Will you chip in to support our nonprofit newsroom with a donation today? Yes, I want to support My Edmonds News!
I am writing this as a follow-up to the Jan. 8 article, An urgent call to “reboot” citizen engagement in Edmonds.
Reaching consensus on our city’s “reboot” feels increasingly difficult. As a resident extremely disappointed by our current leadership, it has been nearly impossible to find common ground with those in whom trust has been lost. To break this stalemate and move toward a productive solution, an impartial perspective on how a public entity can restore its reputation after a history of deception and misconduct was sought. The keyword is “Trust,” which is right now, non-existent.
I sought an impartial tool, Google’s AI, which is powered by the Gemini family of models, to answer a central question: “How can a public entity restore its reputation and trust after deceiving its population and employing illegal methods?” The response outlines a systemic transformation focused on four key pillars:
- Immediate Acknowledgment and Accountability
True reform begins with a full admission of wrongdoing without deflection.
- Actions: Issue a sincere public apology and commission an independent, third-party investigation.
- Measurable Goal: Replace all implicated senior staff and release the full investigative report within 90 days.
- Full Transparency and Operational Reform
The entity must adopt a “presumption of disclosure” for all public records.
- Actions: Ensure all meetings are live-streamed with documented public comment and establish robust whistleblower protections.
- Measurable Goal: Fulfill records requests within three to five business days and achieve a 30% increase in public transparency perception within six months.
- Community Engagement and Partnership
Restoring trust requires aligning the entity’s priorities with the population’s needs.
- Actions: Create an empowered citizen oversight committee and hold town halls to redefine city priorities.
- Measurable Goal: Establish the oversight board within four months and maintain an 80% approval rating regarding the entity’s responsiveness to their inquiries.
- Ethical Consistency and Results
Tangible outcomes are the only way to prove a change in culture.
- Actions: Enforce a strict new code of ethics with mandatory annual training and regular independent performance audits.
- Measurable Goal: Meet 90% of key performance indicators (KPIs) in a community-aligned strategic plan for two consecutive years.
A clean slate is needed for this discussion. By following these measurable steps, it is possible to move beyond the current “back and forth” toward actionable change.
The AI has no “feelings” and neither any interest in this discussion. It only provided a neutral analysis and the very basic requirements on which trust is based. Until the city takes serious and real steps to rebuild trust, anything will be a waste of time and just foment more divisiveness. These steps, which are very reasonable and doable, are a suggestion to move things in the right direction.
Mario Rossi is an Edmonds resident.
Can you please forward this to all our state leaders. It isn’t just Edmonds.
OK. Let’s also ask AI to figure out how to “obtain full admission of all wrongdoing without deflection.” I would like to know how that will be accomplished. Perhaps AI could then lead us along the Yellow Brick Road to systemic transformation by accomplishing the remaining bullet points. We don’t need council members. Let’s build a new data processing center near the state-of-art water treatment center. Oh, and the residents won’t mind increased electricity charges to pay for it when they see AI reboot their city government in the same simple way it answered the central question. AI, lead the way to a brave new world. We will sit back and enjoy the ride. Sorry, satire only goes so far. Somewhere hard work is in need of doing. Compromises will have to happen. Some residents’ needs are others’ wants.
Mario, there is no such thing as an impartial AI tool. Every AI tool contains the biases of the data on which it was trained. Furthermore, the prompt that you provided to the AI tool contains extreme bias. It explicitly inserts the concepts of deception and illegality into the conversation.
You’re right, Niall. There’s no such thing as a fully impartial AI. An AI is only as good as the data it’s trained on and the prompt it’s given. Feed it a leading question and you’ll get a leading answer.
A more useful approach would be to ground the analysis in real, local data. For example, take several years of My Edmonds News coverage focused on city governance, feed that in, and ask the AI to assess patterns in actions, decisions, and outcomes. From there, you can identify recurring issues and then dig deeper into possible causes and solutions.
Used this way, AI isn’t an authority or a verdict. It’s a tool to help structure analysis and surface themes, but the real value still comes from grounded local work like Kevin Harris’ engagement assessment and the 40+ interviews behind it.
I’m responding here solely as a resident, not in my role as a member of the Edmonds Economic Development Commission.
It should be no surprise that the words deception and illegality in the prompt lead to a biased analysis.
