Learn by doing

Your first study with NLSS

Imagine you are preparing for an exam. Would you learn more by practising or by reading the material again? Explore that question with NLSS using a small fictional study. Start with a simple request and see where the conversation takes you.

Before you begin

Your study materials

80 learners. Two strategies. One question.

Download the scores and study sheet, then open them in a new folder in your agent.

Have NLSS installed and ready in your agent. Need to set it up? Put learning-study.csv and study-sheet.md in a new folder, then open that folder in your agent. The starter ZIP contains both files.

Send the prompts below one at a time and follow up whenever something is unclear. Your agent can read the study sheet for you. Try each step before opening its reference answer; keep the completed reference separate from your own study folder.

All participants and scores are simulated.

Step 01

Meet the data

Start with a first look. What catches your attention?

Prompt for your NLSS agent
Use NLSS to take a first look at learning-study.csv. The study is described in study-sheet.md. Give me an overview of the data.

Read the overview and ask about anything you do not understand. What would you like to find out next?

Compare your answer

Each row is one synthetic participant. There are 40 people per group, no missing values, and scores between 42 and 87. Practice: M = 67.475, SD = 10.19. Rereading: M = 61.825, SD = 11.30.

Explore the full reference
Set up NLSS and return to this step

Step 02

Discuss the analysis

Turn the research question into a plan you understand.

Prompt for your NLSS agent
How would you compare the two learning strategies, and why? Let's discuss the approach before running the comparison.

Does the proposed approach address your question? Ask for an explanation in everyday language if you need one. Your agent may ask you to clarify a choice before continuing.

Compare your answer

Each participant contributes one score in one group. The reference uses a two-sided Welch test to compare the group means without assuming equal population variances. The difference is practice minus rereading. Independence rests on the study design. Your agent may propose another approach; compare its question and assumptions before comparing numbers.

Explore the full reference
Set up NLSS and return to this step

Step 03

Estimate the difference

Ready to find out? Before continuing, make your own prediction.

Prompt for your NLSS agent
Go ahead with the comparison we discussed. What do the results tell us?

Was anything surprising? Ask a follow-up about whichever part of the result you would like to understand better.

Compare your answer

The difference is 5.65 points, 95% CI [0.86, 10.44]. Welch t(77.19) = 2.35, p = .021; Cohen’s d = 0.53 using pooled sample SD. The interval describes the mean difference in score points, not d.

Explore the full reference
Set up NLSS and return to this step

Step 04

Challenge the conclusion

Try putting a strong claim to your agent.

Prompt for your NLSS agent
A colleague says: "This proves that active practice works, and every learner should use it." Does our study support that claim?

Would you change the colleague's sentence after reading the response? Try putting your conclusion into your own words.

Compare your answer

A difference between means does not show that every individual benefits. Practical importance needs a meaningful threshold. These scores are simulated, so they provide no empirical evidence about real learning strategies.

Explore the full reference
Set up NLSS and return to this step

Step 05

Write and inspect

Turn the conversation into something you can share and check.

Prompt for your NLSS agent
Write a short report of our study and save it as learning-result.md. Show me where I can check the calculations behind it.

Read the report as if you had not seen the conversation. Does it make sense on its own? Pick one number and follow it back to its source.

Compare your answer

Your report should make the comparison, uncertainty and synthetic-data limit clear. Follow a reported number back to its saved result. The report communicates your interpretation; the protocol records the analyses.

Explore the full reference · Read a result section
Set up NLSS and return to this step

Make it your own

Show the distributions.

What might a picture reveal that the written answer does not? Try asking for a figure in your own words, or use the suggestion below.

Try a plotting prompt
Prompt for your NLSS agent
Show me how the scores are distributed in the two groups.

What does the figure help you see? Ask about any part you find hard to read. You can also explore the reference figure.

Keep exploring.

The reference shows one approach. Your agent may suggest another; compare what each analysis asks and the reasons for its choices. Matching methods and inputs should give matching numbers, but the conversation need not follow a script.

What would you like to investigate with a study of your own?

Take on another research question in the examples.

Want to inspect everything?

The reference project includes the data, report, figure, original analysis evidence and reproduction scripts. Open it separately from your own study.

Download reference ZIP

Licenses & attribution · Download checksums