KakiPal

Companion system · In pilot

KakiPal

Knows how you're doing. Cares how you're feeling.

A personalized, proactive companion for university courses — closing the loop between academic performance and emotional state, and telling teachers when, where, and how to step in.

What It Does

One system, both sides of the loop

Serving teachers and students at once, so learning data actually flows between both sides.

For teachers

The concept map builds itself

Upload slides and quizzes; the system extracts the concept map and tags every question. Teachers review and adjust instead of labeling by hand.

Alerts worth acting on

Every alert ships with a suggested action; volume and precision are controlled to prevent alert fatigue.

Sample alerts

High priority

Concept "Visual Encoding": 58% of the class missed Q4 — three misconceptions identified; suggest revisiting in the next lecture

Student Y failed three quizzes on the concept "Chart Selection"; suggest reaching out

For students

Socratic guidance

Heuristic questions lead students to the answer instead of giving it away; every response cites the exact slide, so answers are verifiable.

Personalized practice

A per-concept mastery model drives question selection: weak concepts get reinforcement, strong ones get advancement.

Lightweight emotional support

When frustration or confusion is detected, it adjusts its approach, offers encouragement, or suggests pacing changes.

Design Principles

Two design principles

Personalized

Practice, explanations, and encouragement are generated from each student's own mastery model and emotional state — tailored to the individual, not the class average.

Proactive

It doesn't wait to be asked: weak concepts reach students before the exam, and struggling students reach teachers before problems escalate — prevention instead of firefighting.

Affect Sensing

Reads the room — without cameras

No cameras, no biometric devices. Three compliant, low-friction signals do the job.

Self-reported mood

😣

😕

🙂

😄

A quick mood check before and after each session: the delta shows the session's effect, the trend shows the trajectory. Fully voluntary and fully visible to the student.

Language signals

"I still don't understand this part…"

Distress phrasing, repeated rephrasing, and mid-problem abandonment in tutoring conversations support non-intrusive affect inference.

Behavioral signals

Abnormal answer speed, long idle gaps, and repeated retries map to confusion, frustration, boredom, and flow — the states that matter for learning.

No cameras

No biometrics

No face or voice emotion recognition

Mastery × Mood

Same struggle, different fix

Each quadrant routes to whoever is best placed to act — the student side adapts automatically, or the teacher is alerted with context.

High mastery

Quietly burning out

High mastery · Low mood

Emotional check-ins

Pacing adjustments

Thriving explorer

High mastery · Feeling good

Stretch challenges

Keep the flow going

Low mastery

Needs attention now

Low mastery · Low mood

Teacher proactively alerted

With full learning context

Confident beginner

Low mastery · Feeling good

Targeted practice

Fresh explanations

← Low mood

Feeling good →

Ethics & Privacy

Three principles the system is built on; the escalation path is settled with each institution before rollout.

Fully transparent

Students see exactly what teachers can see about them; mood check-ins are voluntary and can be turned off at any time.

A defined escalation path

Teacher outreach or counseling referral — the escalation path is agreed with the institution before deployment.

Data minimization

Only what the features require; no biometric data, no third-party sharing.

Rollout

Built. Heading into the classroom.

Done

System built

Fully working, end to end

In progress

First pilot

One semester in a real course

More courses

Pilot evidence recruits instructors; onboarding stays light

Planned

LMS integration

Connects to campus LMS via the LTI standard

Evaluation

How we'll know it works

At the end of the pilot, these three questions decide whether it earns a place in the classroom.

Learning

Quiz-score gains on concepts the system flagged as weak.

Teacher value

Alert precision — whether teachers act on alerts, and how useful they rate them.

Experience

Self-reported sense of support and engagement.

KakiPal

Intelligent learning companion

Pilot stage — running in one real course