AI Score

Every question gets a quality score from a built-in evaluator. The criteria are public and shown in full inside the app.

Why score a question?

PingBack can generate follow-up questions based on how the audience reacted. The question this raises is simple: is the follow-up actually sharper than the original, or just different? The evaluator scores both against the same criteria so you can see whether anything improved.

You can also score any original question. Scoring is available on every card, not only after a follow-up is generated.

Three criteria, 0 to 10 each

The score has three parts. The total is their average.

Provocation
Does the question make it hard to just agree and move on? A high score means it names something the team has been treating as settled. A low score means it can be answered politely without anyone changing their mind.
Premise
Does the question challenge the assumption behind the idea, not just the idea itself? A high score means it questions the frame. A low score means it accepts the frame and stays on the surface.
Brevity
Is it one tight sentence with no filler? A high score means it is short, direct, and free of workshop language ("How might we...", "Let's explore...", "In what ways could..."). A low score means it is padded or softened.
Try it

Pick a question to see how each criterion is scored and why.

Quality score
Which step are we keeping because we built it, not because users need it?
Provocation
9
Premise
8
Brevity
10
Total (average of three axes)
9.0 / 10
Comparing the original and the follow-up

When you score a follow-up question, PingBack scores both the original and the follow-up at the same time and shows them side by side. You can see immediately whether the score improved.

Original
Which step are we keeping because we built it, not because users need it?
Total 9.3 / 10
Follow-up
What would have to be true for step 3 to be the part new users are most grateful for?
Total 8.0 / 10
The follow-up scored 1.3 points lower than the original.
A lower score does not disqualify the card. Sometimes a different angle is worth more than a higher number.
How to use it

Open any presentation session. On any question card, click Eval in the top-right corner. The scores appear instantly. On follow-up cards, both the original and the follow-up are scored at the same time and shown side by side.

The full criteria are printed at the bottom of every score panel. Every number is explained.

Is the evaluator independent?
Yes. Scoring is a separate AI call with its own prompt, independent from the one that wrote the question. This prevents the AI from grading its own output with the same mindset it used to write it.
Questions or issues?

Visit the Support page or write to hello@marcosrezende.com.