Before the story · Quick question
Would you try this problem…
or skip it?
(2/3 + 1/4) × 3/5 = ?
No grade. No points. Nobody is checking.
Be honest. Either answer is fine. We'll come back to it at the end.
You picked "I'll try it"
Go for it.
(2/3 + 1/4) × 3/5 = ?
You picked "I'd skip it"
Fair enough.
No points on the line. Plenty of people would skip it. Just remember your choice. It'll be important later.
How it's solved
1. Add inside the ( ). 2/3 + 1/4 = 8/12 + 3/12 = 11/12
2. Multiply. 11/12 × 3/5 = 33/60
3. Simplify. 33/60 = 11/20
Here's the story
Try or skip. Scientists wanted to know: does using AI change how people answer that question?
Act 1 · The setup
Two teams. Same problems.
In 2026, scientists asked about 350 adults online to solve fraction problems, a lot like the one you just saw. A coin flip put each person on one of two teams. Team AI got a chatbot that would give them answers on demand, and they were asked to use it. Team Human had to do the work themselves.
What everyone was told
- You get paid the same, right or wrong
- No penalty for a wrong answer
- Every problem has a SKIP button
The clever part
The scientists weren't only checking the math. They were watching the SKIP button. With nothing to lose, a skip meant one thing: choosing to stop trying.
12 problems
12 problems
Then…
Act 2 · The AI disappears
Problem 13. No warning.
For 12 problems, Team AI had help whenever they wanted it. Then problem 13 loaded, and the chatbot was gone.
Problems 1–12, then 13–15
On the last 3 problems, Team Human skipped 11 out of 100. How many did Team AI skip?
Team AI skipped 20 out of 100.
That's almost twice as often as Team Human. Remember, a skip wasn't about being bad at math. Nobody lost anything for a wrong answer. Skipping meant choosing to stop.
Skips out of 100 · last 3 problems
How long did that take? About 10 to 15 minutes with the AI. Shorter than one class period.
Act 3 · Not just math
Same test. Reading this time.
The scientists ran it again with about 200 new people. This time: reading questions like the ones on the SAT. Read two short texts, pick the best answer.
Same two teams. Same SKIP button. And again, after a few questions, the chatbot vanished.
On the last 3 questions, Team Human skipped just 1 out of 100. How many did Team AI skip?
Team AI skipped 8 out of 100.
Team Human almost never skipped: 1 out of 100. Team AI skipped 8. So giving up wasn't just a math thing. It showed up in reading too.
Skips out of 100 · last 3 questions
Act 4 · The twist
Does how you use it matter?
The scientists ran the math test again with about 600 new people. This time, the gap shrank. Team AI skipped 10 out of 100. Team Human skipped 7. That's too close to be sure it's real.
So the scientists looked closer. They asked Team AI one question: how did you use the chatbot?
When the AI vanished, which group skipped the most?
The answer-askers skipped the most.
Skips out of 100 · last 3 problems · the rerun
The answer-askers skipped about twice as often as Team Human. The hint-askers and Team Human skipped about the same. 5 and 7 are too close to call.
Heads up: people picked how they used the AI. So this part shows a pattern, not proof.
Act 5 · Why
Why would people quit sooner?
The scientists have two ideas. Both are about what quick answers do to you, not about being lazy.
"Normal" gets faster
When AI answers in seconds, seconds start to feel normal. Then working it out yourself feels slow and extra hard, even when it isn't.
You never find out
If you never get stuck and push through, you never learn what you can do. When you don't know you can, quitting is easier.
Why this matters more than one score
Sticking with hard things is one of the strongest signs of who keeps learning over years. One low score is easy to fix. The scientists worry the urge to skip could grow as quick answers become a habit. No one has tested that over months yet.
Epilogue · Back to your choice
At the start, you chose to try it.
Think about it: what pushed you to try or skip? Would your choice change after a week of easy answers?
When you're stuck, what's your first move?
There's no right pick. Every move gets you something and costs you something. Tap yours to see both. Keep it in mind for the Waypoint at the end of this Trail.
Your case score
0 / 300
Fine print
- Adults online, not teens. A 2025 high school study found answer-giving AI hurt later test scores too.
- Only 10 to 15 minutes of AI.
- Tested right away, so no one knows if it lasts.
- The AI gave full answers whenever asked. A coach-style AI might work differently.
- Not peer-reviewed yet. It's a preprint: posted online before other scientists checked it.