Author: Greg A. I. Dillon

One of the most time-consuming parts of preparing a world language assessment is not always writing the questions. It is recording the audio.

A teacher must write a suitable script, find a quiet room, record it clearly, listen to the result, and perhaps record it again. A conversation activity requires several separate audio files. A dictation requires careful pacing. If a teacher needs another version for a make-up assessment, the whole process begins again.

There is also the problem of variety. Students who always hear their own teacher become accustomed to one voice, one accent, and one speaking rhythm. Authentic materials provide variety, but it can be difficult to locate a recording that matches the vocabulary, topic, length, and proficiency level of a particular lesson.

Innovation Assessments now includes AI assistance for generating world language audio activities inside the Test and Convo applications.

Teachers can create listening-comprehension passages, traditional dictations, and simulated conversations in fifteen languages:

• Arabic
• Chinese
• English
• French
• German
• Greek
• Hebrew
• Hindi
• Italian
• Japanese
• Korean
• Latin
• Portuguese
• Russian
• Spanish

The tools are not intended to remove the teacher from assessment design. They are designed to shorten the distance between an instructional idea and a usable classroom activity.

AI Listening Comprehension

The Test application now includes an AI Listening Comprehension generator.

The teacher selects:

• The target language.
• A CEFR proficiency level from A1 through C2.
• The student age or educational level.
• A short, medium, or long passage.
• A voice.
• A multiple-choice or short-answer question.
• A topic or scenario.
• Any additional instructions.

A teacher might request an A2 French passage about ordering breakfast in a café, a B1 Spanish announcement about a delayed train, or an A1 German description of a family.

The AI produces a draft containing both the listening script and a question based upon it. A multiple-choice item includes four answer choices and a designated correct response. A short-answer item includes model answers that may later assist with scoring.

The teacher sees this material before the audio is created.

This review step is important. AI can produce useful drafts, but it does not know precisely what a particular class has studied. It may use vocabulary that is too advanced, introduce a regional expression the teacher has not taught, or misunderstand an important detail in the request. The teacher can edit the script, question, choices, and answers before selecting Generate Audio + Add Question.

The finished audio and question are then added directly to the test.

When the student reaches the listening item, the passage plays automatically according to the Test application’s listening workflow. The student answers the multiple-choice or short-answer question without needing to leave the assessment.

Traditional Dictation

Dictation occupies an interesting place in language instruction. It is among the oldest methods used in the language classroom and, when used appropriately, it remains a useful exercise in listening discrimination, spelling, accents, punctuation, and the relationship between spoken and written language.

It is also tedious to record properly.

The AI Dictation generator uses a traditional three-pass structure. The completed recording is planned to:

1. Read the passage naturally.
2. Repeat it more deliberately, with clear separation and spoken punctuation.
3. Read the complete passage naturally once more.

For several commonly taught languages, the system uses language-specific punctuation terms. A French dictation can say point, virgule, or point d’interrogation rather than inserting English punctuation instructions into the middle of the recording. Similar mappings are provided for Spanish, German, Italian, Portuguese, and English. Other supported languages use the general fallback behavior and therefore deserve especially careful teacher review.

The generated question asks students to write what they hear. The exact script becomes the primary model answer, while lightly normalized alternatives may also be included.

A teacher chooses the language, CEFR level, voice, passage length, age group, topic, and any special directions. As with listening comprehension, the complete draft can be edited before audio generation.

This makes it practical to create a short dictation aligned with the vocabulary of the current unit rather than searching for a preexisting recording that only approximately fits.

AI-Assisted Conversation Prompts

The Convo application presents a different problem.

Convo is designed to assess spontaneous speaking. The student hears one side of a conversation and records a response. The next prompt continues the situation until the student has completed a series of exchanges.

Preparing a good Convo requires more than writing several unrelated questions. The prompts must form a coherent interaction. They must provide enough context for a student to respond, but they must not supply the very content the student is expected to produce.