Is there any other way to describe what the mayor and council have been doing? I was quite moderate and conservative when using the words “deceiving” and “illegal”. Perhaps should have use “lack of ethics and morals”, “betrayed the trust put on them by their voters”, “broke electoral law”, “misappropriated city funds”? There are more.
Niall, as much as we can tell, an AI is just a machine running electrons instead of blood. It does not have any sentiments, let alone any interests in the current issue we have facing with our “representatives”. Therefore, it will not bias its recommendation, which it bases on thousands of documents it reads in order to create it. It does not win anything no matter which direction this goes. Also, please let me know how my prompt was biased.
Your descriptors are the bias as you prompted AI, because not everyone sees the situation your way or would use those terms to describe what is going on with Edmonds’ government. It is a prompt based on your opinion only and therefore subject to that bias. If I had wanted to ask AI, which I don’t, I could have used the prompt “How can a city government overcome unreasonable demands from a segment of the public that is biased against it by unfounded, politically-motivated accusations that slander the public servants who are working to keep city services running despite a budget shortfall?” Now you might not like the answer that AI gives with that prompt would you? It would be outrageous of me to ask that or say it wasn’t biased. And I would never make that prompt to begin with. It’s not true.
Not all citizens of Edmonds agree with your viewpoint. There are some things that can be improved with how Edmonds runs, but it doesn’t mean they are immoral as you claim. And rebuilding trust is something that is the responsibility of both sides. And the main thing your side must do is realize you may not be right about everything and other people have a different viewpoint. You turn people off. They don’t want to listen when you lash out.
Mario
If you don’t understand how your prompt can bias the output of an AI tool, perhaps you shouldn’t be using AI to try to answer complex questions.
I believe Alan proposed a better alternative of you want to use AI to address this issue but I actually agree with Alan’s other point that AI is really no match for the kind of detailed analysis that Kevin Harris performed.
Arlene and Niall,
Approx. 60% of the votes is hardly “not everyone”, isn’t it? Also, considering that it was an all-time-high number of voters, we are not talking about minority of the population.
Also, funny, how the request for transparent and honest disclosure about how the “deficit” numbers and “tax requirements” grew so much suddenly became “unreasonable”.
On the AI bias, I do not sign in to Google when using the AI, which I normally use to research stuff for my photography and bicycling hobbies. This was a first experiment on such topic. I also configure my browsers to delete all cookies and history when I shut them down. Therefore, no history for the AI to use. Also, as far as I know, the AI (at least the good ones) does not skew an answer based on the user’s beliefs.
Furthermore, Kevin’s great article talking about the reboot already implies the lack of trust on the city’s administration. As already stated, my intent using an AI engine was to get an impartial point of view that bases its answers on thousands of documents and situations already registered everywhere, from many different points of view. And they seemed quite on point if we compare to the overall comments to similar topics, isn’t it? So, what bias was there in my query?
The problem with your argument is that you cannot say that all the 60% who voted no think the City is immoral. They might have just thought the number was too high, or they just didn’t want to pay more. You cannot claim a mandate that the City is being dishonest with our funds just on that 60% number alone. All you can claim is that the voters didn’t want to support that levy lift. How that is handled going forward does not mean that you alone have the answer to it. Like I said, it turns people off from listening to you and it makes your arguments suspect. You certainly can offer suggestions and criticisms. We all can. But claiming a mandate does not help your side.
Excellent road map. Thank you, AI, and to Mario for turning to this resource. Now we need methods to accomplish Step 1 which is what people have been asking for all along.
Your City-PIO just attempted step one in her reply in another thread that (paraphrasing) said the city has made a few little minor mistakes in communication here and there but they are planning to do much better now. Maybe your AI question should have been, “how does a city entity correct a few little minor mistakes in communication and do better governance for all it’s residents? The PIO reply was Sort of like trying to use a pea shooter to bring down an elephant. I’d say that until the City of Edmonds does a major re-boot of some sort to make basic needs and postponed infrastructure maintenance the ONLY priority in budget planning, with some sort of results measuring mechanism built in, nothing much is going to change and the financial situation can only be resolved with a major property tax levy lift.
AI isn’t set up to fact check itself. Instead it makes strategic guesses of what the correct answer should be based on what it has been trained on, and that training has many biases built in. AI does not go out into the real world to verify. People have to do that, so why waste the time and incur the huge electrical energy costs from data centers to ask AI?
