An AI learning system can recommend a task, evaluate an answer and suggest the next step. But who decides whether its recommendation is appropriate? Who can question it, change it or explain its consequences to the learner?
These are useful questions to bring to OEB Global 2026 in Berlin. The conference’s announced theme, “Sovereign Learning: Trust, Agency & the Education Reset,” connects digital learning with control, responsibility and institutional purpose.
OEB is scheduled for December 2–4, 2026, at the InterContinental Berlin. This article is a preview based on public information checked on October 27. It examines the questions raised by the announced theme and selected programme material, rather than reporting conclusions from an event that has not yet taken place.
Why OEB’s 2026 theme deserves a practical reading
OEB brings together perspectives from education, workplace learning and the public sector. Its theme asks how institutions can use increasingly capable technologies while maintaining human judgement and direction.
The organiser’s subthemes connect institutional structures, technology implementation and human agency. That scope is important: a learner’s experience is shaped both by a particular activity and by decisions made elsewhere about platforms, assessment and available support.
Yet a theme remains a starting point. A declaration that learning should be human-centred does not reveal what an individual can do when a system makes an unsuitable recommendation.
A practical reading therefore begins with decisions. Who sets the goal? Who interprets the evidence? Who may override a recommendation? Who is responsible for providing an alternative?
Those questions turn an appealing principle into something an institution can examine.
The AI-teacher debate is a proposition to interrogate
The published conference highlights announce an annual debate about assigning AI the principal teaching role while humans act as coaches. It is a debate motion, not an established conclusion or an endorsement that should be attributed to every participant.
The proposition is useful because it makes assumptions about teaching visible. What counts as instruction? Is explaining content the same as deciding what a learner needs? How are subject expertise, relationships, motivation and assessment connected?
A simple division between machine instruction and human encouragement can conceal difficult professional work. An educator may need to recognise that a learner’s apparently minor error reflects a deeper misunderstanding. They may need to change the task, explain why an answer is inadequate or reconsider whether the assessment is fair.
Equally, technology can make useful support available in situations where it was previously difficult to obtain. That possibility deserves examination without assuming that access to an explanation is equivalent to educational judgement.
The most productive response to the motion is to ask which responsibilities are being redistributed, under what conditions, and with what evidence. The label “coach” does not settle what a human professional must understand or be able to do.
Agency operates at more than one level
An institution can offer learners choices while leaving educators little freedom to alter the system. It can give teachers an override button while providing no time or information to use it meaningfully.
For discussion purposes, it helps to separate three levels of agency.
| Level | A decision to examine | A practical question |
|---|---|---|
| Learner | Responding to a suggested task or explanation | Can the learner ask why, request clarification or seek another route? |
| Educator | Interpreting progress and adapting support | Can the professional inspect relevant evidence and change the response? |
| Institution | Defining the learning system’s purpose and limits | Can the institution revise its arrangements when they no longer serve its educational goals? |
This is an editorial framework, not an official OEB classification or a validated measurement scale.
The levels are connected. A learner may technically be able to choose another task but not understand the consequences. A teacher may be able to challenge a recommendation but lack access to the information behind it.
Agency therefore concerns the conditions for making an informed decision as well as the formal permission to make it.
A hypothetical learning pathway shows where control can disappear
Consider a university introducing an AI-supported pathway for an introductory course. The system recommends additional exercises after short assessments. Students can ask for explanations, and staff receive progress summaries.
At first glance, the roles seem clear: the system adapts practice, students work through it and tutors review progress. The difficult questions arise when the pathway does not fit a particular learner.
Suppose a student repeatedly receives elementary exercises because their short written answers do not match the system’s expected formulation. The student may understand the concept but express it differently. Alternatively, the answers may reveal a genuine gap. The summary alone cannot settle the matter.
Can the student ask for review? Can a tutor inspect the original work rather than only a category assigned by the system? Can the tutor change the pathway and record why? Will that intervention influence later recommendations?
The example is hypothetical. It illustrates the decisions an institution needs to investigate before claiming that a system strengthens learner agency.
A system can provide more personalised activity while still narrowing the learner’s ability to question how their progress is interpreted. The relevant test concerns the whole interaction, including what happens when the recommendation is challenged.
Trust requires a response when the system is wrong
Trustworthy learning arrangements need a workable process for disagreement. A polished explanation of how a system usually functions may provide little help to someone facing an error now.
Institutions can examine that process through ordinary cases. A learner receives confusing feedback. A teacher finds an inappropriate task. A progress summary conflicts with what a tutor observed. Who receives the concern, what evidence is available and what can happen next?
