The Problem Every EFL Teacher Knows
You have 25 students. Some have been watching Netflix in English since middle school and understand everything. Others freeze when asked "How are you?"
You know the truth: one lesson can't serve both.
Pitch to the middle (B1) and the A2s disconnect while the B2s zone out. Differentiate with tiered worksheets and you spend more time preparing than teaching. Either way, actual speaking time per student stays painfully low.
That's the mixed-level paradox: the more diverse the group, the more each student needs individual attention — and the less you have to give.
Why the Usual Solutions Fall Short
The Single Textbook
It targets one level. Everyone else suffers.
Ability Grouping
Logistically impossible in most institutions. And pedagogically questionable: A2 students never hear B2-level English, which slows their natural acquisition.
Tiered Worksheets
They work on paper. In practice, you prepare 3 versions of every exercise, manage 3 parallel streams, and students feel labeled. "You're in the weak group" has never motivated anyone.
Vocabulary Apps
Useful for memorization. But memorizing words has never taught anyone to hold a conversation with a fast-talking Australian using slang.
The Idea: Every Student Speaks at Their Level — in the Same Scenario
Imagine a class where:
- The same scenario is presented to everyone: "It's your first day at a London startup."
- Characters speak at different levels: the canteen barista is patient and speaks simply (A2). The British manager uses fluent professional English (B2). The Indian tech lead speaks precisely and quickly (C1).
- Each student chooses who to talk to based on their comfort level. The beginner starts with the barista. The advanced student goes straight to the manager.
- Characters adapt: if a student hesitates, makes errors, or asks to repeat, the character simplifies. If the student is fluent, the character speeds up and uses idioms.
The result: every student speaks English for the entire session, at a level that pushes them forward. No dead time, no "I don't understand anything," no "this is too easy."
How It Works in Practice
1. You Set the Level Range
Not a single target level. A range: A2 → B2, for example. MEτiS automatically spreads characters across that spectrum.
2. AI Generates Calibrated Characters
Each character has:
- A precise CEFR level (the level they speak at)
- An accent (British, American, Australian, Indian…)
- A personality (patient, direct, formal, laid-back)
- Hidden information the student must uncover by asking the right questions
The Australian barista speaks simple English with slang. The British director uses conditionals, indirect politeness, and doesn't rephrase unprompted.
3. Students Actually Speak
Text mode or phone call mode (speech recognition + text-to-speech). Each character has their own voice, their own accent. The student doesn't type an answer — they talk to someone.
An A2 student calling the barista to order coffee practises:
- Formulating simple requests
- Food and drink vocabulary
- Understanding an Australian accent
A B2 student negotiating with the manager practises:
- Conditionals and reported speech
- Management vocabulary
- British cultural codes (understatement, indirect politeness)
Same room, same scenario, same hour — but everyone works at exactly their level.
4. Bilingual Task Instructions
Assignment briefs exist in both English AND the students' native language. A toggle button switches between them. Beginners read instructions in their L1 first to understand what's expected, then work in English. Advanced students stay in English throughout.
5. The Debrief Is Rich
At the end of the session:
- Team 1 talked to the barista and HR — they got the vibe but missed the strategic issues.
- Team 2 went straight to the manager — they have the professional vocabulary but missed the cultural context.
- Team 3 cross-referenced all the characters — their deliverable is the most complete.
The debrief isn't about "who has the best grammar" but about "who conducted the best investigation in English." Language is the tool, not the subject.
What the Research Says
Krashen's Comprehensible Input hypothesis (i+1) states that learners progress when exposed to language slightly above their current level — not three levels up, not at the same level.
The problem with traditional classes: everyone receives the same input. For some it's i+3 (incomprehensible), for others it's i-2 (boring).
With level-calibrated characters, each student naturally receives their i+1: an accessible challenge in a motivating context.
Research on differentiated instruction (Tomlinson, 2001) confirms: the most effective differentiation doesn't create separate groups — it offers different entry points into the same task.
That's exactly what a scenario with multi-level characters does.
Honest Comparison
| | Traditional Class | MEτiS | |---|---|---| | Speaking time per student | 2-3 min / hour | 30-40 min / hour | | Level adaptation | One level targeted | Each character calibrated | | Prep time | 1-2h per tiered worksheet | 5 minutes | | Skill practised | Grammar, vocabulary | Communication in context | | Beginner motivation | Low (sense of failure) | High (patient characters) | | Advanced motivation | Low (boredom) | High (demanding characters) |
How to Get Started
- Identify the range of your group (e.g. A2 → B2)
- Pick a setting: first day at the office, hotel complaint, job interview, school trip…
- Let the AI generate characters with levels spread across the range
- Launch the session — students enter in teams, talk to characters, submit a deliverable
- Debrief on what they discovered, not what they conjugated wrong
The first lesson takes 5 minutes to prepare. The debrief will take care of itself.
Experience the situation. Understand the stakes. Take a stand.