繁中简中EN Become a partner

PRINCIPLES AND COMMITMENTS

Responsible AI

The stronger the technology, the more human agency matters.

This page states the beliefs, design commitments and rules TOIVO has set for its product.

Core beliefs

AI should extend people’s capacity to think, learn, connect and choose. Judgement, relationships and growth remain in human hands.

We design AI as learning scaffolding with clear limits:

  1. Helping students clarify the question; the answer is theirs to find
  2. Helping students compare information; the conclusion is theirs to judge
  3. Helping students take the first step; the task is theirs to complete
  4. Helping teachers see where students are stuck; the judgement stays with the teacher
  5. Helping parents see their child’s growth while respecting the child’s private space
  6. Helping principals drive school-based innovation, with every analysis explainable and open to checking

Seven commitments

These are the design commitments we have set for the product.

Students think first, then AI supports

Students should first form their own question, hypothesis, draft or idea. Only then does AI offer hints, examples, structure, fact-checking and feedback.

Teachers keep professional authority

AI can draft, organise, prompt and assist. Teaching, assessment, guidance, discipline and high-impact decisions always remain the responsibility of teachers and qualified professionals.

No labelling of students

TOIVO does not generate fixed labels such as “low-motivation student”, “problem student”, “not suited to this subject” or “high-risk personality”.

The system may only show learning patterns that can be explained and corrected, for example:

“In recent tasks the student has often stopped at the first step, and may need clearer task steps or success criteria.”

No mental health diagnosis or life prediction

TOIVO will not use a student’s conversations, assignments, clicks, login times, voice, images or emotional writing to diagnose anxiety, depression, ADHD, addiction, personality or family problems.

Design requirement: when a student says they are struggling, the platform offers ways to get help and refers them to a real person. It does not replace counselling, medical or clinical services.

Students own their work and what they share

Step by step, students can understand, manage, add to, correct and choose to share their own work, digital credentials, TOIVO Passport and growth records.

People review high-impact decisions

High-impact decisions on admissions, discipline, special educational needs, psychological support, financial aid, recruitment, allocation of opportunities or data sharing must have clearly defined permissions, human review and an appeal process.

No addictive design to drive usage

TOIVO does not compete for students’ attention with infinite scroll, late-night push notifications, login streaks, loot boxes or public rankings.

What we value:

  1. Completing meaningful tasks
  2. Knowing when to stop
  3. Returning to real relationships
  4. Offline observation, interviews, experiments, sport and service
  5. Sleep, recovery and a sustainable pace of learning

GOVERNANCE

Three ground rules.

Human review, data use and the protection of children and young people are preconditions for delivering the product. Every task must meet all three.

01 Human review The role of AI and how much help it gave are labelled. Decisions on assessment, promotion and retention, admissions, discipline and diagnosis are made by people.
02 Data use Only necessary data is collected. Showing a student’s work needs the student’s permission, and for minors a parent’s consent as well. Reports use aggregated data only, and personal data is never sold.
03 Protecting children and young people Tasks are graded by age and risk. High-risk tasks open only with parental consent, partner vetting and human support in place.

Certification plans

01 ISO/IEC 27001 information security management systems
02 ISO/IEC 42001 AI management systems

TECHNOLOGY PLATFORM

Designed and governed in Hong Kong.

TOIVO’s technology platform is designed on the four principles below.

Replaceable models
Controlled model routing chooses established model services task by task, balancing quality, cost and safety. The design allows providers to be changed to meet performance, data classification and regional requirements.
Isolated data
Access is managed through role-based permissions and data isolation. Sensitive data may not be entered into unapproved tools.
Traceable decisions
AI use, key outputs, source checks and the reasons for adopting an output are recorded, and tasks, work, revisions, feedback and capability are linked in one record. A student’s capability is shown through real evidence, and every judgement can be traced.
Human in the loop
AI can help produce suggestions, language scaffolding and first-round feedback. People make the decisions on issuing digital credentials, allocating opportunities, escalating risk and other high-impact judgements.

For the security position of this website itself, see Security.

DATA AND ACCESS

Data minimisation and role-based access

A design rule for the product: collect only the data needed for a specific educational, task, service or governance purpose.

DESIGN RULES

  1. AI may read and process only the data a task needs
  2. Schools, families and partners can see clearly what data is used for

Student

Can seeTheir own work, tasks, reflections, capabilities and sharing permissions

Cannot seeOther people’s personal data and assessments

Parent

Can seeGrowth outcomes the student chooses to share, support needs and resources for parent-child conversations

Cannot seePrivate reflections, emotion-related content, comparisons with other students

Teacher

Can seeTasks, work, revisions and class trends relevant to teaching and student support

Cannot seePrivate journals or sensitive data they are not authorised to see

Principal

Can seeAggregated trends in teaching, adoption, outcomes and governance

Cannot seeUnnecessary private content about individual students

Company or NGO partner

Can seeWork and Challenge outcomes that students have shared with explicit consent, for a specific purpose and period, plus anonymised insights

Cannot seeSensitive personal data, school records and private reflections

University

Can seeTOIVO Passports that students have chosen to share

Cannot seeData students have not agreed to share, and internal school data

Mentor or assessor

Can seeNecessary evidence for the students assigned to them

Cannot seeData on students not assigned to them. Data they view may not be freely downloaded or passed on.

Platform administrator

Can seeData needed for the job, under least privilege, separation of duties and audit logs

Cannot seeAnything beyond what the job needs

For what this website itself collects, see the Privacy policy and Children’s data.