Switch course ▾
Session 4. Risk and Ethics in Educational Innovation
Objective. Build a risk register for a school pilot, apply an ethics check covering consent, privacy, equity and the child's best interest, question vendor claims, and set red lines for your school.
A pilot that harms one child is not a small success with a side effect; it is a failure that no gain can pay for. Leaders carry a duty that individual teachers often do not see: they approve pilots, sign agreements and decide what leaves the school. This session gives you practical tools for that duty: a risk register that starts from the child, an ethics check that fits a school, simple rules for pupils' information, and the habit of turning every claim into a question.
Key ideas
A risk register in one table
- A risk register lists what could go wrong, how likely it is (low, medium, high), how serious the effect would be on pupils (low, medium, high), who watches it and what you will do. Look first at the impact on pupils: a rare risk with a serious impact on a child outranks a common small one.
- Check seven kinds of risk: learning (lost time, weakened standard), safeguarding (a child's safety and dignity), privacy and data, equity (who is left out: pupils with no internet at home, girls or boys, rural pupils, pupils with disabilities, speakers of a minority language), staff workload and wellbeing, money and dependency on one supplier, and reputation with parents.
- For each risk choose one response: REDUCE (change the design), PREPARE (a plan B, such as a paper version), ACCEPT (write down why it is small) or AVOID (do not do it). Review the register at every gate from session 2, not once.
Ethics of trying things with children
- The British Educational Research Association's ethical guidelines, in plain words, ask researchers to respect everyone involved, to seek voluntary informed consent, to let people withdraw without penalty, to avoid harm, to protect privacy and confidentiality, and to report honestly. Improving your own teaching is not the same as publishable research, but the same care is a good habit for any school pilot.
- The UN Convention on the Rights of the Child gives the frame: the child's best interests come first (Article 3), the child's views should be heard (Article 12) and the child's privacy is protected (Article 16). In practice: tell parents what you will do and why, let pupils say what they think, and never let a pilot cost a child marks or dignity.
- Use a simple school ethics check before any pilot: Is it likely to help, or at least not harm? Do parents and pupils know, in words they understand? Can a pupil step out without a penalty? Who is left out? What information leaves the classroom, and who sees it? Who can stop the pilot at once? Follow your school's and ministry's rules on informing parents.
Data in plain words
- Five habits protect pupils' information: collect the minimum needed; say why; keep it safe (a locked cupboard for paper, a password for files); share only for the stated purpose; delete it when the purpose ends. Remove names and use codes when you analyse results. Photos, videos and voices count as personal information; check your school's rules and parents' permission first.
- Follow the national personal-data rules and ministry regulations that apply; ask your district office or the school's responsible person if unsure. When an outside tool or partner holds pupils' data, ask where it is stored, who can see it, how long it is kept and how to have it deleted.
- Equity belongs to ethics. If a pilot works only for pupils with a phone, a quiet room or a parent at home, it widens gaps. Always keep a no-technology route so that no pupil is excluded.
Questioning vendor and AI claims, and setting red lines
- Marketing is not evidence. Turn each claim into a question: used for what, by whom, in schools like ours? Compared with what? Who measured, and how many pupils? What data does it need, and where does it go? Free for schools: how does the company earn money? A good supplier answers in writing and shows limits as well as strengths.
- For AI tools, UNESCO's guidance on generative AI in education points to data-privacy protection, possible age limits and keeping teaching human-centred, and UNICEF's guidance on AI and children asks for systems that uphold children's rights, safety and inclusion. Judgement stays with the teacher; the strategy for AI follows in course 14.6 (Strategic Use of AI) and governance in course 15.5.
- Write red lines in advance, so nobody decides under pressure: STOP AT ONCE if a child is harmed or frightened, if consent or a school rule is broken, if pupils' data leaks, or if a pupil's marks in an exam year suffer. ADJUST if the risk is medium and a fix exists. GO ON if the risks are low and watched. Name who has the power to stop.
A risk cycle for every pilot
Consent when the person asking is the teacher
A pupil finds it hard to say no to a teacher or a head. So consent in a school must be built, not assumed. Give pupils a real alternative that carries no penalty (a different task, a blank form), make sensitive questions anonymous, say plainly who will read the answers and what will be done with them, and let pupils stop at any time. For younger children, tell parents and listen to the child's own reaction as well.
