AI Advent Speaking and Writing

Set up a useful practice task

Teachers create writing tasks (emails, translation or free writing) and speaking prompts, then add a situation, source text, target length, formality and task points. A point can ask for more than one idea. Add a time plan, phrase bank and guidance where they help learners get started.

School, activity and task tips can teach the language your class is practising. Keep each tip specific and usable; examples help learners see the pattern. Add grammar patterns as needed. The most specific matching tip takes precedence, and the tip version used for a marking is recorded.

Assign activities to a class, group or selected learners, set due dates, and configure per-learner daily marking and rewrite limits. Teachers can use the sample-answer tool to check how current tips work before learners begin.

How learners practise

Learners choose an assigned task, read its instructions and write a response or complete a speaking exercise. Writing includes a word count, special-character helpers, a time plan, school-tip checklist and phrase chips. Drafts are saved as learners work, so they can return to an unfinished response.

Speaking practice captures a transcript using the learner’s browser speech-recognition feature. This release can mark that transcript, but the AI partner that would take conversational turns is disabled pending price approval. Speaking recognition availability depends on the browser; the plugin documentation identifies Chrome, Edge and Safari as supported options, while Firefox may not provide the required built-in recognition.

Think first, or see the explanation

The activity’s correction setting is chosen by the teacher. In Make me think mode, a learner first sees a clue pointing to a school tip or language rule, without the correction. They can try a rewrite before revealing the explanation and examples. In Explain mode, feedback shows what to change, why it is wrong, the corrected form, the rule and short examples. Teachers can choose a two-step sequence, think-only or explain-only; learners do not switch the activity’s mode themselves.

On a later draft, feedback can note what improved since the previous marked version. It also highlights recurring tip or grammar issues, so the learner has a focused next step rather than a list of corrections alone.

Read the feedback

Each result gives a summary and a score for task, communication and form. Scores use 0, 3, 9 or 15 bands: the school’s stricter score is shown first, with a realistic exam score alongside it. Feedback considers task points, clear connections and suitable register, as well as grammar, word order and spelling at the learner’s level.

Task-point feedback quotes the learner’s response and labels each point as covered, extended, short, missing or mismatched. When an idea needs more work, it offers phrases to help. Other feedback calls out strong expressions, language corrections and communication strengths, and may rewrite one of the learner’s own sentences using the school’s tips.

In think mode, correction details remain hidden until the learner chooses to reveal them. Use the feedback as a teaching aid: review it against the task and the learner’s level, especially before relying on a score for a high-stakes decision.

Speaking: practise, then self-correct

Feedback considers whether the learner was understood and kept the exchange flowing, alongside their transcript, pauses and words to practise. A learner who notices a slip and corrects it while speaking is showing a useful skill: self-correction is praised, not treated as an error that lowers the score. Normal hesitations and restarts should not be penalised.

The plugin may offer browser-based text-to-speech or a repeat/correction practice action where the browser supports it. Browser speech recognition and speech playback do not consume LMS Labs voice credits; browser providers may handle speech data under their own service terms.

What teachers can and cannot control

Teachers set the correction mode, class tips, task-specific criteria and tips, assignments, daily marking limits and rewrite limits. Reports and class analytics help identify repeated learning needs and assign practice. Sample answers let a teacher preview marking against current guidance; they are charged like other valid marking requests.

Teachers should review generated feedback and scores in context. The system applies the configured task and school guidance, but does not replace a teacher’s judgement. Conversation turns from an AI partner are not available while that service is disabled.

Privacy and the services involved

For marking, Moodle sends the answer or speaking transcript, task instructions, relevant school tips, feedback language and limited learning context to LMS Labs. The marking request does not include a learner’s name, email address or Moodle user ID. Moodle stores drafts, marked answers and feedback, scores, focus records, assignments and paid-request records under its own privacy controls.

The plugin does not record or store speech audio. Browser speech recognition can involve processing by the browser maker or another browser provider (for example, Google or Apple). Check the browser and provider terms and your organisation’s policies before learners use speech features.

The supplied integration notes say the marking provider request uses a Europe-addressed API endpoint with storage disabled, but also warn that this does not establish the provider project’s EU configuration or a signed data-processing agreement. Do not treat this as a promise of EU data residency, GDPR compliance or a completed privacy review. Agree applicable retention, provider and data-protection arrangements before classroom rollout.

If marking is slow or interrupted

A valid result costs one credit when it is delivered, including a teacher’s sample marking. Empty or too-short answers and refused or invalid marking results are not charged. Reopening feedback does not submit another marking request.

If Moodle is still waiting, let it check the saved request again. A retry uses the same saved request key and answer; replaying that request returns its original result without another debit. Avoid starting a fresh submission while the existing one is pending. If Moodle reports that the old request has expired, submit again only when you are ready to create a new marking request. A network interruption can leave the acknowledgement uncertain, so check the saved request status before trying again.

Declared package support

The supplied 1.0.1 beta README declares Moodle 4.4–5.3, PHP 8.1 or later (PHP 8.3 or later on Moodle 5.3), and PostgreSQL, MySQL or MariaDB. These are package declarations, not evidence that all supported combinations have passed completed environment tests. Run your own Moodle acceptance checks before use.