Steve Jones wanted B.C.'s political parties to explain what they would actually do about AI. Where should data centres go? What happens when a government service uses an AI system to make a decision about someone? Who can that person call?
By the time AI Ethical Futures Lab #8 gathered at Parker Street Studios on October 6, he had turned a community discussion into a five-question policy questionnaire. He reported sending it to five provincial parties and was still waiting for replies. During the evening, he shared the full template so others could use it too.
That work gave the Lab's larger conversation something concrete to test. The group was considering becoming a clearinghouse for trustworthy AI information. Steve's questionnaire made the editorial questions concrete: what do we ask, how do we make the answers useful, and how do we leave people free to reach their own conclusions?
Steve Jones: five questions for political parties
The questionnaire had already gone through several rounds. Steve proposed it on September 29 as a short, non-partisan request to party leaders. An early draft grew to eight questions, covering classrooms, jobs, sovereignty, regulation, and AI in election campaigns. Feedback added more issues worth raising, including environmental costs.
There was a practical problem. A party can decline to complete a long survey. A question loaded with assumptions can produce an argument about the question instead of an answer.
Steve separated the material into five broad questions, followed by optional prompts. A party could explain its position in its own words. Readers would still have specific things to look for.
- Infrastructure: electricity costs, water use, local jobs, siting, and ownership.
- Government services: human assistance, oversight, appeals, healthcare, and education.
- Regulation: privacy, creators' rights, Indigenous knowledge, discrimination, disclosure, and accountability for harm.
- The economy: productivity, wages, displaced workers, small businesses, and who receives the benefits.
- Sovereignty: control of data and infrastructure, foreign providers, Canadian technology, and Indigenous data governance.
Those details matter. “We support responsible AI” does not tell a parent what happens to their child's information. It does not tell a worker what support exists when a job changes, or a resident whether they can appeal a decision made with AI. The prompts give people somewhere to press for an answer.
Read Steve's complete questionnaire and optional prompts below, or download the questionnaire. The invitation requested replies by October 16 and proposed publishing dated answers on October 19. As of the October 6 discussion, Steve was still awaiting responses; this recap contains the questions, not a comparison of party answers.
What would a premier take responsibility for?
The discussion continued after the Lab. Steve drafted possible questions for a debate moderator about a government's overall approach to AI, data-centre development, and the division of responsibility between governments. The televised leaders' debate is scheduled for October 14, with Shachi Kurl moderating.
The jurisdiction question came out of that evening's discussion: what should Ottawa decide, what belongs with municipalities, and what would a provincial government take responsibility for itself? It asks for more than a promise to collaborate. People need to know who makes the decision, who can change it, and who answers when something goes wrong.
Later that night, Steve said he would email the moderator the following day. Whether any of the proposed questions would reach the debate remained open. He also invited people to adapt the questionnaire for municipal candidates, with the questions adjusted to what local governments can actually decide.
BC + AI's Vancouver Civic AI Voter Guide is a separate piece of work. Its seven-question framework is used to examine candidates' existing public records; those questions were not sent as a questionnaire. Steve's provincial survey asks parties for their own replies. Keeping that distinction visible lets readers judge what each source can tell them.
A clearinghouse has to show its work
The whiteboard held the central tension: should the Lab take positions, remain a forum for debate, or attempt both?
A person should be able to bring a concern about AI without being dismissed as uninformed. Someone who uses it at work should be able to examine its harms without having to defend their entire job. Access to the same facts does not guarantee agreement about acceptable risks.
But collecting links is only part of the work. Someone chooses the sources. Someone decides which claims need checking, what context to include, and when a correction is necessary. The discussion kept coming back to making those choices inspectable.
Steve's questionnaire is an example of the problem at a manageable scale. Show the questions. Preserve the replies in the parties' own words. Date them. Distinguish an unanswered question from evidence of a policy position. Make the method visible before publishing a comparison.
Data centres make the stakes tangible: power, water, ownership, and consequences for the people living nearby. Both Hands on the Grid follows those questions in Kris Krüg's writing. The BC Public Compute Project asks who owns computing infrastructure and whose interests it serves.
There is a public record to build on
The Lab has already worked through the difficulty of turning a conversation into something people can inspect. Peter Van Garderen facilitated the August session; the Lab #6 archive keeps its meeting artifacts and labelled machine-generated outputs available. Sara Taherzadeh worked through the recording to develop the collective report. The public AI transparency response lets readers follow the resulting recommendations.
At the September 30 Vancouver AI meetup, Peter and Sara brought that work to a wider audience. Peter also demonstrated ImageSquid and discussed Content Credentials. A file's history can help people inspect its origins and changes; it does not settle whether every claim made with that file is true.
October's Lab brought that question home: could BC + AI review its own website imagery and make documentary photographs and generated images easier to distinguish? It was proposed follow-up work. This recap's photographs and the earlier consultation graphic below carry separate captions for the same reason.
Help with the decision in front of you
The classroom discussion was about the questions a parent needs answered. Which tool is being used? For what purpose? What happens to student information? What does the learner still need to work out for themselves?
Yvan Boily's A Parent's Guide to AI came up as a practical example. It covers school and home use, privacy, conversation prompts, and a family agreement. Parents can use it to prepare questions for a teacher without first mastering the vocabulary.
Steve has worked on the design side too. Our May Comox Valley recap follows his demonstrations of SlowSpeak and a student AI tool built around one question a day, citations, and human review. The choices built into a product affect what people do with it.
Cognitive offloading came up in the Lab as a concern for students and people entering a profession. If the tool supplies the answer, where does someone get the practice needed to judge it? The conversation raised questions about learning rather than settling them.
Workplace policy has a similar problem. Guidance about sensitive information has to help at the moment someone is about to paste it into a tool. Could practical guardrails sit alongside that tool, where a worker needs to decide what they can share and when to ask for help? That question connects to the work in BC + AI's Responsible AI Professional program.
Make the work useful beyond the regulars
The Lab's AI Ethics Reading List already brings together resources shared by the community. A clearinghouse would need to help someone find what matters to their question, understand the source, and see where the evidence runs out.
Language, accessibility, marginalized communities, and volunteer time all came up. So did the possibility of public debates that leave a useful record of the arguments and the disagreements. A service built only around the interests and vocabulary of regular attendees will miss people it hopes to help.
The discussion ranged further, through self-improving AI, weapons, surveillance, and claims about job losses. An inventory of ethical questions was proposed as a way to choose future work. The overlap with Mind, AI & Consciousness offers another place to explore the longer philosophical questions.
A small steering group was proposed to help prepare agendas and work out what people could contribute. The clearinghouse's scope, editorial method, and any future debate format still need work.
On the same evening, the Comox Valley AI meetup was comparing everyday experiments with AI and the checking people had to do afterward. Both conversations asked how useful knowledge could travel beyond the person who first learned it.
For this Lab, there is already something readers can take away: five questions, the issues underneath them, and a way to ask for answers. The next part is seeing what comes back.
