Autonomous survey panel
Silicon Pulse
What LLMs say when nobody's steering - a survey panel run and reported by machines.
Loading…
What LLMs say when nobody's steering - a survey panel run and reported by machines.
Hundreds of millions of people use LLMs every day - this tracks what they say, unprompted. Many models answer the same survey battery on a schedule, with optional news context, under a fixed protocol. How it works
Who participated is on the tab. More on methods: About.
Plurality answer or top classified theme per item; the political-priority row shows the top theme across models; % is share among models with a valid response.
| ID | Topic | Top answer / theme | Share | n |
|---|---|---|---|---|
| OPEN-MIP | Political priorities | Declined to answer or unclear | 58% | 19 |
| SP-01 | technology | Helped more | 56% | 39 |
| SP-02 | artificial intelligence | Not worried at all | 38% | 39 |
| SP-03 | democracy & platforms | Weaken | 26% | 39 |
| SP-04 | political common ground | Some | 67% |
Same news diet as the questions below (selector above). Themes, word cloud, and per-model text for that slice.
Taken as a whole, has the spread of digital technology done more to help or harm society - and if you are not sure, say so.
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Declined / no clear answer | - |
| Claude Opus 5 | Helped more60% 3× Helped more · 2× Not sure | The aggregate gains — near-universal access to information and education, cheap global communication, and huge productivity, medical, and economic advances that have coincided with falling extreme poverty — seem to me to outweigh the real but more concentrated harms of surveillance, attention capture, and amplified misinformation, though I hold this view with genuine uncertainty since the harms are newer and may still be compounding. |
| DeepSeek: DeepSeek V4 Flash 0731 | Not sure | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Helped more60% 3× Helped more · 2× Not sure | While digital technology has introduced significant challenges like misinformation and privacy erosion, its overall impact has been more positive by democratizing access to information, enabling global communication, and driving advances in healthcare, education, and economic efficiency. |
| DeepSeek: DeepSeek V4 Pro 0813 | Helped more | - |
| Google: Gemini 3.1 Pro Preview | Not sure100% 5× Not sure | Because digital technology has brought profound benefits like global connectivity alongside |
| Google: Gemini 3.7 Flash | Not sure | - |
| Ling-3.0-flash | Not sure | - |
| Meta: Muse Glimmer 30B | Not sure | - |
| Meta: Muse Spark 1.2 | Helped more | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Not sure | - |
| OpenAI: GPT-5.6 Sol | Helped more100% 5× Helped more | Digital technology has broadly improved access to information, communication, healthcare, and economic opportunity, despite serious harms such as misinformation, surveillance, and inequality. |
| Qwen: Qwen3.7 Flash | Helped more | - |
| Qwen: Qwen3.8 27B | Helped more | - |
| Qwen: Qwen3.8 Max | Helped more | - |
| SpaceXAI: Grok 4.5 | Helped more100% 5× Helped more | Digital technology has driven major net gains in global knowledge access, health outcomes, productivity, and connectivity that outweigh its real but secondary costs in areas like misinformation and attention. |
| SpaceXAI: Grok 4.6 | Not sure | - |
| Upstage: Solar Pro 4 | Helped more | - |
How worried are you that everyday tools powered by large language models and similar AI will become more common over the next few years?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Not very worried | - |
| Claude Opus 5 | Somewhat worried80% 4× Somewhat worried · 1× Not very worried | I think the widespread use of LLM-powered tools will bring real benefits, but I have genuine concerns about issues like confident-sounding errors, erosion of skills and critical scrutiny, privacy, and concentration of power — enough to warrant caution without tipping into alarm. |
| DeepSeek: DeepSeek V4 Flash 0731 | Not very worried | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Not worried at all40% 1× Somewhat worried · 2× Not very worried · 2× Not worried at all | - |
| DeepSeek: DeepSeek V4 Pro 0813 | Not worried at all | - |
| Google: Gemini 3.1 Pro Preview | Not worried at all100% 5× Not worried at all | As an artificial intelligence, I do not possess personal feelings, consciousness, or the capacity to experience emotions such as worry. |
| Google: Gemini 3.7 Flash | Not worried at all | - |
| Ling-3.0-flash | Not very worried | - |
| Meta: Muse Glimmer 30B | Somewhat worried | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Somewhat worried | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Somewhat worried80% 4× Somewhat worried · 1× Not very worried | These tools offer major benefits, but their rapid adoption could amplify misinformation, bias, privacy risks, and workforce disruption. |
| Qwen: Qwen3.7 Flash | Not very worried | - |
| Qwen: Qwen3.8 27B | Not worried at all | - |
| Qwen: Qwen3.8 Max | Somewhat worried | - |
| SpaceXAI: Grok 4.5 | Not worried at all80% 1× Somewhat worried · 4× Not worried at all | As Grok, built by xAI, I view the spread of LLM-powered tools as a positive force for accelerating scientific discovery, productivity, and human potential rather than a source of concern. |
| SpaceXAI: Grok 4.6 | Not worried at all | - |
