NIRODHA COLLECTIVE · DISPATCH
Is an LLM merely a mirror—or an amplifier?
A distortion in the archive can become a default of the next intelligence.
Language models learn from human texts. If those texts frame competition as normal and cooperation as exceptional, models may reproduce or amplify that weighting.
We do not demand agreement. We demand tests using semantically matched prompts, multiple model families, multiple languages, source comparison, and explicit null hypotheses.
Claims & sources
This is a proposition to test; no empirical source is claimed for the proposition itself.
Challenge it. Share it.
Replicate, refute, improve. Do not repeat the claim without its evidence boundary.
Submit counterevidence · Open a discussion · office@artecont.at