Responsible AI
AI that supports human decisions in education
Kalonesis is designed to support human decision-making in education. Organizations define policies, roles, review steps, and the boundaries within which AI agents operate.
Our position
AI should not replace teachers, trainers, instructional designers, or learning professionals. It should help them understand learners more deeply, personalize support, reduce repetitive work, identify the right next action, intervene earlier, and make decisions using observable evidence — while human judgment and responsibility stay in control.
Our principles
- 01
Human oversight
People decide where agents may act autonomously, where review is required, and who is accountable for the outcome.
- 02
Purpose limitation
Learner context is used to support learning, and configured for that purpose rather than repurposed silently.
- 03
Data minimization
Agents work with the context a task actually requires, not with everything available about a person.
- 04
Role-based access
What a person or an agent can see and do follows from their role, not from convenience.
- 05
Approved knowledge sources
Organizations decide which material agents may ground their answers in, and can restrict them to it.
- 06
Transparency and traceability
Agent activity, the context used, the workflow step executed, and the outcome are recorded and reviewable.
- 07
Monitoring and review
Prompts, models, quality, and outcomes are monitored so that problems are found by the organization rather than by a learner.
- 08
Safety and escalation
Defined paths route a situation to a person when it falls outside what an agent should handle.
- 09
Continuous quality improvement
Observed outcomes feed back into pathways, prompts, policies, and platform behaviour.
- 10
Respect for educators and learners
The platform is built to extend professional judgment and to treat learners as people, not as throughput.
What we do not claim
We do not claim guaranteed learning outcomes, perfect personalization, complete accuracy, zero hallucinations, fully unbiased AI, or total regulatory compliance. Explainability has real limits, and a Learning Twin is an evidence-based model of a learner rather than a complete or infallible one. We would rather state those limits plainly than have you discover them later.
Certifications and compliance
We publish certifications, compliance statuses, and regulatory approvals only once they have been formally obtained and documented. Where this page does not name one, it is because we do not yet hold it. Kalonesis provides features that support your compliance work; the compliance obligation itself remains with your organization.