AI Rules: steering entry text, citations and judging guidelines

Q: AI Rules: steering entry text, citations and judging guidelines

AI Rules Found at Admin → AI Rules , with three sub-pages: Entry Text Rules : steers the AI-drafted finalist-listing text (see howto-finalists-winners for how the text itself gets generated). Set rules once for the whole program, or, if you're using Category Types, set different rules per type (e.g. a different tone for "Achievement" categories vs "Event" categories). Citation Rules : steers tone, length and style for AI-drafted winner citations (the short lines read out at an awards ceremony). Judging Guidelines : the standard judges' comments are checked against. Set the guidelines themselves plus a few good and bad example comments; the AI uses these to judge whether a submitted comment is substantive enough or too generic ("great entry!" with nothing specific), flagging generic ones back to the judge before they can submit. This is what powers the comment-quality check mentioned in howto-judge-scoring . None of these rules are required (leave them blank and the AI uses sensible defaults), but they're useful if your program has a house style or a specific standard you want judges/AI-generated text to match consistently.

Question

Accepted Answer

AI Rules

Found at Admin → AI Rules, with three sub-pages:

Entry Text Rules: steers the AI-drafted finalist-listing text (see howto-finalists-winners for how the text itself gets generated). Set rules once for the whole program, or, if you're using Category Types, set different rules per type (e.g. a different tone for "Achievement" categories vs "Event" categories).

Citation Rules: steers tone, length and style for AI-drafted winner citations (the short lines read out at an awards ceremony).

Judging Guidelines: the standard judges' comments are checked against. Set the guidelines themselves plus a few good and bad example comments; the AI uses these to judge whether a submitted comment is substantive enough or too generic ("great entry!" with nothing specific), flagging generic ones back to the judge before they can submit. This is what powers the comment-quality check mentioned in howto-judge-scoring.

None of these rules are required (leave them blank and the AI uses sensible defaults), but they're useful if your program has a house style or a specific standard you want judges/AI-generated text to match consistently.