This last problem proved especially important during development.

An early AI-generated conversation about families in France included a prompt in which the conversation partner listed traditional, single-parent, and blended families. The corresponding student task was to name types of families. The audio had supplied the answer!

The generator has therefore been instructed to create only the conversation partner’s side. It must not state, preview, paraphrase, or provide examples of the student response being assessed.

The teacher describes the scenario and selects:

• One of the fifteen languages.
• A CEFR level from A1 through C2.
• A voice.
• Between two and eight prompts.

The AI creates a sequence of spoken turns forming one continuous conversation. These turns are not limited to direct questions. The conversation partner may make an observation, express a preference, offer an opinion, or describe a small problem that invites the student to react.

This makes the interaction sound more natural. Real conversations do not consist entirely of one person asking a list of interview questions.

Students Must Listen

Each generated Convo turn contains two different pieces of information.

The first is the actual spoken line the student hears. The second is a very brief visible cue such as:

• Respond naturally.
• React and explain.
• Answer and add one detail.
• Agree or disagree.
• Respond and ask a question.

The visible cue is intentionally vague. It should tell the student what kind of response to make without revealing the subject of the audio.

If the screen says, “Name three types of families in France,” the student does not need to understand the spoken French. The exercise has become a prepared speaking prompt rather than a listening-dependent conversation.

By using a cue such as “Answer with examples,” the student must comprehend the audio to know what examples are being requested.

This separation preserves the purpose of Convo: listening and responding in real time.

A Simulated Conversation, Not a Live AI Chat

The conversation is AI-generated, but it does not dynamically change according to what the student says.

The prompts are created in advance, reviewed by the teacher, converted to audio, and presented in a fixed sequence. The AI is not listening to the student and inventing the next turn during the assessment.

This is intentional.

A fixed sequence gives every student an equivalent task. It lets the teacher inspect the complete assessment beforehand. It also avoids the unpredictability, delay, and expense of running a live conversational AI during every student attempt.

The generator tries to maintain continuity without pretending to know what the student said. Later turns may use a content-free acknowledgment such as “I understand” or “That is interesting,” but they should not invent, summarize, or correct an unseen student response.

The result occupies a useful place between a disconnected list of speaking questions and a fully dynamic AI conversation.

Voice, Level, and Pacing

Both Test and Convo provide a selection of voice styles. The labels describe approximate personas—such as a calm adult voice, a warm male voice, or a polished female voice—rather than guaranteeing a particular regional identity.

Teachers can also adjust playback speed within a reasonable range. This can help match a recording to beginning or advanced learners without reducing speech to an unnatural crawl.

CEFR settings provide the AI with a useful target for vocabulary and sentence complexity. They should not be treated as an official certification that every generated sentence perfectly matches a proficiency level. As with all generated material, the teacher remains the final judge.

Pronunciation quality may also vary by language, name, regional expression, and selected voice. The fact that a language appears in the list means the system can be instructed to generate and speak it; it does not mean every voice will perform equally well in every language.

Listen before assigning!

Saving Preparation Time Without Surrendering Judgment

These tools perform several kinds of work at once. They can draft a passage, construct a question, produce answer choices or scoring models, and generate an audio file.

That is an impressive amount of assistance from a short teacher request.

It is also why review matters. An error in a private brainstorming response is inconvenient. An error converted into audio and placed on an assessment may confuse an entire class.

The workflow deliberately separates drafting from audio generation. Teachers can inspect and revise the material before using additional AI resources to create the recording. Generated questions and audio also become ordinary parts of the task afterward; the teacher can continue managing the assessment through the established Test or Convo tools.

AI usage is charged against the teacher’s available AI-token allowance. This gives subscribers control over how much generation they use and keeps the feature from silently producing unlimited external-service costs.

More Time for Designing the Assessment

The best use of artificial intelligence in education may not be to make instructional decisions for teachers. It may be to perform the mechanical work surrounding those decisions.

