Blogs

Using AI to Address Misinformation

By Karl Zarins posted 5 hours ago

  
Ohio has seen numerous changes to school funding and property tax laws in recent years. As these changes unfolded, we found ourselves struggling with how best to communicate their impact to our community. The issues are complex, and public discussion is often polarized. The sheer volume and speed of information shared online can make it difficult for residents to separate fact from fiction.

The Challenge

Like many districts, we understood how rapidly misinformation can spread. When inaccurate information circulates without being addressed, it can be accepted as fact. We wanted to provide our community with a single source of information that was accurate, comprehensive, and straightforward. 
 
To accomplish this goal, we decided to create a dedicated district webpage explaining property tax issues, recent legislative changes, and the link between school funding and property taxes. Unlike larger districts, we do not have a communications professional on staff. That led us to explore whether artificial intelligence could help us develop the page.
 
We use Anthropic for finance-related work within the district, so we turned to that platform as a starting point. Rather than relying on AI to gather information on its own, we first identified trusted sources and fed those materials directly to the tool. Sources included the Ohio Education Policy Institute (OEPI), the Ohio Department of Education and Workforce (DEW), Ohio Capital Journal reporting, and Legislative Service Commission (LSC) analyses of recent property tax legislation.
 

Our Approach 

While it requires some time and effort, using AI to create websites should not be intimidating.  First, we gave AI some ground rules.  We asked it to assess our current websites and documents and create a “skill” to ensure that kept branding consistent (“skill” is the Anthropic term).  We also directed it to keep it simple at a 6th-grade reading level and informational vs persuasive. We defined length parameters and uploaded reliable source materials and website URLs to direct its research.  
 
Overall, we spent approximately three and a half hours creating and refining prompts, reviewing drafts, checking for accuracy, and organizing the content into a format that is useful for the public. We asked the AI to summarize complex topics, compare policy changes, identify common misconceptions, and present information in a clear, easy-to-read manner. Multiple rounds of prompts and revisions helped ensure the final product was complete and easy to navigate.  For our own comfort, we asked our legal counsel to review the finished product to ensure we were within the bounds of public discourse and advocacy.
 
The result is a resource that we can confidently share with residents, taxpayers, staff, and community stakeholders. Most importantly, it gives us a place to direct people when questions arise or when inaccurate information is being shared. Rather than responding individually to every rumor or misunderstanding, we now have a centralized source that sets the record straight and provides the facts people need to make informed decisions.  Maybe one day we will assess the use of a specialized agent to answer the community’s specific questions, but we are taking things one step at a time.
0 comments
13 views

Permalink