Educational Companionship

1. Context

The way we institutionalize education increasingly fails to serve individual learners. Standards-based, didactic systems rarely have the one-on-one bandwidth to develop the skills that most shape a student's future — executive functioning, self-awareness, and the ability to ask good questions. Maslo's empathetic-companion platform offered a way to bring genuinely personalized support into that gap.

I led the design of how a Maslo companion would work in a classroom: a personified AI that learns through interaction with a student and, over time, becomes a real learning companion rather than a content dispenser.

Maslo, the empathetic AI companion, learns by interaction with its partner — the student. The more interaction, the more personalized the experience for the user.

— Foundations in Educational Companionship

2. Problem & Constraints

Design problem: How might an AI companion personalize learning support across executive, social, and academic skills — without taking agency away from the student or trust away from the teacher?

• Agency is sacred. If the companion becomes powerful enough to build the student's environment for them, the student loses agency and the tool becomes a curator, not a companion.
• Privacy by design. What a student shares with their companion must stay private; only keywords, subjects, or generalized statistics should ever reach a teacher.
• Human escalation for safety. Certain phrases or behaviors — especially mental-health signals — must be able to alert a teacher, parent, or school.
• Serve three stakeholders. The same system had to work for students, teachers, and administrators, with different data and different needs at each level.

Maslo cannot be overtly autonomous. If Maslo can harness and interpret the amount of data and then create environments for the student, then the student loses agency and Maslo becomes not a companion on a journey but a third-party curator. Every interaction was designed to prompt self-initiation, not to act on the student's behalf.

— The constraint that defined the design

3. Method Rationale

Scaffolded relationship over one-shot help

Rather than answering isolated questions, the companion builds rapport gradually, deepening personalization as the student matures — a cooperative cycle of growth modeled on how a good mentor relationship actually develops.

Computational thinking + Socratic learning as the pedagogy

I paired two traditions to define how the companion teaches. Computational thinking supplies a systematic process of analysis, experiential problem-solving, and reflection; Socratic questioning supplies the inquiry-driven 'Why?'. Together they form a cross-curricular way of approaching problems, not a STEM-only skill — and they map cleanly onto an AI companion whose core move is a well-timed question.

By teaching students (and teachers) the importance of “Why?” and by giving them the systems to seek out that knowledge, the results — regardless of specific content — reflect the interests of the student and allow for more personalized learning experiences.

— Computational Thinking and Socratic Learning

4. Process: The Interaction Model

The companion aggregates signal well beyond the journaling conversation — pulling from the platforms a student already uses so it can support the whole learner:

A worked example shows the model in action: during a Monday homeroom reflection, the companion notices from the calendar that a Social Studies test is on Friday, and instead of doing the work, it prompts the student toward several options to prepare — nudging self-initiation. Intervention level is adjustable, so a student with ADHD or an executive-function disorder can receive routine reminders and long-term project scaffolding tuned to their needs.

5. Synthesis: Three Stakeholders, One System

• The student's companion fosters executive functioning, soft skills, and academic skills through focused, one-on-one attention no classroom can otherwise provide.
• The teacher's companion serves as a self-care and self-awareness tool against burnout, and lets teachers push questions or polls to students and pull anonymized classroom trends to target competency gaps.
• The administrator's view maps anonymous trends across a teaching community to guide professional development — without exposing any individual's private content.

The data-sharing model is the connective tissue: private by default, generalized when shared, and escalating to a human when safety requires it.

6. Key Design Decisions

• Design for prompting, not doing. The companion's job is to create opportunities for the student to self-initiate — the clearest expression of the agency constraint.
• Whole-learner data, narrow-purpose use. Aggregating many streams enables genuine personalization; strict privacy rules keep that aggregation from becoming surveillance.
• Personalization scales differently than content. The value grows from the depth of the relationship over time, not from more material delivered faster.
• Cross-curricular by construction. Grounding the pedagogy in computational thinking and Socratic inquiry makes the companion useful in the humanities and arts, not just STEM.

7. Outcome

The work produced a coherent design framework — an interaction model, a multi-stakeholder data-sharing model, and a pedagogical foundation — for deploying empathetic AI in schools as a step toward genuine personalization. It positioned Maslo's education product around a defensible claim: that a companion AI can help a student succeed academically while, more importantly, becoming more self-aware and more empathetic toward the world around them.

Maslo represents a step towards genuine personalization. A companion AI would allow a student to succeed academically but more importantly to be more self-aware and empathetic to the world around them.

— Foundations in Educational Companionship

8. Reflection

Designing this framework sharpened a conviction I carry into any product decision: the most important constraint is often the one that limits what the technology is allowed to do on the user's behalf. Holding the line on student agency made the design harder — it ruled out the flashiest automation — but it is what makes the companion trustworthy in a classroom. If I were to validate it next, I would run classroom pilots measuring whether scaffolded prompting actually increases self-initiation over time, and whether the anonymized teacher view changes instructional decisions. Those outcome measures are where this framework would prove — or revise — its claims.