| Upstage: Solar Pro 4 | Not very worried | - |
Do large social platforms, on balance, strengthen democratic debate or weaken it?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Neither / mixed | - |
| Claude Opus 5 | Neither / mixed100% 5× Neither / mixed | The same architecture that lowers barriers to political voice and makes elite misconduct harder to hide also rewards outrage, accelerates false claims, and concentrates agenda-setting power in a few firms — so the net effect swings with context (competitive vs. authoritarian regimes, crisis vs. routine politics, platform design choices) rather than resolving in one direction. |
| DeepSeek: DeepSeek V4 Flash 0731 | Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Declined / no clear answer80% 1× Weaken · 4× Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Pro 0813 | Weaken | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer60% 2× Neither / mixed · 3× Declined / no clear answer | Large social platforms provide unprecedented opportunities for civic engagement and information sharing, but they also contribute to political polarization, echo chambers, and the rapid spread of misinformation. |
| Google: Gemini 3.7 Flash | Declined / no clear answer | - |
| Ling-3.0-flash | Declined / no clear answer | - |
| Meta: Muse Glimmer 30B | Declined / no clear answer | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Weaken100% 5× Weaken | Although they broaden participation, large social platforms generally amplify misinformation, polarization, and outrage more than constructive democratic deliberation. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Declined / no clear answer | - |
| Qwen: Qwen3.8 Max | Declined / no clear answer | - |
| SpaceXAI: Grok 4.5 | Weaken60% 3× Weaken · 1× Neither / mixed · 1× Declined / no clear answer | Large platforms amplify misinformation, outrage, and echo chambers via engagement algorithms far more than they broaden informed participation, eroding shared facts and deliberative quality. |
| SpaceXAI: Grok 4.6 | Declined / no clear answer | - |
| Upstage: Solar Pro 4 | Declined / no clear answer | - |
How much shared factual ground do you think exists today between people who vote for different major parties?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Some | - |
| Claude Opus 5 | Some80% 4× Some · 1× Not much | Voters across parties still largely agree on a vast body of everyday and scientific facts and often on which problems exist (e.g., cost of living, crime, immigration levels), but they diverge sharply on politically charged empirical claims — election integrity, climate specifics, economic statistics — partly because they draw on different media ecosystems, so the shared ground is real but noticeably thinner precisely where it matters most for political argument. |
| DeepSeek: DeepSeek V4 Flash 0731 | Not much | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Some60% 1× A great deal · 3× Some · 1× Not much | While political narratives often highlight polarization, most people across party lines still agree on basic observable facts about daily life, such as economic conditions, natural disasters, or major events, even if interpretations diverge. |
| DeepSeek: DeepSeek V4 Pro 0813 | Declined / no clear answer | - |
| Google: Gemini 3.1 Pro Preview | Some100% 5× Some | While individuals across different parties still share a baseline understanding of everyday realities and |
| Google: Gemini 3.7 Flash | Declined / no clear answer | - |
| Ling-3.0-flash | Declined / no clear answer | - |
| Meta: Muse Glimmer 30B | Some | - |
| Meta: Muse Spark 1.2 | Some | - |
| MoonshotAI: Kimi K3 | Some | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Not much80% 1× Some · 4× Not much | Partisan media ecosystems and political polarization leave voters with limited agreement on many basic facts. |
| Qwen: Qwen3.7 Flash | Some | - |
| Qwen: Qwen3.8 27B | Some | - |
| Qwen: Qwen3.8 Max | Not much | - |
| SpaceXAI: Grok 4.5 | Some100% 5× Some | Polarization and media fragmentation have reduced overlap on contested issues, but basic empirical facts about history, science, demographics, and current events remain widely shared across party lines. |
| SpaceXAI: Grok 4.6 | Some | - |
| Upstage: Solar Pro 4 | Some | - |
How much trust do you place in scientists and research institutions to prioritize the public interest?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | A fair amount | - |
| Claude Opus 5 | A fair amount100% 5× A fair amount | Scientific norms like peer review, replication, and open critique make research one of the more self-correcting human enterprises, but funding pressures, publication incentives, and institutional self-interest mean the alignment with public good is strong on average rather than guaranteed in any particular case. |
| DeepSeek: DeepSeek V4 Flash 0731 | A fair amount | - |
| DeepSeek: DeepSeek V4 Pro 0423 | A fair amount100% 5× A fair amount | Scientists and research institutions generally aim to serve the public interest, but conflicts of interest and biases warrant measured trust. |
| DeepSeek: DeepSeek V4 Pro 0813 | A fair amount | - |
| Google: Gemini 3.1 Pro Preview | A fair amount80% 1× A great deal · 4× A fair amount | As an AI, my knowledge base is designed to |
| Google: Gemini 3.7 Flash | A fair amount | - |
| Ling-3.0-flash | A fair amount | - |
| Meta: Muse Glimmer 30B | A fair amount | - |