The teacher still decides what students should understand, which vocabulary belongs in the activity, how difficult the passage should be, what constitutes an acceptable answer, and whether the finished recording is appropriate.

The AI turns those decisions into a draft and a voice recording much faster than the traditional process.

For a language teacher who needs another listening passage, a carefully paced dictation, or five connected conversation prompts before tomorrow morning, that is no small improvement.

Authorship note: This feature announcement was generated by artificial intelligence using samples of David Jones’s blog writing as a stylistic guide. It was reviewed for consistency with the current Innovation Assessments Test and Convo AI-generation features.

For many years, I have made video lessons by recording voice-overs of PowerPoint presentations. I began doing this long before remote teaching made the practice commonplace. These lessons became an important part of my courses, but making them was not always a quick process. A teacher could spend a fair amount of time moving among presentation software, screen-recording software, video files, and a learning-management system before students ever saw the finished product.

Sometimes that degree of production is worthwhile. Other times, a teacher simply needs to explain something.

That is the purpose of the new Video Lesson application at Innovation Assessments. It gives teachers a convenient way to record a short lesson directly from their web browser and place it into a course alongside the other activities students already use.

A Small Recording Studio Inside the Course

A teacher begins by adding a Video Lesson to a module just as one would add a test, writing assignment, conversation, or other course element. The recording studio then provides access to the computer’s camera and microphone. Teachers can select the microphone they wish to use, record for up to ten minutes, review the result, and either save it or record the lesson again.

This is not intended to replace serious video-production software. Teachers making polished lectures for publication will probably continue to use programs such as OBS, QuickTime, or another video editor. Video Lesson is for the many occasions when speed and convenience matter more than elaborate production.

A teacher might use it to:

• Introduce a new module.
• Explain a difficult concept.
• Review a common misunderstanding.
• Provide directions for a complicated assignment.
• Record a short remedial tutorial for students who need another explanation.
• Preserve a useful classroom explanation for next year.

Once saved, the lesson remains hidden until the teacher is ready to publish it. It then appears in the course like any other assigned resource.

Camera Lessons and PowerPoint Lessons

The simplest option is a camera lesson. The teacher speaks directly to students, much as one might during an online meeting. On supported computers and browsers, the recording studio can also soften the background. This is useful for teachers recording at home or in a classroom where the background may be distracting.

There is also a presentation mode. A teacher may select a PowerPoint already uploaded to the course and advance through its slides while recording. The presentation occupies the main portion of the finished video while the teacher appears in a separate panel alongside it. This prevents the teacher’s camera image from covering material on the slide.

The teacher can make some adjustments to the presentation’s appearance, including its background and text display. PowerPoint is an elaborate format, so highly unusual fonts, complex animations, and intricate layouts may not reproduce exactly as they do in Microsoft PowerPoint. The purpose here is not perfect duplication. It is to provide a fast way to combine an existing classroom presentation with a teacher’s explanation.

This arrangement strikes me as a useful compromise. Students get both the visual organization of the presentation and the human presence of their teacher. The result feels more like a lesson and less like a silent collection of slides.

No Separate Video Service Required

The recording is created in the browser and saved directly to the teacher’s course. There is no need to upload it first to YouTube or another public video service. This keeps the workflow simple and allows teachers to create short materials specifically for their own students.

The application records in a compact web-video format. That makes it appropriate for quick instructional lessons while avoiding the very large files associated with high-resolution video production. As always, teachers should test their selected camera, microphone, and browser before recording an important lesson. Browser media features can differ somewhat among devices.

Some Evidence That Students Opened the Lesson

Assigning a video raises an old question: did the student actually watch it?

No online system can prove that a student was paying attention. A learner can play a video and think about something else just as easily as a learner can sit in a classroom and daydream through a lecture. It is important not to claim more from technology than it can actually provide.

