Pedagogical scenario · AI-LANG CELV Graz 2026

SimuLang

Preparing a job interview in the target language using generative AI, on Moodle.

Moodle + LLM Adult job seekers FR · NL professional
CEFR Progression
A2 B1
Professional French
Professional Dutch
Purpose

Develop professional language competences (French or Dutch) through AI-simulated job interview practice, fully integrated into a Moodle sequence. The AI handles volume, immediacy and judgment-free feedback; the trainer focuses on what AI cannot assess: prosody, pragmatics, cultural adaptation.

Language objectives · CEFR
💬

Coherent oral discourse in a professional formal register

📖

Sector-specific vocabulary relevant to the learner's target field

🧱

Structured responses using the situation → action → result framework

🔎

Self-correction of syntactic and lexical errors through CEFR-based reflection

Sequence · 4 sessions on Moodle
1
Resources & lexical activation
Sector vocabulary sheet, formal register phrases, annotated interview video. Learners log 5 reusable expressions in their Moodle learning journal.
Pre-simulation
2
LLM simulation · 20 min
Learners access a custom simulation interface via a Moodle link — no account required. The LLM is pre-prompted by the trainer to play a sector-specific recruiter, flags register errors and proposes reformulations in the target language. At the end of the simulation, a single button sends the transcript automatically to Moodle via a n8n webhook (Moodle REST API). Interface powered by Mistral AI (EU infrastructure, GDPR-compliant). No personal data stored beyond the session.
AI simulation
3
Self-assessment on CEFR grid
Moodle submission: skills in progress, recurring errors flagged by AI, retained reformulations. Builds language-learning reflexivity.
Reflexivity
4
Trainer-led debrief · 10 min
Individual session focused exclusively on what the AI cannot evaluate: prosody, silence management, cultural adaptation, discursive posture.
Human
Role distribution
Artificial intelligence
Unlimited practice volume, no fatigue
Instant feedback on syntax, vocabulary, register
Available outside sessions, 24/7
Judgment-free space (key lever for job seekers)
🎯
Trainer
Designing the sector prompt and calibrating difficulty
Regulating pragmatic and cultural competences
Assessing what the AI cannot simulate
Validating CEFR progression
Full pedagogical scenario · AI-LANG CELV Graz 2026

SimuLang

An AI-supported language learning scenario for adult job seekers, built around job interview simulation using a Large Language Model, fully integrated in Moodle.

Moodle + LLM Adult learners · Professional insertion French & Dutch professional CEFR A2 → B1

Context & rationaleBackground

SimuLang is a language learning scenario designed for adult job seekers enrolled in professional training programmes. It addresses a concrete pedagogical challenge: developing oral and written communication skills in a second professional language (French or Dutch) under real workplace constraints, while maximising practice time within a limited number of face-to-face sessions.

The scenario relies on a Large Language Model (LLM) integrated into Moodle to simulate a sector-specific recruiter, providing learners with repeated, judgment-free, immediately corrective practice. The trainer is freed from repetitive linguistic correction and can focus on higher-order competences that AI cannot assess.

Audience & CEFR objectivesWho is this for?

Target audience: Adult learners in professional reintegration programmes. Entry level: A2. Target level: B1 (CEFR). Language: French professional or Dutch professional, depending on the learner's target sector.

Language learning objectives at B1 level:

Coherent oral production in formal professional register
Produce extended spoken discourse adapted to a formal professional context, using appropriate tone and register without major breakdowns in communication.
Sector-specific vocabulary
Mobilise and correctly use vocabulary relevant to the learner's target professional field (healthcare, logistics, administrative services, etc.).
Structured argumentation
Construct responses following the Situation → Action → Result framework, a standard professional communication structure across sectors.
Linguistic self-correction and reflexivity
Identify recurring syntactic and lexical errors through structured self-assessment, and apply targeted corrections using a simplified CEFR grid.

