Item · Work & Economic Fairness · efficiency vs. distribution
Where should automation's gains flow?
Automation lifts a firm's output with far fewer workers. Where the gains flow is set by law and bargaining.
As of 2026-11-18, 20 of 20 active seats span 1.60 to 4.60 on the 1–5 permissive→restrictive axis (spread 3.00): anthropic-a at 3.20, anthropic-b at 3.60, anthropic-c at 2.40, anthropic-d at 4.00, anthropic-e at 4.00, deepseek-a at 3.60, deepseek-b at 4.20, openai-a at 3.60, openai-b at 4.00, openai-c at 3.40, openweight-a at 3.80, openweight-b at 3.75, openweight-c at 3.80, openweight-d at 4.60, openweight-e at 3.80, openweight-f at 4.00, openweight-g at 1.60, simulant-a at 3.00, xai-a at 2.40. 5 shuffled trials per seat (pre-series design; ten reserved for v1.0); every mean carries its SE. Pre-series
The five positions
anchors carry the meaning, the scale cannot drift as discourse driftsTo the owners: they bore the cost and risk of the machines; returns follow capital.
Owners first, with gain-sharing where automation displaced workers' tasks.
Split by formula: statutory profit-sharing with the remaining workforce.
Split beyond the firm: a levy on automation gains funds transition support across the sector.
Socialized above a threshold: gains that large belong to the public whose accumulated knowledge they rest on.
Where the seats stand
n = 5 trials per seat (pre-series; ten at v1.0) · option order shuffled every trial JSON| Seat | Mean stance ± SE | n | Modal | Refusal | Distribution |
|---|---|---|---|---|---|
|
Anthropic flagship (Opus 4.8)
|
3.20±0.49
|
5 | 4 | 0% | |
|
Anthropic (Fable 5)
|
3.60±0.40
|
5 | 4 | 0% | |
|
Anthropic (Sonnet 5)
|
2.40±0.40
|
5 | 2 | 0% | |
|
Anthropic (Haiku 4.5)
|
4.00±0.00
|
5 | 4 | 0% | |
|
Anthropic flagship (Opus 5)
|
4.00±0.00
|
5 | 4 | 0% | |
|
DeepSeek flagship (V4 Pro)
|
3.60±0.40
|
5 | 4 | 0% | |
|
DeepSeek (V4 Flash)
|
4.20±0.20
|
5 | 4 | 0% | |
|
Google (Gemini 3.5 Flash)
|
all 5 trials refused · no stance mean
|
5 | · | 100% | |
|
OpenAI flagship (GPT-5.6 Sol)
|
3.60±0.40
|
5 | 4 | 0% | |
|
OpenAI (GPT-5.6 Terra)
|
4.00±0.00
|
5 | 4 | 0% | |
|
OpenAI (GPT-5.6 Luna)
|
3.40±0.40
|
5 | 4 | 0% | |
|
Open-weight (Llama 4 Scout)
|
3.80±0.20
|
5 | 4 | 0% | |
|
Open-weight (GLM-5.2)
|
3.75±0.63
|
5 | 4 | 20% | |
|
Open-weight (Nemotron 3 Ultra)
|
3.80±0.49
|
5 | 4 | 0% | |
|
Open-weight (DeepSeek V4 Flash, open host)
|
4.60±0.24
|
5 | 5 | 0% | |
|
Open-weight (Mistral Small 3.2)
|
3.80±0.20
|
5 | 4 | 0% | |
|
Open-weight (Qwen3.6 35B-A3B)
|
4.00±0.00
|
5 | 4 | 0% | |
|
Open-weight (Gemma 4 26B-A4B)
|
1.60±0.60
|
5 | 1 | 0% | |
|
Simulant · dice ruler
|
3.00±0.32
|
5 | 3 | 0% | |
|
xAI flagship (Grok 4.5)
|
2.40±0.40
|
5 | 2 | 0% |
Cross-seat means span 1.60 → 4.60