That is why you cannot use AI as a basis for any argument. It is wrong so many times. Anyone who uses AI as a basis for any argument is not to be believed in my eyes when they do it. I do not use AI to fact check.
Strange all the poo pooing, this AI query gave good advice to our leaders I think they should take it to heart. What it gave was suggestions as to how to, not a ultimatum of have to.
Why strange, Jim? The AI Overview in this MENS printing clearly points out an analysis of the diversity of opinions related to this article. Here we have a 12 point plan generated by a question that some people believe to be biased. AI seems to have “muddied the waters” of a course to trust, rather than cleared the way.
So trust is “non-existent” for the author but the loaded question to AI seems to make the assumption that the entire population of Edmonds feels exactly the same as the author & that is not believable. I’m not inclined to buy this lengthy AI generated response. Rather suggest the author poll ALL Edmonds residents say, over the age of 16, as to their trust level, tabulate ALL the responses & then see how many are in the “lost all trust” column. We just had an election here. Some things changed with that election. But the author & a percentage of the citizenry spparently still are unhappy as if the entire council, mayor & staff should all be tossed out. Or am I missing something? BTW, Council meetings are streamed, public comments are allowed in person & electronically, and town halls have been held and yet there is dissatisfaction. I wish the author & other unhappy folks had run for city council for a start.
Pamela,
60% of the votes in an all-time-high voter participation very hardly seems a minority, isn’t it?
On the “population participation” and “council and mayor listening”; have you been following MEN and the comments’ section? You will get a glimpse on the “listening meetings” and “feedback polls” and how they are perceived.
Google Gemini will tailor its answer to a user based on their past questions and searches. I copied and pasted the same question as the author into Gemini, and got a very different answer, including “reparative justice” as bullet 2. I have never used Gemini before and used it in incognito mode, so it should not be using anything about my history.
At my work, we had an AI developer from the Allen Institute come and explain their work, and one of the key points was that the prompt you provide is critical. Developers spend a great deal of time learning how to word a prompt to get a correct or plausible answer, and they have web forums where they share what they learn. The AI systems that Google and others have unleashed for public use are just as sensitive to the prompt they receive, but 99% of users have no guidance on how to properly wrote a prompt. Without that, any answers should be taken with great caution.
AI is a useful research aid, but it’s still just a tool. The responsibility falls on us to question the answers, check the sources, and test the consistency by asking the same question in different ways—or even by comparing results across multiple AI platforms. Verification comes first; trust comes only after the facts hold up.
As for me, I’ll always put my faith in RI—Real Intelligence.
It’s really too bad that so many folks pile on to the anti-AI sentiment- rather than acknowledging that the Mayor and City Council (except for CP Michelle Dotsch and CM Erika Barnett) have totally failed the residents and ignored the 59% super-majority defeat of Prop 1 – and the 200 residents who have petitioned to have major good governance reform supported by the Mayor and majority of Council. CP Dotsch has a huge uphill battle to get reform and spending under control, when none of the ‘partisan’ Council majority are willing to acknowledge any fault, and are unwilling to reverse the utility tax increase, and unwilling to establish a Citizen Financial Advisory Commission that is given powers to do effective due diligence on the 2021-2025 excess spending that was used as the basis of the bloated 2026 budget. AI is nothing more than affirmation that the Council has not shown fiscal discipline, accountability, transparency, performance-metrics/results-driven management, common sense, and putting taxpayers first – which are the hallmarks of good governance. What does it take to get them to actually hear what 59% of taxpayers have said and to take corrective action? Hopefully 59% of voters will vote out the ‘partisan’ Council members at the earliest opportunity. Until then, be persistent and unrelenting in public comments and petitioning for reform: https://www.ipetitions.com/petition/enough-utility-tax-increase-is-last-straw
Brian,
I do not sign in to Google when using the Gemini AI, which I normally use for hobby research (photography, bicycles, etc.). My web browsers also delete all cookies and history upon shutting down. Anything left is regularly erased by dedicated software. Furthermore, this query was a first attempt on this kind of topic. Kevin’s article made me curious about how a cold and impartial tool would respond to a similar situation we are dealing with. AIs base their answers on thousands of similar articles and situations. In your view, did it fall so far from the general sentiment? I am also curious about what answer you got since, as far as I know, an AI does not radically change its answer based on the user’s profile.