I recently posted a deep-dive analysis on Reddit about Massachusetts’ 15-year literacy crisis. While the data and substance weren’t seriously contested, a portion of the reaction focused on the fact that I used Large Language Models to help map context, cross-reference data, and organize findings. Some readers immediately dismissed the work on that basis alone.
That pushback was actually useful. It highlighted a widespread misconception: that using AI automatically means you’re not thinking for yourself.
In truth, there’s a vast difference between asking an AI to generate lazy content and using it as a powerful context and research engine.
For decades, the political system — lawmakers, agencies, and entrenched interests — has protected itself through density. This comes in two forms: the sheer volume of hundred-page bills and the complex, dry legal language packed even into shorter ones. Massive lump-sum budget lines with little detail make it extremely difficult for working people with jobs and families to follow what’s really happening. As a result, most citizens end up relying on short headlines, soundbites, or partisan commentary.
LLMs break that barrier. They reduce the cost of serious policy analysis to zero. They give ordinary citizens research capabilities that once required full-time staff or expensive consultants. Browsers are still incredibly valuable, but LLMs supercharge them by providing instant context, relevance filtering, synthesis, and reduced friction — helping you understand, learn, and discover far more efficiently. (Think of it like using a spreadsheet versus an abacus.)
Take the recent Massachusetts literacy bill now on (or recently signed by) Governor Healey’s desk. Tools like these made it possible to map 15 years of MCAS (Massachusetts Comprehensive Assessment System) data — the state’s standardized testing program — and pinpoint exactly where third-grade reading proficiency has stagnated in specific Gateway Cities and districts. The same approach helped clarify the real differences between failed three-cueing / balanced literacy methods (guessing from pictures and context) and more effective Systematic Synthetic Phonics instruction.
It also revealed how key funding and accountability provisions changed between bill versions. You can read the bill text here.
Bonus capability worth knowing: Google’s NotebookLM lets you upload bill PDFs and other documents, then instantly generate your own audio “podcasts” with two hosts discussing the legislation. It’s an excellent way to absorb complex material while commuting or doing chores.
This isn’t replacing human judgment. It’s sharpening it.
Try It Yourself – Quick Start Guide
- Go to your state legislature’s website (in Massachusetts: malegislature.gov).
- Search for a bill that interests you.
- Download the PDF version of the bill text.
- Paste sections into an LLM (or upload to NotebookLM) and start asking questions.
Prompt 1 – Bill Auditor (Low-Hanging Fruit)
“Act as a non-partisan legislative analyst. Here is the full text of Massachusetts bill [paste the relevant sections or summary]. Break it down plainly: What does this bill actually require? What funding is authorized, and is it earmarked for specific uses or left vague? What are the main enforcement mechanisms? What loopholes or vague phrases could limit its impact in high-needs districts?”
Prompt 2 – Reading Science Decoder
“Explain in plain language the cognitive difference for a struggling 4th grader between ‘three-cueing / balanced literacy’ (guessing from pictures and context) and Systematic Synthetic Phonics. How does each approach affect working memory and the shift from ‘learning to read’ to ‘reading to learn’?”
Prompt 3 – Long-Term Outcomes
“What are the statistically documented long-term outcomes for students who are not reading at grade level by the end of third grade? Include impacts on high school graduation rates, dropout risk, future earnings, involvement with juvenile justice or social services, and lifetime likelihood of needing public assistance.”
You can run Prompt 3 on any LLM (ChatGPT, Claude, Gemini, Grok, etc.). Try it on two or three different models and compare the answers.
Additional Resources (great starting points to explore further)
- ExcelinEd Fact Sheet on Why Three-Cueing Doesn’t Work – Clear explanation of the limitations of guessing-based methods.
- Lexia Learning – Shifting from Three-Cueing to Science-Based Reading – Good contrast between the two approaches.
- Annie E. Casey Foundation – Third Grade Reading & Long-Term Outcomes
- Chapin Hall – Third Grade Reading as a Predictor
When you read those outcomes, ask yourself honestly: Would you want this to be your child’s future?
As voters, our job doesn’t end on Election Day. The real decisions that shape our schools, taxes, and communities are made every single day in committee rooms through the fine print of bills and budgets. Most of us have been forced to guess or trust headlines. That no longer has to be the case.
The public data is already available. Free and low-cost AI tools can now help any motivated citizen read it, question it, and hold lawmakers accountable.
The tools exist. The choice is whether we use them.
Is A Bill Just A Bill? 
Leave a comment