At aihorizon R&D, we’re introducing Adaptive AI Tutors — a new capability in Horizon, our responsible AI learning environment, designed to deliver personalized, evolving instruction for every learner.

This marks the first experimental phase of Adaptive Tutors. As we explore and refine these systems, we are actively evaluating our approach — from how we define adaptation rules, to how we assess learning progress, to how we manage known challenges like hallucination risks across different foundation models. This is a phase of critical reflection and iteration, where the feedback and insight of early testers will be essential in shaping the path forward.

These AI-powered tutors don’t just provide content — they build on your learning history, adapt to your progress, and tailor instruction to match your goals and preferences over time.

​

Built to Learn With You

 

Each session starts with context — not from scratch. The system draws on prior sessions to assess:

• Learning scores and performance history

• Patterns in comprehension, struggle, and success

• Preferred engagement and interaction style

• Feedback, motivation levels, and responsiveness to methods

 

From there, the tutor adjusts both what it teaches and how it teaches it — creating a dynamic, continuously personalized learning path.

 

Learn in the Way That Works Best for You

​

Adaptive Tutors support a range of pedagogical styles. You can select a preferred approach, or let the system identify the best fit as it learns from your interaction:

• Socratic Method – Guided questioning and critical thinking

• Guided Discovery – Learning through exploration and scaffolding

• Direct Instruction – Clear, structured content delivery

• Montessori Style – Autonomy-driven, self-paced learning

• Feynman Technique – Deep understanding through teaching concepts back

 

Tutors can shift methods as needed — for example, beginning with structure and transitioning to more learner-driven approaches — based on your evolving needs.

 

Personalization Features

Adaptive Tutors can be configured through a combination of static settings and conversational interaction, offering both flexibility and control. Customization includes:

• Learning goals and subject domains

• Current skill level and assessments

• Preferred learning style(s) and cognitive strategies

• Feedback preferences and motivation patterns

• Accessibility needs (e.g., visual, auditory, interaction speed)

• Contextual information (e.g., prior knowledge, curriculum alignment)

• Custom instructions and updates over time

 

No two learners are the same — your tutor reflects that from the start.

 

Ethically Grounded, Transparently Tracked

True to our mission, Adaptive AI Tutors are closely monitored and improved through our Responsible AI Coach — an internal oversight system ensuring ethical, fair, and inclusive use of AI agents.

​

In addition, every use of the tutors contributes to our Sustainability Module, which provides insights into CO₂ emissions, compute resource usage, and cost efficiency. This level of transparency aligns with our broader commitment to Responsible AI: AI that’s not only effective, but accountable.

Adaptive AI Tutors in Horizon combine personalization, pedagogical flexibility, and responsible design — offering a more meaningful and efficient way to support learning at scale.

​

Contributors: Dr. Michael Jülich (Conceptual Lead), Moritz Goeke (Technical Lead), Leon Kollofrath (Design & Engineering)