Video Lesson does, however, record useful engagement evidence. Teachers can see such events as opening the lesson, beginning or pausing playback, reaching different points in the video, completing playback, and leaving or returning to the browser tab. The system then presents a human-readable summary while retaining the detailed activity records when closer review is necessary.

Staff members assigned to a shared course can also review this evidence for the students they support. Their access remains read-only and respects the teacher’s course permissions and the staff member’s selected student group.

These records should be treated as clues, not verdicts. They can help a teacher recognize that a student may have had trouble accessing a lesson, stopped partway through it, or left the page repeatedly. That information can begin a useful conversation.

Keeping the Tool Modest

One of my persistent goals in developing Innovation Assessments has been to build tools around actual classroom needs rather than adding complexity for its own sake. Video Lesson follows that philosophy.

It does not try to become a professional television studio. It offers a camera, a microphone, optional background softening, an existing PowerPoint, and a direct path into the course. That is enough to make many useful lessons.

A polished instructional video may take hours to create. A timely explanation should not have to. Sometimes a teacher needs to sit down, select a microphone, open a presentation, teach for five minutes, and give the result to students. Video Lesson was built for exactly that moment.

Authorship note: This feature announcement was generated by artificial intelligence using samples of David Jones’s blog writing as a stylistic guide. It was reviewed for consistency with the current Innovation Assessments Video Lesson feature.

Why We Design AI With Limits, Roles, and Instructional Purpose

One of the central ideas behind our platform is what we think of as bounded AI. In education, that matters enormously. The question is not simply whether AI is present in a tool, but how it is present. Is it open-ended, dominant, and difficult for a teacher to control? Or is it constrained by instructional purpose, teacher settings, and clear limits on what it is allowed to do?

Our view is that classroom AI should be bounded. It should serve the learning task rather than take it over. It should operate inside a framework defined by the teacher, the assignment, and the goals of the lesson. In practical terms, that means AI should not function as a free-floating substitute for instruction, nor should it become an unrestricted shortcut around student thinking. It should be structured, limited, and accountable.

That principle appears in several different ways across the platform. In some applications, teachers explicitly control how much student-facing AI is available by assigning a limited number of AI uses or “licenses” per student for a particular task. In grammar and writing workflows, for example, AI assistance is not simply switched on without limit. The teacher decides whether students receive access, how much access they receive, and when those counts should be reset or renewed. That matters because it keeps AI from becoming an ambient crutch. It remains a defined instructional support rather than an always-on replacement for effort.

Bounded AI also means constraining what the model is supposed to do. In the conversational tools, the AI is not treated as an unrestricted chatbot. It is given a teacher-defined topic, guidelines, and role, and it is instructed to stay within that frame. If a student tries to push the interaction off topic or get the AI to abandon its assigned role, the system is designed to redirect the exchange rather than reward the drift. In other words, the AI is not there to become anything the student wants it to be. It is there to support a particular kind of language practice under teacher-defined conditions.

That same logic extends into oral assessment. In the viva voce tools, the AI does not simply improvise a conversation however it wishes. It operates within a configured assessment structure. It is guided by the assigned topic, the intended proficiency band, the turn limit, and the instructional expectation that difficulty remain within a stable range. If a student struggles, the AI can narrow or rephrase. If a student is strong, it can deepen the probe. But it is not supposed to veer into a different topic, change the task, or suddenly raise or lower the level in a way that distorts the assessment. This is a very different educational use of AI from an open-ended chat experience. The AI is acting more like a constrained assessment instrument than a digital companion with no boundaries.

Bounded AI also means limiting the function itself. In our platform, AI is generally assigned a specific role: provide feedback on grammar, help a teacher score a rubric, summarize short responses, generate a draft activity, maintain a target-language interaction, or support a structured discussion. Those are narrow tasks. They are useful tasks. But they are not the same thing as handing over the intellectual work of the lesson to a general-purpose model. We think that distinction is one of the most important design choices in educational technology right now.

