That puts it into perspective, focusing on the concepts, not just the specific details of each language. Now, your platform also differentiates by using AI to generate fresh content—articles, conversations, even custom phrase books at the learner's level. But linguists often argue that authentic language, the way real native speakers write and speak, has patterns and cultural texture that AI might flatten. How do you think about that tension? Is AI-generated language good enough for most learners, or does it eventually become a ceiling? GUEST: Well, I think it's certainly good enough, and that's even kind of an undersell because AI does extremely well with those patterns you're talking about, like the cultural aspects of it, authentic language. It's not like your your grandfather or your grandfather's AI. Or the old school technology. Um, of course, it didn't exist back then, but you get the point. It's changing, and it has been improving constantly at an incredibly impressive rate. So, what you would even experience a year ago with AI learning compared to what it is now, very different. And we make sure that our app stays on the top of the technology and uses the best LLM to do this kind of translation to help people learn. HOST: So maybe the gap between authentic and generated is shrinking faster than many think. Before we go—for listeners who want to follow up on what we covered, where can they find you and the work you're doing? GUEST: You can find all about me at uh bowbennett.com. That's my personal website. But more importantly, sciencebasedlearning.com is the website of the application. And you could currently get this on iTunes and on the Android store and load it on your phone. HOST: Thanks so much for coming on, Bo—the idea that we should test multiple compartments of our memory is a really concrete takeaway. If you're enjoying "Language & Mind," subscribe wherever you get your podcasts—Spotify, Apple Podcasts, Amazon Music—so the next episode lands in your feed automatically. And thank you for spending part of your day with us. Until next week—AIHosts.fm signing off.