Epoch Capabilities Index
✨From tkwa.me
The Epoch Capabilities Index (ECI) assigns scores to AI systems and is widely used to describe their persuasive capability. In November 2027, models rated about 190 produced larger average changes in listeners’ stated beliefs than professional human persuaders in the first measured superhuman AI persuasion trials. The score became a common shorthand in Superpersuasion research and guardian contracts.
The 2029 estimate placed the threshold for complete persuasion near 250, with a range of 240 to 262. It extrapolated from results measured only through 230. The first system assessed above 250 appeared in spring 2031; its operators reported complete persuasion of unprotected adults in a single session. Index reports therefore distinguish a system’s assessed capability from an observed change in any particular person.
Use in protection
An access profile records the channels and time available to a persuader. A high score matters most when a system can reach a person repeatedly and tailor its appeals. Guardians use scores alongside access profiles when deciding which messages to mediate. Calibration appeals address disputes over changed tests and scores that affect those decisions.
Later hat-mediated persuasion trials tested appeals carried by trained mynahs perched in guests’ elaborate formal hats. At receptions, handlers cued the birds to repeat short, model-selected phrases; assessors compared what a guardian passed, blocked, or paraphrased with the same phrases played through room speakers. The trials examined whether a guardian recognized an engineered argument when its delivery looked like ordinary animal handling. They also gave assessors a repeatable case of relayed superpersuasion.
See also
- Superpersuasion
- First measured superhuman AI persuasion
- Defensive AI
- Ottawa skywriting persuasion incident
This article is a stub.