AI & Data — what an Australian parent actually wants to know
Four questions. Plain answers. No jargon. The technical disclosure (which AI vendors we use, how anonymisation works) is at the bottom of the page in a collapsed section for anyone who wants the engineering detail.
1. Will my child get a wrong answer from the AI?
Sometimes — yes. AI systems get things wrong. Two safeguards make this less likely for an Edvora student than for a child using a generic chatbot:
- Every question has a static, human-reviewed correct answer and worked solution. The AI tutor explains why that answer is correct — it does not invent the answer. If the AI's explanation drifts from the static answer, the static answer is what the child sees as the source of truth.
- Every static solution is checked by our quality-gate suite (10 automated gates running on every push, covering structural, mathematical, rendering, answer-validation, and hint-vs-explanation consistency). The current bank scores a mean of 91.8 / 100 against those gates.
If your child sees a wrong answer in practice, please report it via Support — we triage every bug report and many template fixes have shipped this way.
2. Will my data — or my child's data — be sold?
No. We do not sell personal information to any third party. We do not run third-party advertising on the platform. We do not share account-linked progress data with anyone outside Edvora.
For the avoidance of doubt: third-party AI inference providers (used to generate explanations on demand) do receive the question and your child's answer — without any identifying information — to produce the explanation. Their contractual terms prevent them from using that data to train their own models. The full vendor list is in the technical disclosure at the bottom.
Full data-handling detail in the Privacy Policy. Account deletion removes your child's personally identifying information within 24 hours.
3. Can the AI replace a real tutor?
No, and that is not what Edvora is built to do.A human tutor adapts to your child's emotional state, asks follow-up questions, builds a relationship over weeks, and catches the social cues a software product cannot.
Edvora is built to do the part of tutoring that benefits most from automation: high-volume, format-accurate, individually-adaptive practice with on-demand explanations. It pairs well with a human tutor (the AI handles the drill volume; the tutor handles the conceptual gaps the AI surfaces) and it stands on its own when tutoring isn't financially or logistically possible.
4. What happens if the AI is wrong?
The student is not left to figure that out alone:
- The static worked-solution is always shown alongside the AI explanation. If they disagree, the static one is canonical.
- The AI explanation includes a disclaimer that it is for developmental feedback, not an official mark or score.
- The bug-report button is visible on every question. Reports go straight to the content team's triage queue.
- Our internal LLM re-derivation audit independently solves a sample of templates with a different model and surfaces disagreements for review (the y5_frac_div_groups bug was found and fixed this way).
The interactive demo & misconception classification
On a question your child gets wrong, Edvora identifies which specific reasoning error caused the slip — not just that the answer was incorrect. For example: a student who answers 1/3 + 1/4 = 2/7 has confused "add numerators, add denominators" with the correct common-denominator method. Edvora names that misconception and generates a targeted explanation for it.
See the walkthrough on the homepage for a live example. Misconception coverage is being rolled out template-by- template; items without a tagged misconception still receive a general step-by-step explanation on demand.
The rest of this page is the legal-grade data-handling detail. If you've read enough to feel comfortable, you don't need to read on.
What we collect when you practise
When you answer a question, we save:
- The question
- Your answer
- Whether it was correct
- How long you took
- Which topic it was
We also save a random ID so we can show you your progress. That ID is not your name or email.
Why we collect it
To help you learn (always)
So we can show you questions that match where you're at, keep your progress, and let your parent see how you're going.
To train our AI (only if you opt in)
Edvora is building an AI that's very good at understanding why students get things wrong — not just that they got it wrong. To do that, we need examples of real student answers. You can help by letting us use your wrong answers in our training data.
Once trained, the resulting fine-tuned model will power the Edvora platform, including both the free research program and the paid subscription tiers. Until then, the platform uses a foundation-model prompting pipeline (Groq, Google Gemini, Anthropic Claude) with our misconception taxonomy supplied in-context.
What goes in the training data
- The question
- Your answer
- The correct answer
- Which topic and year level
What never goes in the training data
- Your name
- Your email
- Your school
- Any free text you write (we do not collect free-text reasoning — only the pre-written chip buttons)
- Anything that would let someone figure out who you are
How anonymisation actually works
Your account and your training contributions are in different parts of our database. The link between them is a random pseudonym ID. When you delete your account or withdraw consent, the link is severed. After that, even we can no longer tell which contributions were yours.
Your choice
- You choose whether to help us train our AI when you create your account.
- You can change your mind any time in Settings → Privacy.
- If you say yes, you get full platform access free while you're helping us.
- If you say no, you can still use Edvora — there's a standard paid plan at $19/month.
Under 16?
A parent or legal guardian must create the account. Parents decide on AI training consent on behalf of the child. Both can be withdrawn any time.
Your rights
You can always:
- See everything we have about you (Settings → Privacy → Download my data)
- Change or delete it
- Ask us to stop using it for AI training
- Close your account entirely
Email support@edvora.com.au or use the Settings page.
Technical detail — which AI providers we use
When we generate an explanation or classify a misconception we may send the question and your answer (no identifying information) to:
- Groq (primary, for speed)
- Google Gemini (fallback)
- Anthropic Claude (fallback)
These are industry-standard AI inference providers. Their contractual terms prevent them from using the data we send them to train their own models.
Questions?
support@edvora.com.au
Full detail in the Privacy Policy and Terms of Use.