Reading & writing · Grades 9–12
A score without a guarantee
An original passage with comprehension questions and a teacher key.
A score without a guarantee
A school trial used software to flag possible missing steps in worked solutions. The tool correctly flagged several incomplete examples but also flagged a concise valid method it had not been designed to recognize. The teacher kept the software as a prompt for review rather than an automatic marking authority. Pupils could explain their reasoning, and final decisions considered the actual mathematics. In the report, the team separated the number of flags from the number of confirmed errors. A warning was useful evidence to inspect, not proof of a mistake.
Try it yourself
Which case showed a false positive?
- It flagged an incomplete example.
- The tool flagged a concise but valid method.
- The teacher reviewed a solution.
- Pupils explained their work.
Why were flags and confirmed errors counted separately?
What role did the teacher retain?
- only counting software messages
- final judgement based on the actual reasoning
- accepting every flag unchanged
- hiding valid alternatives
How did allowing pupils to explain their method improve the process?
Which statement best matches the report?
- Every flag proves wrongdoing.
- No automated tool can ever help.
- A warning identifies something to inspect, not a guaranteed error.
- Concise methods are always invalid.
Suggest a useful performance measure beyond the total number of flags.
Answers and marking notes
Answers and marking notes
- The tool flagged a concise but valid method.
A valid answer was incorrectly treated as suspicious.
- A flag is a prediction or prompt; review can show that some flagged work is valid.
Credit the difference between model output and verified outcome.
- final judgement based on the actual reasoning
The software did not replace mathematical review.
- It provided context and evidence that the tool might not recognize.
Credit review of alternative valid reasoning.
- A warning identifies something to inspect, not a guaranteed error.
The report treats the tool as a limited aid.
- The proportion of flags confirmed as errors, missed errors, or performance across different valid solution methods.
Credit a measure that distinguishes useful detection from mere activity.
Preview material · independent educational review required. Grade bands are approximate, not official curriculum alignment.