4-session sequence on MoodlePedagogical design

1
Pre-simulation
Resources & lexical activation
Learners access a Moodle resource page containing a sector vocabulary sheet, formal register phrase bank, and an annotated video of a professional interview. Task: identify and log 5 reusable expressions in their Moodle learning journal. Learning output: Moodle learning journal entry.
2
AI simulation · 20 min
Job interview with the LLM
Learners access a lightweight simulation interface via a direct Moodle link. The interface is deployed on institutional infrastructure and calls a language model API (Mistral AI, EU-hosted) with a pre-configured system prompt — no learner account or registration required. The prompt, designed by the trainer, instructs the LLM to: ask standard interview questions adapted to the learner's target sector, flag register and syntax errors at the end of the exchange, and propose reformulations in the target language. The simulation runs in written form (with an optional oral variant using a transcription tool).

Transcript tracking: At the end of the simulation, a single button triggers a webhook to an n8n automation server (deployed on the same infrastructure). N8n automatically pushes the transcript to the learner's Moodle assignment via the native Moodle REST API — no manual copy-paste required. As a next step, xAPI statements can be sent to a Learning Record Store (LRS) connected to Moodle via the Logstore xAPI plugin, enabling fine-grained tracking of each exchange, error and reformulation.

GDPR compliance: Mistral AI processes data exclusively within the EU. A Data Processing Agreement (DPA) is in place. No personal data is stored beyond the session. Learning output: Simulation transcript automatically submitted to Moodle assignment.
3
Reflexivity
Self-assessment on CEFR grid
Learners submit a structured self-assessment on Moodle using a simplified CEFR grid: which language competences are progressing, which recurring errors were flagged by the AI, and which reformulations they are retaining. This step develops metalinguistic awareness and learning autonomy. Learning output: CEFR self-assessment grid.
4
Human · 10 min
Trainer-led individual debrief
The trainer reviews Moodle submissions and conducts a short individual session focused exclusively on competences the AI cannot evaluate: prosody and intonation, silence and turn-taking management, cultural adaptation in professional communication, discursive posture and nonverbal communication (where applicable). Learning output: Debrief summary note.

Human-AI collaborationRole distribution

⚡ What the AI handles
  • Unlimited practice volume without fatigue
  • Immediate feedback on syntax, vocabulary and register
  • Availability outside sessions, 24/7
  • Judgment-free environment (a key lever for adult learners in vulnerable situations)
🎯 What the trainer handles
  • Designing and calibrating the sector-specific prompt
  • Regulating pragmatic and intercultural competences
  • Assessing non-simulable competences (prosody, posture)
  • Validating overall CEFR progression

Why this mattersTransformative value for language learning

Volume problem solved. A single trainer cannot provide 20 learners with sufficient individual oral practice time. The LLM multiplies practice opportunities without multiplying trainer workload. Each learner can run several simulations per week, at their own pace and outside class hours.

Affective barrier reduced. Adult learners in professional reintegration frequently report anxiety about making mistakes in front of a trainer or peers. The AI creates a low-stakes, private space where errors are corrected without social cost, increasing willingness to take linguistic risks.

Trainer time redirected. By offloading surface-level linguistic correction to the AI, the trainer's intervention time is concentrated on higher-value, non-simulable competences: the subtleties of professional communication that no LLM can currently model reliably, and the human relationship that sustains motivation.

Moodle as the connective tissue. The scenario does not require any external platform. All learner productions, reflections and trainer notes are centralised in Moodle. Transcripts are pushed automatically from the simulation interface to Moodle assignments via an n8n webhook and the native Moodle REST API. As a next step, xAPI statements can feed a Learning Record Store (LRS) for fine-grained competence tracking.

GDPR by design. The simulation interface relies exclusively on Mistral AI, a European provider processing data within the EU. A Data Processing Agreement is in place. No personal data is stored beyond the session. Learners are informed prior to use. This architecture makes the scenario deployable in institutional contexts without legal friction.