Did we tell
the same story?
Students made pictures without words. Classmates told the stories they saw. We score three things: is it different? Did readers understand it? And did they prefer it?
?How did we score each picture?Three equal parts · Real examples · Click a score bar to filter
What students did
- Make a picture. Show a story without words. Write down what happens. The written story stays private.
- Read ten pictures. Tell the stories you see. We keep the voice recordings private.
- Rank the pictures. Put the pictures you read in order, from best to worst. Confirm your order.
How we score each story
Every score uses a 0–10 scale. Each story gets three scores. Each score has the same weight.
We take the student's private written story. A model turns it into numbers. We compare these numbers with those for the other 347 stories.
We find the closest story. A very close match earns fewer points. The least distinct story gets 0. The most distinct gets 10.
An AI reads each classmate's spoken narration. It selects one of five private story descriptions. We count how often the AI selects the correct story.
Formula: 10 × correct matches ÷ readings. For example, 8 of 10 correct gives 8/10.
Students rank the pictures they read. We combine their orders. We adjust for pictures with few rankings.
Formula: 10 × ranking percentile. A higher rank gives a higher score.
How we calculate the final score
- Find the average. Add uniqueness, understanding and preference. Divide by 3.
- Find the limits. Find the lowest and highest averages across all 348 stories. They are 3.2456 and 8.6911.
- Scale to 0–10. Subtract the lowest average. Divide by the range. Multiply by 10.
Final story score / 1010 × (average − 3.2456) ÷ (8.6911 − 3.2456)
Example: uniqueness 3, understanding 9, preference 6. The average is 6. The final score is about 5.06/10. The lowest final score is exactly 0. The highest is exactly 10.
Final story score distribution
Each bar covers 0.25 marks from 0.00–0.25 to 9.75–10.00. Select a bar to see those pictures.
Select a bar to filter the picture gallery. Select the same bar to clear it.
See two real examples
These pictures have very different uniqueness scores. Open either one to see its closest matches and how its score was calculated.
The nearest match helps explain the uniqueness score. It does not prove that a student copied another story. Common ideas can be told independently.
Each picture opens a full score explanation, its closest stories, and the actual student narration transcripts.
How are the student creator and reader scores calculated?
Creator / 10: 5 × (correct story matches ÷ readings) + 5 × ranking percentile.
Reader / 10: up to 2 marks for completing ten narrations and confirming a ranking; up to 5 marks for understanding the stories; up to 3 marks for matching other readers' rankings.
Student total / 10: 0.8 × creator + 0.2 × reader. The story score uses three criteria. It does not replace the student total.
How do we calculate uniqueness from the nearest story?
We use text-embedding-3-large. It converts each private written story into numbers. We do not embed the picture or the narration.
We use cosine similarity to compare these numbers. A higher value means two stories are more similar.
We compare each story with the other 347. We select the highest similarity. This is the closest other story. We do not average the similarities.
The closest-match values run from 0.315816 to 0.958106. The story with the highest similarity gets uniqueness 0. The one with the lowest similarity gets 10.
Formula: uniqueness = 10 × (0.958106 − closest similarity) ÷ (0.958106 − 0.315816).
Example: closest similarity 0.750 gives uniqueness about 3.24/10. A close match does not prove copying.
Important points
- The instructor accepted all submitted pictures, even when they included written text.
- We compare the meaning of stories, not just matching words. Two different pictures can tell the same story.
- The final story score is a separate analysis. The student total uses the creator and reader tasks.
- The original written stories stay private. Narration transcripts are shown under anonymous reader labels.
Every picture has a story.
Uniqueness, understanding and preference have equal weight. The final story score spans 0–10.
Loading pictures…
Anonymous student results.
Story, creator, reader and student scores all use a 0–10 scale. No name, roll number, or email is shown.
How are student scores calculated?Read the rules and see creator and reader distributions
Each score is out of 10. You can earn creator marks without reading. You can earn reader marks without making a picture.
Creator / 10
Earn up to 5 marks when readers tell the correct story. Earn up to 5 more when readers rank the picture highly.
Creator = 5 × understanding rate + 5 × preference percentile.
Reader / 10
Earn up to 2 marks for narrating 10 pictures and confirming a ranking. Earn up to 5 marks for understanding stories. Earn up to 3 marks when your ranking agrees with other readers.
Reader = completion (2) + comprehension (5) + ranking agreement (3). Missing tasks earn 0 for that part.
Student total / 10
Multiply creator by 0.8. Multiply reader by 0.2. Add them. For creator 8 and reader 6, the total is 7.6/10.
The final story score is a separate score based on uniqueness, understanding and preference.
Reader score distribution
For students who read at least one picture. Each bar covers 0.5 marks.
Creator score distribution
For students who made a picture. Each bar covers 0.5 marks.
Creator counts for 80% of the student total. Reader counts for 20%. Each role is scored out of 10. A score of 0 means no credit for that task.
Loading scores…
| Student | Participation | Uniqueness /10 · distance from closest story | Understanding /10 · correct matches ÷ readings × 10 | Preference /10 · peer ranking percentile × 10 | Final story /10 · scaled mean of three criteria | Creator /10 · 5 × match rate + 5 × preference | Reader /10 · 2 completion + 5 comprehension + 3 agreement | Student total /10 · 0.8 × creator + 0.2 × reader | Details |
|---|
How are the scores related?
The charts show stories and anonymous students. Select a dot to see the source scores. Each axis fits the observed results.
The Pearson r value shows how closely two scores move together. +1 is a perfect positive link. −1 is a perfect negative link. 0 means no linear link. A link does not prove cause and effect.
Four kinds of stories
Student total distribution
Student total marks from 0 to 10. The 40-band final story score distribution is in the help section above.
Reading the colours
Warm colours mean lower story scores. Teal means higher story scores. Colour is not the only cue: each card also shows its score and the exact counts.
The rankings use a Plackett–Luce model. It compares lists even when students saw different pictures. It gives less weight to pictures with few rankings. For a reader's agreement marks, we leave out that reader's own ranking.