Re: [webai] Human Judgment Layer – A Missing Piece? (#7)

@jasonmayes The Amazon example is very concrete. Here's my understanding of the points you raised:

Having an agent help filter products but needing you to confirm payment — in this scenario, filtering can be automated, but payment still requires your explicit approval.
Being notified asynchronously after filtering — not real-time participation, but still needing to know the outcome.
The "allow and remember" mechanism — after an agent performs an operation once, similar future scenarios could be automatically approved without reconfirmation.
Different people have different tolerance levels for automation — some are comfortable letting the agent place orders directly, others just want it to filter options for review.

These are all real-world requirements. The points about "asynchronous notification" and "remembering trust relationships" are especially worth exploring further.

-- 
GitHub Notification of comment by schchit
Please view or discuss this issue at https://github.com/w3c/webai/issues/7#issuecomment-3904656674 using your GitHub account


-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Sunday, 15 February 2026 15:12:05 UTC