Election Content Analysis and Improvement Tool

ABOUT THE PROJECT

To help the NJ Division of Elections share accurate and timely information ahead of a gubernatorial election and anticipated changes to election law, this AI tool analyzes website content against established content guidelines and identifies unclear, redundant, inaccessible texts side-by-side by revised text suggestions. It makes design accessibility recommendations along with source code evaluation. The tool generates sample voter personas in order to recommend site improvements to adjust to various users. This is the first step in building additional tools, such as a voter chatbot to provide reliable, accurate voter information.

Impact: Help New Jersey rapidly update online materials in response to changes in federal and state election law. Tool is reusable across other websites.

Partner: New Jersey Division of Elections

Project Status: Functional and ready for deployment

PARTNER

EXECUTIVE SUMMARY

Challenge: As New Jersey prepares for its 2025 gubernatorial election, approximately half of its residents will rely on the New Jersey Division of Elections (NJ DOE) website for critical information. But as central hubs for information, the website built on outdated infrastructure contains an overload of complex content, making it difficult to navigate. 

Solution: Machine Assistance for Experience (MAX) — an AI-powered tool that analyzes government websites to enhance the user experience. The tool provides guidance for public servants on the user-centered design process, as well as, an opportunity to gain feedback on content clarity, website design, and best code practices (WCAG 2.1 AA.). MAX evaluates language, website layout, and HTML to provide actionable improvements.

Impact: Public sector teams gain automated insights for improving their websites, enabling clearer content, easier navigation, and faster AI tool integration with the long term goal of helping more residents engage with civic services.

Partners: New Jersey Division of Elections, New Jersey Department of State,  New Jersey Office of Innovation, Northeastern University’s Burnes Center for Social Change at Northeastern University

PROBLEM CONTEXT

Background: Government websites are central to civic participation, but often built on legacy systems that are hard to update and confusing to navigate. For example, the Division of Elections site does not utilize a modern content management system,  each page is updated individually using Dreamweaver, and  only two people are able to maintain the site.

Urgency/Need: With a major election approaching in NJ, there’s an immediate need to modernize information delivery for residents including first-time voters, seniors, and long-term community members. 

Target Audience: Residents, Voters, election officials, and government staff 

INNOVATION PROCESS

Approach: 

We used Agile development practices to rapidly prototype and refine features that meet project needs. Further user feedback and testing is required. 

Co-Creation:

While developing, our initial motivation stemmed from the New Jersey Gubernatorial Election (DOE) that will be held in approximately 5 months. As the NJ Division of Elections website will be their go-to source for critical voting information, our scope of project formulated around the website. 

While meeting with the NJ DOE, we found that the current website relied on outdated content management practices, making it difficult to integrate AI tools. Similar to many government websites, the NJ DOE website, as central hubs for public information, suffers from content overload. Designing for clarity, organization, and accessibility is increasingly difficult, especially when accommodating a range of users with different needs and abilities. Integration of AI is difficult when the underlying content is overly complex. 

In order to support the DOE in a website redesign or future integration of AI tools, we built Machine Assistance for Experience. Data Sources: NJ Division of Elections Homepage Website / Public Government Websites

AI SOLUTION OVERVIEW

What was built:

Machine Assistance for eXperience (MAX), is a web application that analyzes a government website for a better user experience. With our tool, public servants can have guidance though the user- center design process, readily improve their content clarity and website layout, and implement code best practices. 

Key Features:

Analyze Audience members to Support User-Center Design: Guidance through the user center design process by developing personas and analyzing websites. 

Content Clarity Review: Reviews all text found on the website with the aim to rephasing it between a six to eighth grade reading level. 

Website Design and Layout Review: Analyzes the webpage screenshot to suggest layout improvements (e.g., contrast, layout, label visibility)

Code Best Practices: Evaluates HTML source code for Website Content Accessibility Guidelines (WCAG) violation and returns specific, fixable HTML snippets

LESSONS LEARNED

What Worked:

• Generating structured feedback 
• Content clarity page provides a consistent output for changes of the text to be between a 6-9 grade reading level 
• Code Best Practices page is very readable and consistent with helpful suggestion 

What Didn’t:

• LLMs hallucinated without strong prompting, sometimes the output is not actionable 
• Large HTML files had to be chunked to maintain response, further testing without chunking could be done as context windows have increased in size
• Web Design page can be generic and not very actionable actionable 

Adaptability:
MAX can analyze any government or non-government website with a url.