Alaska Public Comments
Agent
Government transparency at scale through a Human-in-the-Loop processing engine.
Role
Product Designer
Timeline
6 Months
Overview
Alaska State government receives thousands of public comments for regulatory changes. Traditional review methods were slow, prone to bias, and lacked public-facing transparency. We designed an AI agent to categorize, summarize, and highlight key sentiments—all while ensuring every decision is reviewable by a human expert.
The Challenge
Complexity vs. Trust
Public discourse is nuanced. Automating the summary of public concerns carries significant risk: if the AI misses a critical community grievance, it undermines the democratic process.
The primary challenge was building an interface that could process massive datasets without becoming a "black box," allowing state officials to justify every summarized point back to the original source text.
Process
We began with deep ethnography, shadowed state reviewers as they manually tagged PDFs. We identified that 60% of their time was spent on administrative sorting rather than critical analysis.
User Research
Interviewed 12 state officials to map the legal requirements for comment processing.
Iterative Prototyping
Tested 4 versions of the "Confidence UI" to see how reviewers reacted to AI uncertainty.
The Solution
An Interface for Collaborative Intelligence
Transparency-first interface
Every AI-generated summary includes a direct link to the source paragraph, ensuring no statement is unattributed.
HITL Review Workflows
A specialized dashboard for experts to approve, edit, or reject AI suggestions with a single click.
Comments Processed
50,000+
Across three state regulatory cycles
Reviewer Accuracy
99%
Verified through blind human audit
Time Saved
40%
Reduction in manual sorting time
Transparency
100%
Audit trails for every summary point
Have a similar challenge?
I help organizations build trust in their AI systems through transparent, audit-ready design.