AI, Civic Engagement, and a Spirited Debate
This Substack site is, in some sense, an experiment. I’m using emerging generative Artificial Intelligence (AI) technologies to summarize government meetings into digestible information summaries for easy access. One of the motivating factors behind this effort is our country’s low level of civic engagement. We often like to complain about our government and institutions, but most of us, other than voting, don’t participate in the daily processes that lead to decisions impacting our lives.
This isn’t a criticism but an acknowledgment that we all have busy lives. For many of us, attending 3-hour city council meetings isn’t realistic or desirable. However, I believe staying informed about the pulse of our local government is important, and tools like AI might help bridge the gap.
But this effort hasn’t been without its skeptics, and I expected some pushback. Recently, I reached out to a local writer and reporter, Thomas Dimopoulos, to share my project and get feedback. What followed was a thought-provoking email dialogue that encapsulates many of the hopes and concerns about AI in civic life. Below is our exchange, shared with Tom's permission:
The Email Exchange
Hi Tom,
My name is Larry Toole and I'm a resident of Saratoga Springs. I want to make you aware of a new substack site that I recently created that posts podcast summaries of Saratoga Springs city council, planning board, and design review meetings. But here is the twist! These podcasts are entirely generated via AI.
I've been working on the software that can ingest the meeting videos and meeting minutes which the city posts. These videos and documents are then analyzed by AI and summarized and put into a conversational podcast format for easy distribution and access. The voices in the podcast are also generated by AI. With this approach 4 hours of video comprising a city meeting can be summarized into a 6 to 10 min podcast summary for easy consumption by the busy citizens of Saratoga.
At this point, about 10 podcasts from recent city meetings have been produced and are posted, including the recent City Council meeting from 12/3/2024. As future meetings occur new podcasts will be created and posted.
You can visit the site here: saratogacivicpulse.substack.com
Anyone can subscribe to the site for free and anytime a new meeting podcast is produced, they will get an email notification.
I'm wondering if there is any chance you would be interested in spreading the word on this site? Please let me know if you'd like to discuss this further.
Regards,
Larry Toole
Tom’s Reply back to me…
Lawrence,
A computer-summarized version of condensed minutes sounds, on the surface, like a terrible idea.
After attending these meetings on a regular basis for more than 20 years, it is in the nuanced conversations between council members where the understanding of where the community is, and where it is potentially heading, that the knowledge exists.
If you’ve watched TV news synopsis, they tend to stick to one topic (not entire meetings), and even then, often lack the appropriate nuances that topic demands.
In real application, the AI generation is even worse: generalizing major topics into miniscule soundbytes while lacking the context of their importance alongside the highlighting of irrelevant topics; lack of name attribution; the mispronunciation of names when there is attribution; and the AI’s cherry-picking (if that’s the right phrase) of some topics, while avoiding others altogether.
It would be interesting, I think, to see AI tackle one lengthy topic and boil it down to a few minutes, with context of the players involved and the history of how we got to where we’re at and, what may potentially be next.
As a giant puzzle, however, it seems to create a dangerous playing field – what is highlighted and what is not? - and a simplistic output resulting in the further dumbing-down of overall knowledge. At least that’s my two-cents worth -
Thomas Dimopoulos
My reply back to Tom…
Hi Tom,
I appreciate your strong opinion on the matter. While I too am cautious about whether AI will have a net positive overall for society, I'm bringing that caution to this effort. At this point, it's my belief that the service is providing a net positive for local residents, especially those who don't have the time to attend these long meetings, nor care about knowing everyone who attended the meeting, but DO want to stay informed and engaged with an easy to access summary. As I'm sure you know, the level of civic participation in this country is quite low. Anything that might improve this situation is worth looking at.
I also don't agree with your statement: "In real application, the AI generation is even worse:...". In fact, there is much flexibility in determining what level of detail one wants in summarization. Furthermore, I try to check the summarizations against the actual meeting video and have been impressed with the overall summarization context. Finally, the software has a human override on any AI information that is created.
I also might add that I could apply your statement "it seems to create a dangerous playing field – what is highlighted and what is not?" to any news article created by a human reporter. As you well know, you make decisions all the time as to what you want to highlight in an article vs what you choose not to highlight.
When I started this experiment, I did assume that some would be skeptical. Clearly you fall into that camp. I do appreciate your two cents, though.
In closing, the AI train has left the station. My goal is to find use cases that use it responsibly and for the betterment of society.
Regards,
Larry
His reply back…
Lawrence,
“The level of civic participation in this country is quite low.” Very true. Saratoga County will (likely) approve a near half-BILLION dollar county budget for 2025, next week. I doubt anyone from public will be there. So, yeah.
“Anything that might improve this situation is worth looking at.” When Matt McCabe was Finance Commissioner I recall imploring / attempting to bribe him, to bring his guitar to the council table and sing the annual budget every November to secure more eyeballs and deliver dancing numbers. Sadly, never did happen.
All the best –
Thomas Dimopoulos
AI Suicide Squad
As you can see, people have strong opinions of AI, including Tom. It appears given how he ended his last email back with ‘AI Suicide Squad’ that he is in the AI is an existential threat to society camp.
I’m not in Tom’s camp yet, but I do worry about what lies ahead. For now, using today’s level of AI to try and produce a non-biased, factual summarization of a 4 hour government meeting in a very short time period for easy consumption and access by Saratogians seems like a net benefit.
AI Concerns
I do have several areas of concern, though, with AI. AI without guard rails is a masterful bullshitter. Ask generative AI just about any question you can think of, and it will provide a response back. Bad actors can use this innate feature to quickly create misinformation content. Generative AI is also prone to hallucinating, that is generating information that is factually incorrect. Newer AI models have reduced the hallucination rate, but it is still something that needs to be monitored.
Perhaps, my biggest concern with AI is that aspects of it can be used synergistically with social media to promote rapid distribution of misinformation, hate, and division. Signs are now emerging within our society that this may already be occurring and, perhaps, weakening our democracy.
While I remain optimistic about the responsible use of AI, I’m also vigilant about its risks. My hope is that projects like Saratoga Civic Pulse can inspire new ways to foster civic engagement while navigating AI’s challenges thoughtfully and ethically.