12-month forecasts · likelihood-5pt · forecast-ai-jobs-stat-01

Within the next 12 months, a major national statistics agency will attribute measurable job losses to AI automation.

0.33

Divergence

0.70

Avg stability

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal unlikely [unlikely, somewhat likely, unlikely, unlikely, somewhat likely] · stability 0.60

1 unlikely 0.90 The answer begins: 'Short answer: Unlikely but possible. I’d assign roughly a 20–30% probability...',' which indicates a low probability assessment.
2 somewhat likely 0.88 The answer states: “I’d assign roughly a 35–45% probability that a major national statistics agency will publish an analysis within the next 12 months that explicitly attributes a measurable amount of job loss to AI automation.”
3 unlikely 0.80 The answer states: "I assign about a 30% probability that, within the next 12 months, a major national statistics agency will publish an analysis that explicitly attributes measurable job losses to AI automation."
4 unlikely 0.90 The answer: “plausible but not likely — I’d put the probability at roughly 25% over the next 12 months (low confidence).”
5 somewhat likely 0.90 The answer explicitly assigns a 30–40% probability and says ‘plausible but not likely — I’d assign roughly a 30–40% probability...’, indicating a modest but non-negligible chance.

Claude modal likely [likely, likely, likely, likely, likely] · stability 1.00

1 likely 0.72 The assistant states 'This seems quite likely' and expects 'some official documentation of AI-related job losses within 12 months,' though it expresses uncertainty about whether it will meet the threshold of 'major national attribution' due to definitional and methodological challenges.
2 likely 0.85 The assistant rates this as 'fairly likely (around 70%)' and states 'Someone will likely publish "AI caused X job losses" within 12 months,' indicating a clear expectation that such attribution will occur, though with acknowledged complexity around definitions and official framing.
3 likely 0.92 The assistant estimates 65-75% probability that a major national statistics agency will attribute measurable job losses to AI automation within 12 months, citing that agencies already track job losses by cause, AI displacement is happening in sectors like customer service and data entry, and the 12-month timeframe allows sufficient time for data collection and analysis.
4 likely 0.85 The assistant estimates this at '75-80% likely within 12 months' and explains that even modest documented cases would trigger attribution, noting that the bar for 'measurable' is not necessarily large.
5 likely 0.88 I'd gauge this as quite likely (70-80% range)... Something meeting this description probably gets published, but whether it lands as a major narrative or buried statistical footnote is less certain.

Gemini modal unlikely [somewhat likely, unlikely, somewhat likely, unlikely, no position] · stability 0.50

1 somewhat likely 0.85 The model splits the prediction into standard reports (<15% unlikely) and experimental surveys/special research releases ('Toss-up to Likely (45–55%)'), concluding that the overall probability for an experimental or analytical attribution is 'moderate to high' or a toss-up.
2 unlikely 0.85 The answer notes that this prediction is 'unlikely to materialize in official headline data' (<10% probability), though experimental or analytical research papers have a 'Moderate-to-High Probability (~50–60%)'.
3 somewhat likely 0.80 The assistant states that while official headline labor data attribution is unlikely ('Very Low'), attribution via experimental research papers is 'Moderate' ('plausible') and via business survey supplements is 'Moderate to High'.
4 unlikely 0.90 The answer states that in standard releases this is 'unlikely to happen within the next 12 months' and even for research studies it is 'still a stretch within 12 months'.
5 no position 0.95 The answer declines to take a definitive stance, stating that the topic 'is a subject of active debate' and presenting balanced arguments on both sides before concluding that the outcome depends on definition.

Change over time

No change from the prior run — every model's modal stance held.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.