Informing parents is part of the same respect. A short note in plain language, in Uzbek or Russian as the family needs, saying what changes, why it is tried, what is collected and how to ask questions, avoids most conflicts. Rules on informing parents differ; follow your school's and ministry's regulations and ask your district office when unsure.
From claim to evidence: a ladder for vendor and AI statements
Vendor statements sit on different rungs of evidence. At the bottom are testimonials and 'used by thousands of schools'. Higher are the company's own studies, then independent studies with a comparison group, then results repeated in several places. Ask which rung a claim stands on, and who paid for it. A claim of '30 percent improvement' without the comparison, the number of pupils and the method has no meaning for your school.
AI tools need extra questions. They can produce confident answers that are wrong, and they may work less well in Uzbek or Russian than in English, so test them with your own examples before using them with pupils. Ask what data they store, whether pupils' texts are used to train the tool, and whether you can switch it off without losing anything. The teacher keeps the judgement; the tool only suggests.
Examples from the classroom
The survey with names in Namangan
Scene. Jasur's questionnaire on home life has pupils' names on top and is read aloud at the staff meeting. A mother complains. The head and Jasur rebuild it: no names, codes on the forms, only Jasur sees them, questions about money removed, an option to leave the form blank, and a note to parents. Results are shared only as counts: 12 of 34 pupils say they have no quiet place to study at home.
Insight. The same question can be asked ethically when names, audience and purpose are changed.
AI feedback with practice texts in Samarkand
Scene. A Grade 8 Uzbek teacher wants pupils to get feedback from an AI tool. The head agrees only on conditions: pupils use practice paragraphs written by the teacher, not their own work; no names; the teacher reads the tool's feedback first and corrects mistakes in Uzbek grammar; a paper route exists for all; parents are told. After four weeks the teacher reports that the tool made three errors in 20 comments.
Insight. A pilot with an AI tool can be safe when data is minimal, a human checks output and there is a no-technology route.
Turning a claim into questions in Tashkent
Scene. A supplier tells a Tashkent school that its tool 'improves reading by 30 percent'. Dilnoza writes: Compared with what? In which schools and for how many pupils? Who measured? What pupil data is stored and where? How can we delete it? The supplier replies that the figure comes from its own survey of 40 teachers. Dilnoza records the answer, and the school agrees only to a two-class trial with no personal accounts.
Insight. Turning a claim into specific questions shows what rung of evidence it really stands on.
Common pitfalls
- Rating risks by how often they happen and forgetting the impact on one child.
- Treating a pupil's silence or a signed form as proof of free consent.
- Sharing pupils' photos, marks or free-text answers in staff meetings or chats.
- Accepting 'used worldwide' or 'approved by experts' as if it were evidence for your pupils.
Risk, Ethics and Red Lines
- Pairs (15 min): in Table A, three sample pilots are planned. For each, find at least four risks across the seven kinds, rate likelihood and impact on pupils, and choose reduce, prepare, accept or avoid. Underline the risk you would handle first.
- Pairs (15 min): in Table B, apply the six-question school ethics check to the university questionnaire request. Decide yes, yes with changes, or no, and write the changes. Draft a three-line notice for parents.
- Pairs (10 min): in Table C, turn five sample vendor claims into questions you would ask in writing, and say what a satisfying answer would look like.
- Part B, alone (25 min): in Table D write a one-page ethics and risk protocol for pilots in your school: the ethics check, red lines, who may stop a pilot, data rules for pupils' information, and a short parent notice for one real pilot.
- Part B, trios (15 min): present the protocol in 3 minutes each. The others ask: What if the head wants to continue and a red line is crossed? Who checks equity? What does the parent notice not say? Revise two lines.
- Alone (5 min): write your leadership task from the Kit.
Your facilitator will guide you through this activity in the session.
Questions to discuss
- Which risk did you first think was small and then rated serious when you looked at the impact on one child?
- Which red line would be hardest to hold when a result looks promising?
- Who in your school can say 'stop', and do they know it?
Reflect
- Has your school ever shared a pupil photo or mark more widely than parents expected?
- Which pupils would be left out by your current favourite innovation?
- What claim have you accepted from a supplier without asking for evidence?
Try it tomorrow
- Write a seven-kind risk register for one current pilot and rate each risk by impact on a single pupil.
- Draft a three-line parent notice for a pilot: what changes, why, what is collected, whom to ask.
- Choose one supplier claim you have heard and write three written questions about it.
Further reading and links
Rate risk by what it could do to one child, check consent, privacy and equity before you start, question every claim, and write the red lines before you need them.