Scoring the AI Conversation Process: What a 21-Student Classroom Pilot Can Tell Us
Want Students to Think With AI? Raise the Stakes.

English translation
The author argues that AI conversation logs should be incorporated into assessment and describes a four-week pilot at a Swiss school involving 21 twelfth-grade students: students co-developed a rubric, submitted complete conversations, and 18 reported changing how they used AI. The article later turns to test feedback on the author's company's product; this is not evidence of outcomes from the same study.
The piece may prompt teachers to attend to question quality, critical responses, and who leads the thinking. It is also useful for practicing how to distinguish a small pilot, self-reported experience, and feedback on a commercial product.
Sources and editorial checks
Editorial check:
- Scope
- The pilot took place in Switzerland; the author's product example is from the United States, and the two are clearly distinguished.
- Study design
- Four-week classroom pilot with 21 twelfth-grade students; no control group.
- Sample
- Pilot of 21 students.
- Main findings
- 18 students self-reported changing how they used AI; this figure comes from the author's account and is not equivalent to verified learning gains.
- Funding and conflicts of interest
- The author discloses that AI Friction Labs is their company; the product section describes early customer feedback.
Limitations
- The 21-student pilot had no control group and most measures were self-reported; the author discloses that AI Friction Labs is their company. The Khanmigo NBER paper cited in the article was not obtained in full during this round, so no independent claim about its effects is made here.
Source: EdSurge · edsurge.com