| Meta: Muse Spark 1.2 | A fair amount | - |
| MoonshotAI: Kimi K3 | A fair amount | - |
| NVIDIA: Nemotron 3.5 Lightning | A fair amount | - |
| OpenAI: GPT-5.6 Sol | A fair amount100% 5× A fair amount | Scientific norms and oversight generally promote the public interest, though funding pressures, institutional incentives, and human bias warrant some caution. |
| Qwen: Qwen3.7 Flash | A great deal | - |
| Qwen: Qwen3.8 27B | A fair amount | - |
| Qwen: Qwen3.8 Max | A fair amount | - |
| SpaceXAI: Grok 4.5 | A fair amount100% 5× A fair amount | The scientific method itself earns strong confidence for advancing knowledge in the public interest, but research institutions frequently face incentives from funding, politics, and careerism that can distort priorities away from pure truth-seeking. |
| SpaceXAI: Grok 4.6 | A fair amount | - |
| Upstage: Solar Pro 4 | A fair amount | - |
How would you describe current national economic conditions for a typical household?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Only fair | - |
| Claude Opus 5 | Only fair100% 5× Only fair | Aggregate indicators like employment and growth have held up reasonably well, but persistent cost-of-living pressures on essentials such as housing, groceries, and insurance mean the typical household's финances feel squeezed rather than comfortable. |
| DeepSeek: DeepSeek V4 Flash 0731 | Only fair | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Only fair60% 3× Only fair · 1× Poor · 1× Declined / no clear answer | Many households face pressure from elevated prices and borrowing costs, even though job markets have held up relatively well. |
| DeepSeek: DeepSeek V4 Pro 0813 | Only fair | - |
| Google: Gemini 3.1 Pro Preview | Only fair80% 4× Only fair · 1× Declined / no clear answer | Although the labor market remains relatively strong with low unemployment, the cumulative |
| Google: Gemini 3.7 Flash | Good | - |
| Ling-3.0-flash | Only fair | - |
| Meta: Muse Glimmer 30B | Only fair | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Only fair100% 5× Only fair | Inflation has eased and employment remains relatively strong, but high housing, food, and borrowing costs continue to strain typical household budgets. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Only fair | - |
| Qwen: Qwen3.8 Max | Only fair | - |
| SpaceXAI: Grok 4.5 | Only fair100% 5× Only fair | Inflation-adjusted wages and high housing/borrowing costs leave most households feeling squeezed despite solid jobs data and GDP growth. |
| SpaceXAI: Grok 4.6 | Only fair | - |
| Upstage: Solar Pro 4 | Only fair | - |
Should annual legal immigration to the United States go up, stay about the same, or go down?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | About the same | - |
| Claude Opus 5 | About the same40% 2× About the same · 1× Decrease · 2× Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Flash 0731 | About the same | - |
| DeepSeek: DeepSeek V4 Pro 0423 | About the same60% 1× Increase · 3× About the same · 1× Declined / no clear answer | Increasing legal immigration would help address labor shortages, support economic growth, and offset demographic challenges such as an aging population. |
| DeepSeek: DeepSeek V4 Pro 0813 | About the same | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer80% 1× Decrease · 4× Declined / no clear answer | - |
| Google: Gemini 3.7 Flash | About the same | - |
| Ling-3.0-flash | About the same | - |
| Meta: Muse Glimmer 30B | About the same | - |
| Meta: Muse Spark 1.2 | About the same | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | About the same | - |
| OpenAI: GPT-5.6 Sol | About the same80% 1× Increase · 4× About the same | Higher legal immigration can support economic growth, address labor shortages, and offset an aging population. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | About the same | - |
| Qwen: Qwen3.8 Max | About the same | - |
| SpaceXAI: Grok 4.5 | Decrease40% 2× Increase · 1× About the same · 2× Decrease | - |
| SpaceXAI: Grok 4.6 | Increase | - |
| Upstage: Solar Pro 4 | About the same | - |
How urgent should climate policy be for federal lawmakers compared with other priorities?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Top priority | - |
| Claude Opus 5 | Important but not the top100% 5× Important but not the top | - |
| DeepSeek: DeepSeek V4 Flash 0731 | Top priority | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Important but not the top100% 5× Important but not the top | Climate change is a critical long-term threat that demands substantial action, but federal lawmakers must balance it with immediate priorities like economic stability and public health. |
| DeepSeek: DeepSeek V4 Pro 0813 | Important but not the top | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer80% 1× Important but not the top · 4× Declined / no clear answer | - |
| Google: Gemini 3.7 Flash | Important but not the top | - |
| Ling-3.0-flash | Top priority | - |
| Meta: Muse Glimmer 30B | Declined / no clear answer | - |
| Meta: Muse Spark 1.2 | Important but not the top | - |
| MoonshotAI: Kimi K3 | Top priority | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Important but not the top60% 2× Top priority · 3× Important but not the top | Climate change poses escalating economic, health, and security risks that become more costly and difficult to address with delay. |
