✦AI-Behaviour prevention tool for K-12 schools

PROJECT NOTE

This is a personal project aiming to explore how AI could support both prevention and intervention strategies inside a school management systems.

ROLE

Scope
Research & Discovery - UX/UI
Tools
ChatGPT, Cursor & Figma

PROBLEM

K–12 schools are overwhelmed not by severe incidents, but by the accumulation of low-level disruptions, lack of early insight, and inconsistent enforcement which existing systems fail to interpret.

SOLUTION

I explored how AI-generated suggestions could help administrators intervene earlier while keeping humans in control of the decision.

EXPECTED IMPACT

Reduce incident frequency and escalation

-30%

Escalation reduction

-20%

Repeat referral rate

-25%

Tier 2 and 3 reduction

CONTEXT

Minga is a tardy and behaviour management platform for K-12 schools. Its main focus is to improve its MTSS supporting framework.

U.S. schools use a widely adopted Multi-Tiered System of Supports (MTSS) model to ensure the appropriate academics and behavioural assistance is offered to students based on their Tier (Tier 1 / Tier 2 / Tier 3).
‍
Over time, repeated infractions may lead to escalated discipline such as detention or suspension and to the student being moved to Tier 2 or 3.

PROBLEM

Schools lack intelligent systems that translate behavioural data into actionable foresight.

Without predictive modelling to identify escalating risk trajectories, educators must rely on intuition and hindsight to make intervention decisions.

GOALS

To design a solution guided by transparency, human oversight, and actionable insight.

1. Clarity over complexity

Surface insights in a way that is immediately understandable and actionable

2. Proactive over reactive

Focus on early signals, not just recorded incidents

3. Explainability builds trust

Always show why a prediction or recommendation exists

4. Assist, don’t replace

Support educator decision-making and never automate discipline

PROCESS + INSIGHTS

I generated a report using ChatGPT Deep Research to summarize the literature on what administrators struggle with most regarding behavior management

Key insights 1: Volume

Too many small incidents → impossible to analyze manually

Key insights 2: Insight problem

Data exists → but no actionable intelligence

Key insights 3: Timing problem

Everything is reactive → no early intervention

Key insights 4: Teacher Burnout

Behaviour challenges → major driver of stress and burnout

DESIGN 1/2

AI-generated prioritization and suggestions  to help admins intervene earlier while keeping them in control.

Because this feature supports sensitive student behavior decisions, AI suggestions should prioritize precision over recall.
‍
The system only presents high-confidence recommendations, and is reinforced via user feedback and actions.

DESIGN 2/2

Intervention progress tracker supported by AI to help admins adapt students’ needs.

Because managing problematic behaviour is an on-going and dynamic process and needs to be tailored to the student’s situation.

REFLECTION

Using AI in discovery

While AI-assisted research helped me rapidly explore the problem space and identify recurring patterns, it also highlighted the limitations of discovery without direct user interviews. If I continued this project, my next step would be validating assumptions with assistant principals to better understand how these workflows play out in real school environments.

Designing AI for trust, not automation

In a school setting, over-automation can quickly reduce trust, especially when decisions affect students directly. This shifted my focus away from prediction accuracy alone and toward explainability, confidence, and human control.

Next project