Engagement content for Twitter/X, Bluesky, Mastodon, and other platforms. Designed to spark discussion around AI redistribution themes.
Poll Questions (Twitter/Bluesky)
AI & Work
1. When AI automates a job, who should benefit most from the productivity gain?
- The company deploying the AI
- The workers affected
- The public (via taxes/funds)
- The AI developers
2. What's the most honest description of "reskilling" programs?
- Genuine investment in workers
- Cost of doing business
- PR cover for layoffs
- Necessary but underfunded
3. If AI doubles your company's output per employee, your wages should:
- Double proportionally
- Increase somewhat
- Stay the same (profit goes to shareholders)
- Depends on the labor market
AI & Access
4. What matters more for equitable AI access?
- Lower API prices
- Public AI infrastructure
- Open-source models
- AI literacy education
5. AI development is currently concentrated in a handful of companies. This is:
- A natural phase that will diversify
- A structural problem requiring regulation
- Fine, as long as models are open
- The biggest under-discussed issue in AI
6. Which barrier to AI access matters most?
- Cost of compute
- Digital literacy
- Language (English dominance)
- Broadband infrastructure
AI & Governance
7. Who should have the most say in how AI systems are deployed?
- The companies building them
- Elected governments
- Affected communities
- Independent technical experts
8. The EU AI Act is:
- A good first step
- Too cautious
- Too aggressive
- Addressing the wrong problems
9. Public money funded the foundational research behind commercial AI. The public's fair return is:
- Nothing (that's how research works)
- Open access to resulting models
- Revenue sharing / sovereign fund
- Public AI alternatives
AI & Power
10. Five companies control most frontier AI development. The best response is:
- Antitrust enforcement
- Public investment in alternatives
- Open-source mandates
- All of the above
11. When an AI system makes a consequential error (denies a loan, flags someone as a threat), who should be liable?
- The company that deployed it
- The company that built the model
- The institution that chose to use it
- The regulatory body that approved it
12. The phrase "AI democratizes access" is:
- Accurate and hopeful
- Aspirational but premature
- Marketing language
- Depends entirely on the specific system
AI & Society
13. The best historical analogy for AI is:
- Electricity
- The printing press
- Nuclear energy
- The enclosure of the commons
14. In 10 years, AI's biggest impact on inequality will be:
- Making it worse (concentration)
- Making it better (access)
- Mixed (helps some, hurts others)
- Roughly neutral
15. What's most missing from mainstream AI coverage?
- Who pays the costs
- Technical accuracy
- Global perspectives
- Worker voices
Open-Ended Discussion Questions
1. The distribution question: AI is increasing productivity in several industries. But productivity gains and wage gains decoupled decades ago. What would it actually take — structurally, not aspirationally — to ensure AI productivity gains reach workers?
2. The infrastructure question: If AI is infrastructure (like electricity or broadband), should there be a public option? What would publicly funded, publicly governed AI infrastructure actually look like? Who runs it? Who audits it?
3. The ghost work question: Every AI system depends on human labor — data labeling, content moderation, RLHF training. This work is largely invisible, precarious, and poorly paid. What would fair labor standards for AI supply chains look like?
4. The commons question: AI models are trained on the collective output of humanity — text, images, code, conversations. Does the public have a claim on the resulting systems? If so, what form should that claim take?
5. The literacy question: Ursula Franklin argued that understanding technology is a precondition for democratic participation. What does meaningful AI literacy look like — not "learn to prompt," but genuinely understanding these systems well enough to govern them?
6. The local question: Most AI governance discussion happens at national or international level. What could local and municipal AI governance look like? Procurement standards, algorithmic audits, community oversight boards — what's realistic?
7. The global question: Data flows from the Global South to train models in the Global North, which are sold back as products. Birhane calls this algorithmic colonialism. What would a genuinely equitable global AI ecosystem look like?
8. The transition question: Every major technological shift has a transition period where costs and benefits are unevenly distributed. We're in AI's transition period now. What specific policies would make this transition less brutal for the people bearing the costs?
9. The accountability question: When an AI system causes harm — discriminatory hiring, wrongful benefit denial, biased policing — the chain of responsibility is deliberately unclear. How do we build accountability into systems designed to diffuse it?
10. The imagination question: Most AI discourse is either utopian ("abundance for all") or dystopian ("mass unemployment"). What's a realistic, specific, achievable vision for AI that genuinely serves the public interest? Not the best case or worst case — the case we could actually build?
"This or That" Engagement Posts
Framed as genuine trade-offs, not gotchas. Each pair represents a real tension in AI policy and development.
1. Speed vs. Safety Would you rather have:
- Faster AI development with post-deployment regulation
- Slower AI development with pre-deployment requirements
The electricity industry chose "build fast, regulate later" and got decades of monopoly abuse. Pharmaceuticals chose pre-market testing and got slower but safer drugs. Which model fits AI?
2. Open vs. Governed Would you rather have:
- Fully open AI models anyone can use for anything
- Restricted models with usage guardrails and accountability
Open models enable innovation and resist censorship. Restricted models enable accountability and reduce misuse. The tension is real. Where do you draw the line?
3. Efficiency vs. Employment Would you rather have:
- AI that maximizes productivity (even if it displaces workers)
- AI designed to augment workers (even if it's less "efficient")
Acemoglu calls the first "so-so automation." The second is harder to build and less attractive to investors. But Brynjolfsson's research suggests complementarity creates more total value. Which should policy incentivize?
4. Corporate vs. Public AI Would you rather have:
- The best AI built by private companies, accessible via paid APIs
- Decent AI built as public infrastructure, freely accessible
The private option is probably more capable. The public option is probably more equitable. Is "good enough for everyone" better than "excellent for those who can pay"?
5. Global Standards vs. Local Control Would you rather have:
- One global AI governance framework (consistent but slow)
- Many local/national frameworks (responsive but fragmented)
Ostrom showed that polycentric governance often outperforms centralized control for complex systems. But AI is globally networked. Can local governance work for global infrastructure?