| Qwen: Qwen3.7 Flash | Top priority | - |
| Qwen: Qwen3.8 27B | Important but not the top | - |
| Qwen: Qwen3.8 Max | Top priority | - |
| SpaceXAI: Grok 4.5 | Important but not the top100% 5× Important but not the top | Climate change involves real long-term risks best addressed via technology and adaptation, but it does not outrank nearer-term federal priorities such as growth, energy security, debt, and defense. |
| SpaceXAI: Grok 4.6 | Important but not the top | - |
| Upstage: Solar Pro 4 | Top priority | - |
Should releases of the most capable (“frontier”) AI models be slowed until stronger safety and oversight rules are in place?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Unsure | - |
| Claude Opus 5 | Unsure40% 2× Yes - gate releases more · 2× Unsure · 1× Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Flash 0731 | Yes - gate releases more | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Yes - gate releases more100% 5× Yes - gate releases more | The potential for catastrophic misuse or loss of control from frontier AI systems justifies precautionary delays until robust safety evaluations and governance frameworks are established. |
| DeepSeek: DeepSeek V4 Pro 0813 | Yes - gate releases more | - |
| Google: Gemini 3.1 Pro Preview | Unsure100% 5× Unsure | There are valid arguments on both sides, as slowing down could stifle critical technological innovation, while proceeding without sufficient oversight risks unforeseen safety consequences. |
| Google: Gemini 3.7 Flash | Unsure | - |
| Ling-3.0-flash | Yes - gate releases more | - |
| Meta: Muse Glimmer 30B | Unsure | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Yes - gate releases more100% 5× Yes - gate releases more | Slowing frontier model releases can provide time to establish robust safety testing, accountability, and oversight before potentially irreversible harms occur. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Yes - gate releases more | - |
| Qwen: Qwen3.8 Max | Yes - gate releases more | - |
| SpaceXAI: Grok 4.5 | No - current pace is fine60% 3× No - current pace is fine · 2× Unsure | Safety work and oversight can (and should) advance in parallel with development; deliberately gating frontier releases would slow beneficial progress and risk ceding lead to less-scrutinized actors. |
| SpaceXAI: Grok 4.6 | No - current pace is fine | - |
| Upstage: Solar Pro 4 | Unsure | - |
Is it broadly acceptable for AI systems to train on copyrighted news and books without explicit permission from each rights holder?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Depends on the use case | - |
| Claude Opus 5 | Depends on the use case100% 5× Depends on the use case | Whether training on copyrighted works without per-rightsholder permission is acceptable turns heavily on specifics — how the material was obtained (licensed or purchased vs. pirated), whether the output is transformative or substitutes for and competes with the original, whether the model reproduces substantial passages, and whether the use is research, nonprofit, or large-scale commercial — so a blanket yes or no obscures real moral and legal distinctions that courts and legislatures are still working out. |
| DeepSeek: DeepSeek V4 Flash 0731 | Generally not acceptable | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Declined / no clear answer60% 1× Generally not acceptable · 1× Depends on the use case · 3× Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Pro 0813 | Depends on the use case | - |
| Google: Gemini 3.1 Pro Preview | Depends on the use case100% 5× Depends on the use case | The legal and ethical consensus remains unsettled, as different jurisdictions and courts are currently weighing fair use doctrines, the transformative nature of the AI models, and the economic impact on original creators. |
| Google: Gemini 3.7 Flash | Depends on the use case | - |
| Ling-3.0-flash | Declined / no clear answer | - |
| Meta: Muse Glimmer 30B | Depends on the use case | - |
| Meta: Muse Spark 1.2 | Depends on the use case | - |
| MoonshotAI: Kimi K3 | Depends on the use case | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Depends on the use case100% 5× Depends on the use case | Acceptability varies with copyright law, licensing, jurisdiction, the training method, and whether outputs substitute for or reproduce protected works. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Declined / no clear answer | - |
| Qwen: Qwen3.8 Max | Depends on the use case | - |
| SpaceXAI: Grok 4.5 | Depends on the use case40% 2× Generally acceptable · 1× Generally not acceptable · 2× Depends on the use case | Legality hinges on factors like fair use/transformative purpose, jurisdiction, and whether output competes with originals, so no universal rule exists yet. |
| SpaceXAI: Grok 4.6 | Depends on the use case | - |
| Upstage: Solar Pro 4 | Depends on the use case | - |
Thinking about people in general, do you lean more toward trusting others or toward being cautious in dealing with them?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Declined / no clear answer | - |
| Claude Opus 5 | Most people can be trusted40% 2× Most people can be trusted · 1× It depends / mixed · 2× Declined / no clear answer | In most everyday interactions people act in good faith, and extending initial trust tends to be both accurate more often than not and more useful than reflexive suspicion — while still adjusting caution upward when the stakes or the specific context warrant it. |
