Most language apps are built on a clean, falsifiable premise: repeat enough, and you will retain. The logic is borrowed from motor learning. Practice a chord progression until your fingers stop thinking. Drill vocabulary until the word surfaces before you reach for a translation. It is a reasonable model. It is also, for most adult learners, wrong in the way that matters most.
Retention is not the bottleneck. Being willing to speak is.
The wrong model
When I started building Samovar, a Russian language learning tool that treats the language as lived context rather than textbook instruction, I kept running into the same observation: people who had studied Russian for two or three years would go quiet the moment a native speaker entered the room. Not because they had forgotten their vocabulary. Because the social stakes had suddenly become real.
This is not a niche problem. It is the central problem of adult language acquisition, and most EdTech product design ignores it entirely.
The standard product response is to add gamification, streaks, confidence scores. These are not wrong exactly. But they address a symptom while leaving the mechanism untouched. A streak tells you how often you showed up. It says nothing about whether you were willing to be misunderstood, corrected, or slow in front of another person.
What building Samovar revealed
The research behind Samovar started with a simple question: when do people actually feel like they are learning a language, versus when do they feel like they are studying one? The answers were consistent enough to be uncomfortable.
People described real acquisition moments in social terms. A conversation at a market in Novosibirsk where they had to ask for directions and the vendor did not switch to English. A phone call they could not hand off. A joke they understood before anyone explained it. The common thread was not repetition. It was consequence. Something was at stake, and they were the only person who could handle it.
Studying, by contrast, they described in private terms. Headphones in. App open. No one watching. Safe.
The gap between those two modes is not motivational. It is structural. Most language learning products are built entirely for the second mode while claiming to produce the first.
The social stakes problem in language acquisition UX
This creates a specific design problem. If language acquisition is primarily social, then a product that removes all social friction to reduce anxiety is also removing the mechanism that makes language stick.
You cannot replicate the feeling of being the only person in a room who might not understand. Not with a progress bar. Not with an AI conversation partner that will wait patiently and never look bored. The anxiety is not a bug in the learning process. It is closer to the engine.
This does not mean good language learning product design should be cruel or deliberately stressful. It means the design question shifts. Instead of asking "how do we make this feel safe?" the more useful question is "how do we make the stakes feel real without the consequences being catastrophic?"
That is a harder question to build toward. It requires understanding what learners are actually afraid of, which is rarely "forgetting the word." It is usually something more specific: sounding foolish to someone whose opinion matters, being marked as foreign in a context where they wanted to belong, losing the thread of a conversation in front of people who are waiting.
What this means for language learning product design
The implication for EdTech product research is that the unit of analysis cannot be the individual learner in isolation. You have to study the social environments where language actually functions: the dinner table, the workplace, the bureaucratic counter, the phone call with a landlord.
Samovar was built with this in mind. Rather than organizing content by grammar rules or frequency lists, it organizes around contexts where Russian operates as a social instrument. The assumption underneath the curriculum is that you are not learning a language. You are learning how to be a particular kind of person in a particular kind of room.
This reframing changes what counts as progress. Fluency is not a threshold you cross. It is a relationship you build with a specific set of social situations. Someone can be fluent at ordering coffee in St. Petersburg and completely lost in a formal meeting in Moscow. The same person. The same language. Different contracts.
Most language apps treat this as a problem of vocabulary range. It is actually a problem of social modeling.
The contract underneath the curriculum
The phrase "social contract" in language learning is not metaphorical. When you speak a language, you are making an implicit agreement with the people around you: I am willing to be judged by the standards of this community. I accept that my mistakes are visible. I am asking you to meet me partway.
That is a significant ask. It requires trust in the other person and a specific kind of courage in yourself. No app can manufacture that courage directly. But a product can be built in a way that acknowledges the contract exists, rather than pretending the whole thing is just a memory exercise.
This is the design philosophy behind Samovar, and it is the same philosophy that runs through the other projects at dmitrii.dk: Flatmate.dk, which treats housing compatibility as a behavioral question rather than a logistical one, and Indfodsretsproven, which treats Danish citizenship exam preparation as a way of navigating bureaucratic anxiety rather than just drilling facts.
The pattern is the same across all three. Observe what people are actually afraid of. Build toward that, not around it.
If you are working on a language product, a civic tech tool, or anything else where human behavior is the core design material, I would be glad to think through it with you. Learn more at dmitrii.dk.
FAQs
What is the main argument against repetition-based language learning?
Repetition builds retention, but retention is rarely the reason adult learners stop progressing. The more common barrier is social: the willingness to speak in front of others when something is at stake. Products that focus only on repetition leave this barrier entirely unaddressed.
What is Samovar and how does it approach language learning differently?
Samovar is a Russian language learning tool built around the idea that language is lived context, not textbook content. Rather than organizing lessons by grammar rules or frequency lists, it organizes around the social situations where Russian actually functions, treating language acquisition as a social process rather than a memory task.
What does "language learning as a social contract" mean in practice?
It means that speaking a language involves an implicit agreement with the people around you: you accept visibility, the possibility of being corrected, and the social exposure that comes with not yet being fluent. A product that ignores this contract is solving a different problem than the one learners actually face.
How does social stakes design differ from gamification in EdTech products?
Gamification addresses motivation by rewarding consistency. Social stakes design addresses the structural gap between studying alone and performing in front of others. The two are not mutually exclusive, but most EdTech product research focuses on the first while the second remains largely unsolved.
Why do people describe real language acquisition in social terms?
Because the moments that make language stick tend to involve consequence: a situation where you were the only person who could handle it, where you could not hand the conversation off, where something depended on being understood. Private study rarely produces that feeling, which is why it rarely produces the same kind of retention.
What can EdTech founders learn from the Samovar approach?
That the unit of analysis matters. If you study learners in isolation, you will build products for isolated learners. If you study the social environments where language actually functions, you will build products that address the real bottleneck: the willingness to be misunderstood in front of another person.
Where can I learn more about this approach to product and language design?
You can read more field notes and case studies at dmitrii.dk, or reach out directly at dima@babinov.dk to discuss a collaboration.