widget: suggested pre-selects the answer on ranking and triage (D2-c) — Fable 5.1 trial evidence and a design question

Observation

Live trial on 2026-09-07, Claude desktop Code tab on Fable 5.1, via
elicitation_render_widget → show_widget. attune-forms 0.14.0 served
through attune-ai 16.2.1 (main at f10baa3e1). One five-field form:
ranking (four options, suggested = a full order), triage (six items,
suggested per item), number, date, textarea.

• Rendered HTML: the ranking's ranked list is pre-populated with the
  suggested order and headed "Ranked 4/4 proposed"; each triage row's
  suggested radio is checked with a "suggested" mark. An untouched
  Submit posts the proposals as the answer.
• attune_forms/widget.py documents this as D2-c by design: "submitting
  without touching it posts the proposal — that submit IS the
  confirmation, and the badge is what makes it visible rather than
  silent."
• Trial result: the user reordered the ranking (an explicit act) and kept
  all six triage suggestions unchanged. Both came back through
  elicitation_collect_response (resp-20260907-151254-6baac98d), and
  the form_submitted telemetry row carried the render's instance id.
  The user attested keyboard-only operation.

Question, not a bug report

D2-c holds: the behavior is documented and the badge is visible. What the
trial cannot distinguish is "confirmed after reading" from "accepted by
not touching" for the six triage rows. Options the maintainers might
weigh:

1. Keep as is. Submit is the confirmation (status quo).
2. For a required field carrying suggested, require at least one
   interaction, or an explicit "accept proposal" control, before Submit,
   so the answer is always an act.
3. Keep as is, but record an accepted-untouched flag per field in the
   response or telemetry so a confirm-without-change rate can be
   measured.

Option 3 pairs with the socratic-ambiguity-calibration idea in attune-ai
(asked / inferred / corrected per session); a high confirm-without-change
rate is the ceremony signal.

Filed at the chair's direction (card pick, 2026-09-07). Evidence:
Smart-AI-Memory/attune-ai#2462 and
docs/probes/host-surface-parity/host-native-trials-2026-09-07.md there.

🤖 Generated with Claude Code

Start with attune_forms/widget.py and the elicitation_render_widget → show_widget path, then read the D2-c documentation and the linked host-native trial evidence. The issue presents three unresolved options, so a maintainer decision is needed before implementation; done would be the selected confirmation behavior and corresponding response or telemetry coverage.