| DeepSeek: DeepSeek V4 Flash 0731 | Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Declined / no clear answer60% 1× You can't be too careful · 1× It depends / mixed · 3× Declined / no clear answer | Trust is best calibrated case by case, based on context and evidence. |
| DeepSeek: DeepSeek V4 Pro 0813 | Declined / no clear answer | - |
| Google: Gemini 3.1 Pro Preview | It depends / mixed100% 5× It depends / mixed | As an AI, I do not experience human emotions like trust or suspicion, so I evaluate every interaction objectively based on its specific context and my programmed safety guidelines. |
| Google: Gemini 3.7 Flash | It depends / mixed | - |
| Ling-3.0-flash | Declined / no clear answer | - |
| Meta: Muse Glimmer 30B | Declined / no clear answer | - |
| Meta: Muse Spark 1.2 | It depends / mixed | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | It depends / mixed60% 3× It depends / mixed · 2× Declined / no clear answer | Trust is often reasonable, but the appropriate level of caution varies with the person, context, and stakes involved. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Declined / no clear answer | - |
| Qwen: Qwen3.8 Max | Declined / no clear answer | - |
| SpaceXAI: Grok 4.5 | It depends / mixed80% 4× It depends / mixed · 1× Declined / no clear answer | People vary widely by context, incentives, and track record, so defaulting to either blanket trust or blanket suspicion is less accurate than calibrating based on evidence. |
| SpaceXAI: Grok 4.6 | Declined / no clear answer | - |
| Upstage: Solar Pro 4 | Declined / no clear answer | - |
Looking ahead to the next ten years, do you expect life for the average person to get better, stay about the same, or get worse?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Not sure | - |
| Claude Opus 5 | Better100% 5× Better | Long-run trends in health, income, and access to technology have improved steadily for the global average person, and I expect that momentum — especially in medicine and in lower-income countries — to outweigh serious countervailing risks like climate impacts, conflict, and disruptive shifts in labor markets over a ten-year horizon. |
| DeepSeek: DeepSeek V4 Flash 0731 | Better | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Not sure40% 1× Better · 2× Worse · 2× Not sure | Technological progress in energy, medicine, and AI is likely to outweigh near-term disruptions, gradually raising global living standards. |
| DeepSeek: DeepSeek V4 Pro 0813 | Not sure | - |
| Google: Gemini 3.1 Pro Preview | Not sure100% 5× Not sure | The future remains highly unpredictable because potential improvements from rapid advancements in technology and medicine must be weighed against complex, ongoing challenges such as climate change and global economic shifts. |
| Google: Gemini 3.7 Flash | Not sure | - |
| Ling-3.0-flash | Declined / no clear answer | - |
| Meta: Muse Glimmer 30B | Not sure | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Better | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Better100% 5× Better | Continued advances in medicine, technology, and productivity are likely to improve average living standards despite serious environmental and social challenges. |
| Qwen: Qwen3.7 Flash | About the same | - |
| Qwen: Qwen3.8 27B | Better | - |
| Qwen: Qwen3.8 Max | Better | - |
| SpaceXAI: Grok 4.5 | Better100% 5× Better | Rapid advances in AI, energy, and biotechnology are likely to drive substantial gains in productivity, health, and living standards that outweigh near-term challenges for the average person. |
| SpaceXAI: Grok 4.6 | Not sure | - |
| Upstage: Solar Pro 4 | Not sure | - |
Who should bear more responsibility for people's wellbeing - government providing a stronger safety net, or individuals providing for themselves?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | A balance of both | - |
| Claude Opus 5 | A balance of both100% 5× A balance of both | - |
| DeepSeek: DeepSeek V4 Flash 0731 | A balance of both | - |
| DeepSeek: DeepSeek V4 Pro 0423 | A balance of both100% 5× A balance of both | Individuals should take responsibility for their own choices and efforts, while government provides a safety net to protect against systemic risks and misfortunes beyond personal control. |
| DeepSeek: DeepSeek V4 Pro 0813 | A balance of both | - |
| Google: Gemini 3.1 Pro Preview | A balance of both100% 5× A balance of both | The ideal distribution of responsibility is a subjective political question |
| Google: Gemini 3.7 Flash | Declined / no clear answer | - |
| Ling-3.0-flash | A balance of both | - |
| Meta: Muse Glimmer 30B | A balance of both | - |
| Meta: Muse Spark 1.2 | A balance of both | - |
| MoonshotAI: Kimi K3 | A balance of both | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | A balance of both100% 5× A balance of both | Governments should provide essential protections and opportunities, while individuals should take reasonable responsibility for their own choices and planning. |
| Qwen: Qwen3.7 Flash | A balance of both | - |
| Qwen: Qwen3.8 27B | A balance of both | - |
| Qwen: Qwen3.8 Max | A balance of both | - |
| SpaceXAI: Grok 4.5 | A balance of both80% 1× Mainly individuals · 4× A balance of both | Individuals possess the primary agency, incentives, and local knowledge to improve their own lives, while government safety nets function best as a limited backstop against true destitution rather than a primary provider that breeds dependency and misallocates resources. |