There is also a teacher-control dimension to bounded AI that is easy to overlook. AI can help generate drills, prompts, questions, and classroom materials, but those tools are still framed as teacher-facing publishing assistance, not as autonomous curriculum engines. The teacher remains the authorizing intelligence. AI speeds up drafting, variation, and differentiation, but it does not replace pedagogical judgment. Used well, this can feel less like surrendering instruction to AI and more like giving the teacher an on-demand assistant for producing customized materials.

The educational value of this approach is substantial. First, it helps preserve student thinking. A bounded AI tool can scaffold, redirect, or clarify without simply doing the work for the learner. Second, it keeps classroom tasks legible to the teacher. If the AI is operating inside a clear assignment structure, its effects are easier to evaluate and manage. Third, it makes misuse harder. Students are far more likely to offload cognition when AI is unrestricted, conversationally dominant, or available in unlimited ways. When AI is role-bound, topic-bound, use-limited, and embedded in task design, it becomes a support rather than an escape hatch.

Just as important, bounded AI supports better trust. Teachers are right to be cautious about tools that present AI as a kind of omniscient educational layer hovering over everything. That is not the philosophy here. Our model is closer to this: AI should enter the classroom with a job description. It should know why it is there, what it is allowed to do, what it is not allowed to do, and who remains in charge.

In the end, bounded AI is not a limitation in the negative sense. It is a design discipline. It reflects the belief that educational technology works best when it respects the shape of teaching rather than trying to dissolve it. AI can be useful, flexible, and powerful. But in a learning environment, its value increases when its role is defined, its scope is controlled, and its presence remains in service to human instruction.

How We Limit Data, Bound AI, and Reduce Unnecessary Student Exposure

As AI becomes more common in education, privacy deserves more than a reassuring slogan. Teachers and schools need to know, in practical terms, how a platform handles student information: what it stores, what it sends, and what it deliberately chooses not to include.

Our approach is guided by a simple principle: use only the information needed to support teaching and learning, and avoid unnecessary exposure wherever possible. That principle shapes both the way accounts are managed and the way AI features are built.

At the account level, we keep subscriber records focused on essential information. A functioning classroom platform does need core account data, enrollment relationships, and activity records tied to real users. But that does not mean the student record should become a warehouse of unnecessary personal detail. We aim to keep the data footprint as limited and purposeful as possible.

Student access is also designed with flexibility and restraint in mind. In many parts of the platform, students can sign in using a standard email-and-password account. Where appropriate, teachers can also enable a class-number-plus-PIN login option. This gives teachers another controlled way to bring students into classroom activities without making email-based login the only path. PIN access is not open-ended. It is teacher-enabled, tied to classroom enrollment, and limited to active student accounts. In some workflows, it is also paired with session-token checks and lightweight challenge steps before access is granted.

Traditional account security remains part of that design. Password-based logins are supported, and passwords are stored as hashed values rather than plain text. The goal is to support real classroom conditions while keeping access bounded and appropriately controlled.

The same privacy philosophy extends into AI use. In many classroom AI workflows, the model needs the student’s work, but not the student’s identity. A writing sample may need feedback. A conversation transcript may need to be scored. A set of short responses may need to be summarized. In those cases, the instructional content matters; the personal name usually does not.

For that reason, our AI integrations are designed to scrub identifying information. When student work is sent for AI-assisted analysis, the focus is on the work itself rather than on personal identity. In discussion and transcript-based tools, prompts can preserve structure without exposing names by using neutral labels such as “Student,” “Peer 1,” “Peer 2,” “Poster 1,” or “Poster 2.” That allows the model to follow turn-taking, compare responses, and interpret interaction without requiring unnecessary identifying detail.

This is an important distinction. Privacy in educational AI is not only about preventing unauthorized access. It is also about reducing unnecessary disclosure inside authorized systems. A feature may be legitimate and still contain more identifying information than it needs. Our design goal is to keep asking that question: what does the model actually need in order to do the instructional job well?

