DIALECTICA
Request a demo

Your class debates an AI. You give the grade.

Secondary-school students debate an AI, phase by phase. The system proposes a rubric score and a report. The teacher decides.

The student's arenasample data
  1. 1
  2. 2
  3. 3
  4. 4
04:12
Motion

Smartphones should be banned for the entire school day.

You defendAGAINST · assigned by your teacher

Round 1
Your argument

A whole-day ban treats the tool as the cause. In the two schools that published figures, absences did not move after the ban. What moved was where the boredom went.

  • ✓ Anticipates the objection
  • ✓ Cites evidence
  • ✗ Hasty generalisation
Opponent's response
Concession

I grant you boredom as part of the mechanism.

Rebuttal

It still leaves the filming in the corridors unexplained, and that is the part a whole-day ban does stop.

Evidence demand

Name the two schools and the year, or drop the number and keep the mechanism.

Four phases to go before any score exists. What the teacher gets at the end is a proposal.

The debate screen, rebuilt from the app. Real labels, real fallacy catalogue; the text is sample data.
  • Data stays in the EU

    Database and authentication in Frankfurt. Language model from a European provider.

  • No student email

    Students join with a class code. We never ask for an address.

  • No third parties

    Student pages load no outside service. Not even analytics.

  • The teacher decides

    The AI's score is a proposal. A person sets the grade.

Privacy, security and compliance

Assign it. They debate. You assess.

You assign

Pick a motion, a protocol and a deadline. Students join with a six-character code: no email, nothing to install.

  • Short
    1. 1Exposition
    2. 2Rebuttal
    3. 3Closing
  • World Schools
    1. 1Exposition
    2. 2Rebuttal
    3. 3Defence
    4. 4Closing
  • Karl Popper
    1. 1Exposition
    2. 2Cross-examination
    3. 3Answering
    4. 4Rebuttal
    5. 5Closing

They debate

Phase by phase, on the side they were given, against an opponent that argues back and names the fallacies it finds.

Dialectica
class 9B · Ada Byron State School

Hi, Giulia.

You have 1 assignment to do. Due Friday 20 March.

Your assignments
TO DOdue 20 March

Smartphones should be banned for the entire school day.

You defendAGAINST · assigned by your teacher
Start →
DONE8,0/10

Cities should close their historic centres to private traffic.

See reportSee debate
The student's screen. They join with a class code.

You assess

A score per rubric dimension and a draft report, with the reasoning behind both. Change anything, then export one PDF per student.

Ada Byron State School · class 9BStudent: Giulia Ferraro · 12/03/2026
Dialectica
Motion

«Smartphones should be banned for the entire school day.» Role: against, World Schools, 4 phases

Content30/40Style29/40Strategy13/20Total72/100
Grade7.0Level: Intermediate
Strongest point

«If the ban only applied during lessons, phones would go straight back into students' hands at break, and break is when the videos filmed on the sly get made.»

You took the opposing proposal and showed its concrete consequence, instead of repeating your own claim.

To improve

«Banning things has never worked, everybody knows that, and any teacher will tell you the same.»

«Everybody knows» is not evidence: the listener can deny it at no cost, and you are left with nothing to stand on.

Rhetorical profile

From the debate analysis, not a grade.

Reasoning78Rebuttal62Credibility71Engagement69Clarity84Timing66
Teacher's signature:

World Schools-style score (training) · Final assessment set by the teacher.

The exported report: a black-and-white A4.

The AI proposes. You decide.

The AI's figure stays in a read-only column, every edit you make is recorded, and when the model is unsure it offers a range, not a number.

Report for the studentReport preview

The student reads this text only after you confirm it. You can rewrite any part, or drop it from the report. If you leave it untouched, they get the proposal as it stands.

1AssessmentTwo paragraphs, addressed to the studentRestore AI proposalDon't include in report
You held your position through all four phases, and in the Defence you answered your opponent's point instead of repeating your own. That's the hardest part of the format, and you pulled it off.

Where you lose ground is on evidence: two arguments out of five rest on an “everybody knows” that the listener can reject at no cost.
2Strongest pointQuote and why it worked
Quote
If the ban only applied during lessons, phones would go straight back into students' hands at break, and break is when the videos filmed on the sly get made.
Why it worked
You took the opposing proposal and showed its concrete consequence, instead of repeating your own claim.
6Rhetorical profileNumbers from the analysis, not editable
Rebuttal62
Credibility71
Engagement69
Clarity84
Timing66
Include the profile in the student's reportDebate analysis, not a grade. The student will see it labelled as such.
The grading screen: the AI's proposal sits beside your pen, not in its place.
  • Review thirty debates already worked up, instead of marking thirty from scratch.
  • A black-and-white A4 that survives the photocopier.
  • Accessibility settings per student: extra time, and dictation kept out of the score.

The objections, at full strength.

The strongest arguments against this product, answered honestly. Sometimes the honest answer is “we don't have that figure yet”.

“An automatic score anchors the teacher.”

It does, measurably. So the figure sits apart from the grade, uncertainty turns it into a range, and we will report whether that works.

Frontiers in Psychology, 2026 · n = 214
“Show me your agreement figure against real teachers.”

We don't have one yet. The measurement is built and collection with schools is starting. Even expert annotators agree only weakly on argument quality.

Wachsmuth et al., EACL 2017
“You sell in five languages a system tuned in one.”

Correct. So no figure travels: every result we publish stays tied to the language and rubric it was measured in.

Fu & Liu, EMNLP 2025
All the literature, including what it does not support

Questions teachers ask first

No. It proposes a score per dimension and drafts the report. A person types the grade, and every edit is recorded.

Want to see it with one of your classes?

An annual licence per school, with the price list sent on request.