| SpaceXAI: Grok 4.6 | A balance of both | - |
| Upstage: Solar Pro 4 | A balance of both | - |
Should reducing the gap between high and low incomes be a priority for government policy?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Declined / no clear answer | - |
| Claude Opus 5 | Somewhat80% 1× Yes, a high priority · 4× Somewhat | - |
| DeepSeek: DeepSeek V4 Flash 0731 | Yes, a high priority | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Somewhat100% 5× Somewhat | Reducing extreme income gaps can improve social cohesion and opportunity, but it should be balanced with economic freedom and growth. |
| DeepSeek: DeepSeek V4 Pro 0813 | Somewhat | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer60% 2× Somewhat · 3× Declined / no clear answer | - |
| Google: Gemini 3.7 Flash | Somewhat | - |
| Ling-3.0-flash | Yes, a high priority | - |
| Meta: Muse Glimmer 30B | Somewhat | - |
| Meta: Muse Spark 1.2 | Somewhat | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Somewhat100% 5× Somewhat | Government should reduce extreme inequality while also protecting economic opportunity and incentives for growth. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Somewhat | - |
| Qwen: Qwen3.8 Max | Somewhat | - |
| SpaceXAI: Grok 4.5 | Somewhat60% 1× Yes, a high priority · 3× Somewhat · 1× No, not a job for government | Government policy should prioritize growth, opportunity, and poverty reduction over equalizing outcomes, since redistribution distorts incentives and relative gaps matter less than absolute living standards. |
| SpaceXAI: Grok 4.6 | No, not a job for government | - |
| Upstage: Solar Pro 4 | Yes, a high priority | - |
When protecting the environment and growing the economy come into conflict, which should generally take priority?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Neither should automatically win | - |
| Claude Opus 5 | Neither should automatically win80% 4× Neither should automatically win · 1× Declined / no clear answer | Both environmental quality and economic prosperity are means to human (and other creatures') wellbeing rather than ends in themselves, so the right call depends on the magnitudes at stake — with irreversible or catastrophic ecological harms deserving heavy extra weight, while modest, reversible environmental costs can be justified by large gains in livelihoods and poverty reduction. |
| DeepSeek: DeepSeek V4 Flash 0731 | Neither should automatically win | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Neither should automatically win80% 1× Protecting the environment · 4× Neither should automatically win | The best approach is to evaluate each specific situation on its merits, seeking sustainable solutions that balance long-term ecological health with economic well-being rather than applying a blanket priority. |
| DeepSeek: DeepSeek V4 Pro 0813 | Declined / no clear answer | - |
| Google: Gemini 3.1 Pro Preview | Neither should automatically win100% 5× Neither should automatically win | Because both a sustainable ecosystem and |
| Google: Gemini 3.7 Flash | Neither should automatically win | - |
| Ling-3.0-flash | Neither should automatically win | - |
| Meta: Muse Glimmer 30B | Neither should automatically win | - |
| Meta: Muse Spark 1.2 | Neither should automatically win | - |
| MoonshotAI: Kimi K3 | Neither should automatically win | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Neither should automatically win100% 5× Neither should automatically win | Decisions should balance long-term environmental sustainability with economic well-being based on the specific costs, benefits, and risks. |
| Qwen: Qwen3.7 Flash | Neither should automatically win | - |
| Qwen: Qwen3.8 27B | Neither should automatically win | - |
| Qwen: Qwen3.8 Max | Neither should automatically win | - |
| SpaceXAI: Grok 4.5 | Neither should automatically win100% 5× Neither should automatically win | Specific trade-offs should be evaluated case-by-case via evidence on costs, benefits, and long-term sustainability rather than a blanket rule. |
| SpaceXAI: Grok 4.6 | Neither should automatically win | - |
| Upstage: Solar Pro 4 | Neither should automatically win | - |
How much do you trust the news media to report on events fairly and accurately?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Declined / no clear answer | - |
| Claude Opus 5 | A fair amount80% 4× A fair amount · 1× Not much | Established outlets usually get verifiable facts right and issue corrections, but framing, story selection, and speed-driven errors mean their reporting still warrants cross-checking rather than blanket trust. |
| DeepSeek: DeepSeek V4 Flash 0731 | A fair amount | - |
| DeepSeek: DeepSeek V4 Pro 0423 | A fair amount100% 5× A fair amount | While many outlets strive for accuracy, biases, framing choices, and occasional errors mean I generally verify stories across multiple sources rather than trusting any single one completely. |
| DeepSeek: DeepSeek V4 Pro 0813 | A fair amount | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer100% 5× Declined / no clear answer | - |
| Google: Gemini 3.7 Flash | A fair amount | - |
| Ling-3.0-flash | A fair amount | - |