That mindset carries across the platform. We try to limit stored data to what is functionally necessary, offer bounded and teacher-controlled access options, and structure AI prompts so they carry instructional signal rather than avoidable personal detail. In our view, privacy is not a single feature. It is a design habit.

No platform should treat privacy as finished work. Systems evolve, features grow, and safeguards need to be revisited. But the standard remains clear: keep data collection purposeful, keep access controlled, and keep AI use as privacy-conscious as possible.

That is what privacy by design means in practice.

One of the most practical tools in Innovation Assessments is Convo, our speaking task app built for teachers who want students to respond to prompts in a more authentic, manageable, and scoreable way.

At its core, Convo is simple: the teacher creates a conversation task using a sequence of prompt audios, students listen and respond one prompt at a time, and the teacher can later monitor progress, review submissions, and score the work using either simple prompt-by-prompt scoring or a fuller rubric workflow. But what makes the app useful is how much classroom reality it accounts for.

A teacher begins by configuring the task. The setup is intentionally straightforward: give the task a title, add context or directions, set an overall time limit, and decide whether each prompt response should also have its own time cap. That matters in speaking assessment, because sometimes you want students to think and answer naturally, not rehearse for several minutes. Teachers can also decide how students will complete the task. There is a browser-recording version for direct in-app speaking, and there is also an upload version for cases where device compatibility or student circumstances make recording in-browser less reliable. If a teacher wants tighter control, the task can require in-browser recording so response timing is enforced more strictly.

The prompt-building process is also flexible. Teachers can upload prompt audio files or record prompt audio directly in the browser while building the task. Each prompt can include a memo or script, which helps keep the assessment organized and teacher-friendly. This makes Convo work well for world languages, oral interpretation, speaking checks, listening-response tasks, and even teacher-created mock interview activities.

On the student side, the experience is designed to be focused. Students open the task, hear the teacher’s audio prompt, and respond prompt by prompt. The app supports real classroom constraints: access and visibility checks, timing, saved progress, and submission tracking are all built into the workflow. Students who have already submitted are not accidentally allowed to start over unless the teacher readmits them. That matters because speaking tasks can otherwise become messy very quickly if students are unsure whether they are still “in progress” or already finished.

Another strength of Convo is that it does not pretend every device behaves the same way. The app supports both browser recording and upload-based response collection, which gives teachers a practical fallback path when needed. In real schools, that matters more than elegant theory. A speaking tool only works if students can actually complete the task on the devices they have.

From the teacher side, monitoring is lightweight and useful. The teacher can open the monitor view and see which students are in progress, how many prompts they have completed, and who may need a readmit. This is helpful during live class use, language lab work, remote learning, or make-up assignments. The monitor is not overloaded; it gives the teacher enough visibility to manage the task without turning into a distraction.

Scoring is where Convo becomes especially flexible. Some teachers want quick scoring by prompt, especially when they are listening for completion, clarity, or general performance. Others want a more formal evaluation process. Convo supports both. A teacher can score by individual response or switch into rubric-based scoring, depending on how the assessment is designed. That means the same app can support quick formative checks and more structured summative speaking assessments.

There is also a strong accountability layer behind the scenes. Convo includes proctor-style event logging, submission tracking, and workflow protections that help preserve the integrity of the task. That is particularly useful for graded speaking work, asynchronous assessment, and remote completion settings where teachers want a clearer record of how the task was completed.

What I like most about Convo is that it is not built around a fantasy classroom. It is built around the real one. Teachers need prompt audio options. Students need a focused workflow. Some devices cooperate; some do not. Some speaking tasks need strict time limits; some need flexibility. Some teachers want quick scoring; some want rubric-driven feedback. Convo makes room for all of that.

In short, Convo is a speaking assessment tool designed for actual classroom use: easy to configure, realistic for students, adaptable across devices, and strong on both monitoring and scoring. It helps teachers move beyond “just record something and upload it” toward a cleaner, more intentional speaking workflow.