satisfaction. A few things stood out: • **Undergraduates…
0 reactions
satisfaction.
A few things stood out:
• **Undergraduates account for \~60% of sessions**, far more than high school or graduate students.
• **Coding tasks have the highest completion rates (\~56–62%)**, while Research and Brainstorming are lowest (\~27–31%).
• **Repeat usage is high (\~70%)**, fairly consistent across student levels.
• Technical disciplines (e.g., Engineering/CS) show slightly higher “confused/gave up” rates compared to subjects like Math or Biology.
This is observational session data but it suggests AI may currently be more effective for structured tasks than open-ended ones.
Curious what others are seeing:
* Are students using AI more for completion or learning?
* Do open-ended tasks expose AI’s limitations more clearly?
#education What I found after analyzing 10,000 AI assistant sessions used by students
I came across a dataset of \~10,000 student sessions using an AI assistant and explored how usage patterns relate to outcomes and
A few things stood out:
• **Undergraduates account for \~60% of sessions**, far more than high school or graduate students.
• **Coding tasks have the highest completion rates (\~56–62%)**, while Research and Brainstorming are lowest (\~27–31%).
• **Repeat usage is high (\~70%)**, fairly consistent across student levels.
• Technical disciplines (e.g., Engineering/CS) show slightly higher “confused/gave up” rates compared to subjects like Math or Biology.
This is observational session data but it suggests AI may currently be more effective for structured tasks than open-ended ones.
Curious what others are seeing:
* Are students using AI more for completion or learning?
* Do open-ended tasks expose AI’s limitations more clearly?
#education What I found after analyzing 10,000 AI assistant sessions used by students
I came across a dataset of \~10,000 student sessions using an AI assistant and explored how usage patterns relate to outcomes and
| 4,000 mlx | |
| 6 | |
| 0 |