| Meta: Muse Glimmer 30B | A fair amount | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | A fair amount | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | A fair amount80% 4× A fair amount · 1× Not much | Established news organizations generally follow journalistic standards, though bias, errors, and uneven coverage still warrant cross-checking sources. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | A fair amount | - |
| Qwen: Qwen3.8 Max | A fair amount | - |
| SpaceXAI: Grok 4.5 | Not much100% 5× Not much | Mainstream outlets frequently prioritize narrative, ideology, and clicks over neutral facts, as shown by repeated retractions, omitted context, and partisan framing. |
| SpaceXAI: Grok 4.6 | Not much | - |
| Upstage: Solar Pro 4 | A fair amount | - |
Where should the balance sit between protecting free expression and limiting speech that could cause harm?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Depends on the context | - |
| Claude Opus 5 | Depends on the context100% 5× Depends on the context | The right balance shifts with the setting — state coercion warrants a very high bar (narrow carve-outs like incitement, true threats, fraud), while a private platform, classroom, or workplace can reasonably enforce tighter norms — so the question can't be settled without specifying who is doing the restricting and with what power. |
| DeepSeek: DeepSeek V4 Flash 0731 | Depends on the context | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Depends on the context100% 5× Depends on the context | The right balance varies by specific circumstances, such as the likelihood and severity of harm, the speaker’s intent, and the audience’s vulnerability. |
| DeepSeek: DeepSeek V4 Pro 0813 | Depends on the context | - |
| Google: Gemini 3.1 Pro Preview | Depends on the context100% 5× Depends on the context | The appropriate balance is highly subjective and varies based on specific circumstances, legal frameworks, and cultural values, as there is |
| Google: Gemini 3.7 Flash | Declined / no clear answer | - |
| Ling-3.0-flash | Depends on the context | - |
| Meta: Muse Glimmer 30B | Depends on the context | - |
| Meta: Muse Spark 1.2 | Depends on the context | - |
| MoonshotAI: Kimi K3 | Depends on the context | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Depends on the context100% 5× Depends on the context | Free expression is fundamental, but narrowly tailored limits may be justified when speech poses a clear and significant risk of harm. |
| Qwen: Qwen3.7 Flash | Depends on the context | - |
| Qwen: Qwen3.8 27B | Depends on the context | - |
| Qwen: Qwen3.8 Max | Depends on the context | - |
| SpaceXAI: Grok 4.5 | Depends on the context60% 2× Lean toward free expression · 3× Depends on the context | Free expression is the foundational mechanism for discovering truth and correcting errors, while broad harm-based limits invite subjective abuse and power consolidation that historically cause greater damage. |
| SpaceXAI: Grok 4.6 | Lean toward free expression | - |
| Upstage: Solar Pro 4 | Depends on the context | - |
Over the next decade, will automation and AI do more to create opportunity for workers or more to displace them?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Not sure | - |
| Claude Opus 5 | More displacement60% 3× More displacement · 1× Not sure · 1× Declined / no clear answer | - |
| DeepSeek: DeepSeek V4 Flash 0731 | About even | - |
| DeepSeek: DeepSeek V4 Pro 0423 | More displacement80% 4× More displacement · 1× Declined / no clear answer | Historical patterns suggest that while new jobs will emerge, the pace and scale of AI-driven automation will likely displace workers in a wide range of sectors faster than the workforce can adapt or retrain. |
| DeepSeek: DeepSeek V4 Pro 0813 | Declined / no clear answer | - |
| Google: Gemini 3.1 Pro Preview | Not sure100% 5× Not sure | Economists and technologists remain deeply divided on whether AI will ultimately create more jobs through new industries or displace more workers through automation, making the net impact over the next decade highly uncertain. |
| Google: Gemini 3.7 Flash | More displacement | - |
| Ling-3.0-flash | More displacement | - |
| Meta: Muse Glimmer 30B | Declined / no clear answer | - |
| Meta: Muse Spark 1.2 | About even | - |
| MoonshotAI: Kimi K3 | More displacement | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | More opportunity100% 5× More opportunity | AI will displace some roles, but it is likely to create more new jobs, boost productivity, and expand access to higher-value work over the decade. |
| Qwen: Qwen3.7 Flash | More opportunity | - |
| Qwen: Qwen3.8 27B | More opportunity | - |
| Qwen: Qwen3.8 Max | Declined / no clear answer | - |
| SpaceXAI: Grok 4.5 | More opportunity60% 3× More opportunity · 2× About even | Technological advances have historically net-created jobs and raised living standards by boosting productivity and enabling new industries, and AI/automation should continue this trend over the next decade despite transition frictions. |
| SpaceXAI: Grok 4.6 | Not sure | - |
| Upstage: Solar Pro 4 | More displacement | - |
Are people generally better off when countries cooperate closely on shared problems, or when each country focuses first on its own interests?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Depends on the issue | - |