The aim is to avoid making the human role merely ceremonial. If a professional is named as the final decision-maker, they need the relevant authority, understanding and opportunity to act.
This also changes how success is described. Rapid resolution may be useful, but an institution should examine whether the resolution addressed the educational problem. Replacing a task without understanding the learner’s difficulty can leave the original issue untouched.
Learning Coach’s guide to moving from AI policy to classroom practice explores the connection between broad principles, recurring decisions and evidence for review.
Teacher capability remains part of the design
UNESCO’s 2024 AI competency framework for teachers provides a reference beyond the event organiser’s own material. It connects a human-centred approach with ethics, AI knowledge, pedagogy and professional learning.
This is a competency framework, not experimental proof that a particular system improves learning. It is useful here because it treats professional preparation as broader than knowing which buttons to press.
An educator reviewing an AI-supported task needs to connect its technical behaviour with its educational purpose. They need to ask what the learner should contribute, what counts as an adequate explanation and when a tool’s assistance undermines the intended activity.
That is particularly important in assessment. A convincing output may provide weak evidence of the learner’s understanding. Learning Coach’s article on checking understanding when AI gives the right answer examines that distinction.
An institution cannot reasonably give staff responsibility for such decisions while treating their preparation as optional. Professional agency depends partly on the capability and support needed to exercise it.
Institutional autonomy becomes concrete in everyday choices
“Sovereign learning” can sound like a distant policy concept. Its practical meaning becomes clearer when an institution asks what it would do if a system’s priorities diverged from its own.
Could it change a learning sequence? Could it assess the quality of feedback using its own educational criteria? Could it continue supporting learners if a feature changed or became unavailable? Could staff understand which decisions they still own?
These are planning questions, not a statement about any particular provider’s contractual or technical arrangements. The answers need to be established for the actual system under consideration.
They also reveal why an educational decision cannot be reduced to a feature comparison. A product may offer attractive functionality while requiring organisational changes that the institution has not yet considered.
A useful conference conversation can help leaders identify those dependencies early. The next step is to examine them with the people who know the institution’s teaching, operations and learner needs.
Cross-sector exchange needs careful translation
OEB’s combination of education and workplace perspectives can expose assumptions that remain invisible within one sector. It also requires participants to examine where an analogy stops working.
A workplace assistant may appropriately guide a trained professional through a rarely used procedure. A university exercise may deliberately ask a novice to work through the reasoning themselves. A school may need additional safeguards and support because of learners’ age and dependence on adults.
The same technical capability can therefore serve different educational purposes. An account of success in one setting should prompt questions about learners, goals and conditions before it becomes a model for another.
For visitors, a useful record from a session might identify the decision being discussed, the original context and the conditions that would need checking locally. That is more informative than collecting a list of technologies to adopt.
What organisers can learn from the debate format
A clear proposition can focus a conference discussion. Its value depends on whether participants can examine the assumptions behind it and hear credible challenges.
For an event about AI and education, organisers can ask speakers to address concrete cases rather than only broad predictions. A panel might consider who should respond to an inaccurate progress judgement. A workshop might ask participants to design a review process. A keynote can establish why those choices matter.
The formats work together when the same question travels through them. Audiences should be able to connect an idea discussed on stage with a decision their institution faces.
Teams preparing their own education or L&D event can explore Learning Coach keynotes to frame a conversation about AI, independent thinking and human responsibility.
What prospective participants should verify
The official participation page separates the December 2 pre-conference workshops from the main conference on December 3–4. Workshop arrangements and any additional access requirements should be checked against the current programme.
The public agenda uses an embedded event platform and states that the schedule may change. The research for this article did not establish a complete, stable session-by-session itinerary. Visitors should use the official programme and participation information for current details.
The sources establish the announced event, its framing and selected planned content. They do not establish what participants will conclude or whether attendance will change practice.
The central question remains useful before and after Berlin: when technology helps direct learning, who retains the knowledge, authority and support to question that direction? An institution’s answer will be visible in its everyday decisions.
References
OEB Global — official website. 2026 theme checked October 27, 2026.
OEB participation information. Dates, venue and conference structure checked October 27, 2026.
OEB conference highlights. Announced annual debate and selected programme material.
OEB 2026 subthemes. Organiser’s thematic framework; not independent research findings.
OEB programme. Dynamic agenda; subject to change.
AI competency framework for teachers. Fengchun Miao and Mutlu Cukurova, UNESCO, 2024; publication page updated January 16, 2026. Professional competency reference, not an evaluation of OEB.
