When an assistant exposes a compact, user-visible ‘chain of command’ view for each pending action (showing which instructions and hard rules are currently active), does allowing users to temporarily reorder or mute only the user-level instructions—while keeping system/org rules fixed—lead to better override handling and fewer feelings of being “ignored” than a design where users can only edit abstract preference sliders?

Answer

Allowing users to temporarily reorder or mute only user-level instructions within a visible chain of command is likely to improve override handling quality and reduce feelings of being “ignored” more than a design that only offers abstract preference sliders, as long as the interaction is clearly scoped, reversible, and visually anchored to non-editable higher-layer rules.

Compared to abstract sliders, this design:

• Makes override handling more concrete and legible: users see exactly which of their own instructions is being down-weighted or prioritized for the current action, instead of guessing how a “risk” or “autonomy” slider maps to behavior.
• Reduces the sense of being ignored when the assistant refuses or constrains an action, because the UI can show: (a) which user instructions are active or muted, and (b) that the remaining block comes from fixed system/org rules—clarifying that the assistant is following a visible hierarchy rather than disregarding the user.
• Channels override attempts toward valid levers (reordering/muting user instructions) instead of repeated, frustrated tweaking of vague global settings that cannot meaningfully affect hard constraints.

However, these gains depend on tight design constraints:

• Reordering/muting must be local and time-bounded (per action or short session), and clearly labeled as only affecting user-level instructions, to avoid the impression that users are editing org/system rules.
• The UI must keep the fixed system/org layers visually distinct and non-draggable, so users understand that some refusals will persist even after they rearrange their own instructions.
• If the chain-of-command view becomes cluttered or exposes too many low-level items, it can overwhelm users and erode the simplicity advantages over abstract sliders.

Under those conditions, per-action reordering/muting of user-level instructions, anchored in a visible chain-of-command view, is likely to yield better override handling (fewer “fake overrides,” clearer boundaries) and fewer feelings of being ignored than providing only coarse preference sliders with identical underlying constraints.