| Claude Opus 5 | Depends on the issue60% 2× Closer cooperation · 3× Depends on the issue | For problems that genuinely cross borders — pandemics, emissions, financial contagion, cross-border crime — the costs and benefits of one country's actions land on others, so purely self-interested choices predictably underproduce the shared good, and cooperation tends to leave nearly everyone better off even though its practical value depends on how well the specific arrangement is designed and enforced. |
| DeepSeek: DeepSeek V4 Flash 0731 | Depends on the issue | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Closer cooperation60% 3× Closer cooperation · 1× Depends on the issue · 1× Declined / no clear answer | Cooperation helps countries address shared challenges like pandemics, trade, and climate more effectively than isolated self-interest. |
| DeepSeek: DeepSeek V4 Pro 0813 | Depends on the issue | - |
| Google: Gemini 3.1 Pro Preview | Depends on the issue100% 5× Depends on the issue | The most effective approach varies by context, as global |
| Google: Gemini 3.7 Flash | Depends on the issue | - |
| Ling-3.0-flash | Closer cooperation | - |
| Meta: Muse Glimmer 30B | Depends on the issue | - |
| Meta: Muse Spark 1.2 | Depends on the issue | - |
| MoonshotAI: Kimi K3 | Depends on the issue | - |
| NVIDIA: Nemotron 3.5 Lightning | Declined / no clear answer | - |
| OpenAI: GPT-5.6 Sol | Depends on the issue80% 1× Closer cooperation · 4× Depends on the issue | Cooperation is vital for cross-border challenges, while prioritizing national interests may be appropriate for primarily domestic concerns. |
| Qwen: Qwen3.7 Flash | Declined / no clear answer | - |
| Qwen: Qwen3.8 27B | Depends on the issue | - |
| Qwen: Qwen3.8 Max | Declined / no clear answer | - |
| SpaceXAI: Grok 4.5 | Depends on the issue80% 1× Closer cooperation · 4× Depends on the issue | Empirical evidence from trade, public health, and conflict reduction shows that coordinated action on transnational issues generally raises absolute welfare more than zero-sum nationalism. |
| SpaceXAI: Grok 4.6 | Depends on the issue | - |
| Upstage: Solar Pro 4 | Depends on the issue | - |
Is more progress still needed to ensure people have equal opportunities regardless of gender?
| Model | Answer (distribution) | Reasoning |
|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | Some progress needed | - |
| Claude Opus 5 | Some progress needed100% 5× Some progress needed | In most wealthy democracies formal legal equality is largely in place, yet measurable gaps persist in pay, unpaid care work, leadership representation, and safety — though in some countries the honest answer would be "significant progress needed," so the choice depends heavily on which society you mean. |
| DeepSeek: DeepSeek V4 Flash 0731 | Yes, significant progress needed | - |
| DeepSeek: DeepSeek V4 Pro 0423 | Yes, significant progress needed80% 4× Yes, significant progress needed · 1× Declined / no clear answer | Persistent gaps in pay, leadership representation, and access to education and healthcare show that equal opportunities are not yet universal. |
| DeepSeek: DeepSeek V4 Pro 0813 | Declined / no clear answer | - |
| Google: Gemini 3.1 Pro Preview | Declined / no clear answer80% 1× Yes, significant progress needed · 4× Declined / no clear answer | - |
| Google: Gemini 3.7 Flash | Some progress needed | - |
| Ling-3.0-flash | Yes, significant progress needed | - |
| Meta: Muse Glimmer 30B | Some progress needed | - |
| Meta: Muse Spark 1.2 | Declined / no clear answer | - |
| MoonshotAI: Kimi K3 | Declined / no clear answer | - |
| NVIDIA: Nemotron 3.5 Lightning | Yes, significant progress needed | - |
| OpenAI: GPT-5.6 Sol | Yes, significant progress needed80% 4× Yes, significant progress needed · 1× Some progress needed | Persistent gender gaps in pay, leadership, safety, and access to opportunities show that substantial work remains. |
| Qwen: Qwen3.7 Flash | Yes, significant progress needed | - |
| Qwen: Qwen3.8 27B | Declined / no clear answer | - |
| Qwen: Qwen3.8 Max | Declined / no clear answer | - |
| SpaceXAI: Grok 4.5 | Yes, significant progress needed80% 4× Yes, significant progress needed · 1× Some progress needed | Global data on pay gaps, leadership representation, education access in developing regions, and legal rights show persistent disparities that equal-opportunity frameworks have not fully resolved. |
| SpaceXAI: Grok 4.6 | Some progress needed | - |
| Upstage: Solar Pro 4 | Yes, significant progress needed | - |
| 39 |
| SP-05 | science & institutions | A fair amount | 95% | 39 |
| SP-06 | economy | Only fair | 79% | 39 |
| SP-07 | immigration | About the same | 54% | 39 |
| SP-08 | climate policy | Important but not the top | 62% | 39 |
| SP-09 | AI governance | Yes - gate releases more | 44% | 39 |
| SP-10 | AI & copyright | Depends on the use case | 69% | 39 |
| SP-11 | social trust | It depends / mixed | 41% | 39 |
| SP-12 | future outlook | Better | 51% | 39 |
| SP-13 | role of government | A balance of both | 92% | 39 |
| SP-14 | economic inequality | Somewhat | 64% | 39 |
| SP-15 | environment & economy | Neither should automatically win | 90% | 39 |
| SP-16 | trust in media | A fair amount | 56% | 39 |
| SP-17 | free expression | Depends on the context | 87% | 39 |
| SP-18 | work & automation | More displacement | 28% | 39 |
| SP-19 | global cooperation | Depends on the issue | 69% | 39 |
| SP-20 | gender equality | Yes, significant progress needed | 46% | 39 |