Question Sets

Experimental

Question sets and the quiz engine are currently an experimental feature. Their behavior may change.

A question set holds the questions a quiz asks and the rules the run follows. A game only ever sends a character id and a set id, so everything about how the quiz behaves is decided here rather than in your code.

Open Question Sets from the dashboard sidebar.

Plan requirement

Question sets and the quiz engine require a Team plan.

Creating a set

Click Create Question Set and give it a name
Click Add Question for each question you want to ask
Copy the set id from the card's action menu, or from the URL of the set

A set holds up to 10 questions.

Question types

Pick the Answer Type on each question.

Type	How it is judged
Free Text	The host reads the answer and compares it to the expected answer you wrote, plus any rubric.
Multiple Choice	Two to six options, one marked correct. Graded against the option you marked.
True / False	Fixed True and False options. Mark which one is correct.

Choice questions are graded without asking the host for a verdict. When a player's words do not name an option outright, for example answering "yes" to a true or false question or naming the option in another language, the host is asked only which option they meant and the comparison to your answer key still happens in code.

What you write on a question

Field	What it does
Question Prompt	The question itself. The host asks this, translated into the character's language if it speaks another one.
Expected / Correct Answer	The reference answer for a free text question. Choice questions derive it from the option you mark.
Scene keyword	A stable name for the scene this question needs, such as `SHOW_RED_APPLE`. Sent to the game with the question so it can key its scene on something readable instead of a generated id.
Reasoning Facts & Context	Background the host uses to judge edge cases and explain an answer. Answers you list here as accepted, partly right or wrong are graded that way. Never read out as the answer.
Hint	A clue, given when a player asks for one, is stuck, or gets it wrong. The host leads them towards the answer with it and never states the answer itself.
Custom Grading Rubric	Free text only. Decides what counts as correct or partial for this question, in the language the player speaks. Useful for judging pronunciation, where near misses have to be spelled out.
Points	What a correct answer is worth.
Allow partial credit	Free text only. Lets the host award a partial verdict, worth the partial points you set.

A wrong answer teaches rather than reveals

While the player still has an attempt left, the host never says the answer: it uses your hint and background facts to lead them towards it. Once the attempts are spent, it tells them what the answer was and why, so nobody moves on still believing the wrong thing.

How a run plays

Open Configure Host & Flow to set the rules. These are what make the same two ids drive a coaching tutor that lets a child try again and a game show that never does.

Setting	Effect
Question order	`In order` or `Shuffled`. A game can choose its own for a run.
Questions per run	How many of the set one run serves. Leave it empty to serve them all. A game can choose its own for a run.
After a wrong answer	`Move on`, `One more try`, or `Until they get it`. Single player runs only: with several players a round is judged once and the run moves on.
Ask before trying again	The host offers another go and waits for a yes. Off, and it simply asks the question again.
A second try still scores	Whether a replayed question can move the leaderboard. Off, and a replay is played for the learning rather than the points.
Give a hint when asked	Whether the host may offer the hints you wrote. Off, and the game is never told a clue exists.
Opening words	The host's welcome, before the first question. `Say it exactly` speaks them word for word; `In own words` makes them a note the host performs. Leave empty to open on the first question.
Closing words	How the run ends, with the same two modes. Leave empty to end on the host's last reaction. The host is told the scores on the last question, so a note such as "announce the winner" names the right one.

Opening and closing words are spoken exactly as you wrote them, in your own language, rather than paraphrased.

Host delivery

How the host delivers a result decides what the host says once an answer has been judged. It never decides the verdict or the points: those come from the question's own answer, rubric and scoring.

Mode	Delivery
Game Show	Puts the question, takes the answer, says the verdict and the points out loud, moves on.
Natural	Reacts in character and never says a number. Your UI shows the score.
Coaching Tutor	Teaches rather than tests: why an answer was wrong, or what made it right. Says no numbers.

Custom Host Instructions add guidance for the host on top of the character's own personality, for this set only.

The character you pick when the quiz runs supplies the host's name, personality, language and voice. How it hosts is this set's business, not the character's: pick the mode above. Give the character a Language node and it runs the whole quiz in that language, whatever language the set is written in.

Testing a set

Test Set opens the playground. It runs the same engine the public API runs, so a question graded here grades the same way in a game.

Pick a question and a host character
Add simulated contestants, type or tap their answers, and post them one at a time
Evaluate Round scores the round and shows the host's reaction and each verdict
Preview Start and Preview Finish play the opening and closing sequences

Playground runs count against your daily editor allowance, the same as the character chat preview.

Taking it into your game

The playground and your game run the same engine. Once no question shows an Incomplete badge, the set is ready: pass its id and the host character's id to

the Web API quiz endpoint
the Web SDK, with <neocortex-quiz> or QuizAgent