WIECARA - Sexual Violence Support Platform
A campus platform designed to provide a safe, empathetic entry point for students through ARA Chatbot, context-aware reporting logic, consent gates, and emergency safety routing.
Connecting early conversational support, resource navigation, consent-based reporting, and emergency escalation in one interface.
The Problem
Someone who has experienced sexual violence or harassment on campus may not be ready to immediately file a formal report. They may still be processing what happened, fearing retaliation, worrying about privacy, or simply not knowing where to begin.
Most existing institutional platforms assume that every visitor is ready to fill out a long reporting form with complete dates, names, and details. This creates a high barrier to entry.
What We Changed
WIECARA separates early conversational disclosure from formal reporting. Users are never forced into a report automatically.
Assumes users are ready to disclose all details immediately without support or exploration.
The Part I Worked On
My responsibilities centered on designing ARA Chatbot's intelligence, prompt architecture, context-aware escalation rules, user consent flows, and connecting conversation state to backend reporting routes.
ARA Prompt & Logic
Designed empathetic, non-judgmental prompt templates with Groq LLM integration and strict boundary guidelines.
Consent Gate Flow
Built user consent check before converting chat context into formal report drafts or sharing data.
Contextual Reporting
Implemented intent tracking across message history rather than triggering on simple isolated keywords.
Emergency Escalation
Connected critical threat detection in chat directly to emergency campus security & hotline triggers.
Try the Chatbot ARA Playground
Click through scenarios below to explore how ARA handles conversation context, consent gates, and safety triggers.
ARA Chatbot & Prompt Engineering
ARA serves as a low-pressure entry point. Users can describe their situation without immediately committing to formal documentation.
I developed the prompt structure using Groq LLM to ensure ARA remains empathetic, respectful, and strictly bounded within its supportive role.
ARA supports early conversation and guides users to campus resources; case investigation, psychological evaluation, and legal actions remain strictly with qualified human personnel.
Consent Gate & Reporting Flow
In WIECARA, reporting is offered naturally as conversation context develops - not triggered mechanically by a single word or single message.
Before any conversation data is pre-filled into report fields, the Consent Gate requires clear user confirmation. Users can choose to continue talking, postpone, or proceed.
This keeps the user in total control over their data and decision at all times.
AI does not evaluate the case
We designed WIECARA with clear ethical boundaries. AI output should never replace human validation or institutional decision-making in sensitive sexual violence cases.
- • Provide safe conversational entry point
- • Explain available campus resources
- • Draft report forms with user consent
- • Fast-track immediate emergency alerts
- • Case investigation and evidence check
- • Professional counseling & therapy
- • Institutional protection measures
- • Legal & disciplinary enforcement
Desktop Views & Gallery
Landing Page & Entry Point
Overview of campus support services and clear pathways for students.
What Is Still Unproven
WIECARA was developed as an academic web prototype. It has not yet been deployed as an active campus service with real counselors and security response teams.
We cannot claim that WIECARA increases report rates or reduces campus incidents. Those claims require pilot evaluations, security integration testing, and long-term user research with support organizations.
What I Learned
Working on ARA made me realize that building AI for sensitive contexts is not just about prompt accuracy or model performance: it's about system boundaries and user agency.
When should the AI offer an option?
When should it give the user space to pause or step back?
And which decisions MUST remain with trained human professionals?
Designing product boundaries that prioritize user consent over aggressive feature automation was the most valuable technical lesson from WIECARA.
Technology Used
Key Outcomes
Lower-Pressure First Step
Users can start with a conversation instead of being pushed directly into formal reporting forms.
Consent Gate Control
Reporting is offered as a user-controlled next step, not an automatic consequence of sharing.
Clear Product Boundaries
ARA supports conversation and navigation; formal case decisions remain strictly with humans.