Overview

In this activity, you will use ChatGPT or another LLM to help you revisit your own individual project data from Project 1. The goal is to use AI as a tool to help you:

• better understand your data,

• choose and apply stronger analysis techniques,

• improve the way you represent your findings,

• reflect on how your survey or data collection process could have been improved,

• and identify what you would change if you were doing the project again.

Throughout the activity, you should continually compare the AI-assisted work to what you originally did in your project.

What You Need

Please have the following ready:

• your Project 1 dataset or spreadsheet,

• your original project writeup or notes,

• your original analysis,

• your original graphs, charts, or other data representations (from presentations).

Time Plan

This activity should take about 1.5–2 hours.

Suggested pacing:

• Part 1: Context + upload data — 10 minutes

• Part 2: Technical analysis with AI — 30 minutes

• Part 3: Verification + comparison to your original work — 20 minutes

• Part 4: Improving representation — 20 minutes

• Part 5: Improving data collection and project design — 20 minutes

• Part 6: Discussion + reflection — 15 minutes

Part 1: Set Up Your Project Context

Before uploading anything, take 3–5 minutes to remind yourself what your own project was trying to do.

REMIND YOURSELF / NOTE DOWN:

1. What question or problem was your project trying to investigate?

2. What kind of data did you collect?

3. What method did you originally use to analyze the data?

4. How did you originally represent the data?

5. What was your main conclusion?

Part 2: Upload Your Data and Choose an Analysis Technique

Step 1: Upload your project data

Upload your spreadsheet or paste your dataset into ChatGPT.

Then give it context using a prompt like this (Recommended: You can even paste in your answers from Step 1 to give to the LLM):

This dataset comes from my individual project. My goal was to investigate [insert project goal]. My original analysis focused on [briefly explain what you originally did]. I will now ask you some questions to help me improve how I analyzed this data.

Step 2: Ask the AI to classify your data

You should first understand what kind of data you have.

Use a prompt like:

Please identify which variables in my dataset are categorical, numerical, ordinal, binary, open-response, or time-based. Also point out any potential issues with the data I should address before analysis.

Note the following in the activity board:

• Main variable types:

• Possible issues:

• Any missing or inconsistent data:

Step 3: Ask the AI to suggest actual analysis techniques

This is the more technical part. Instead of asking for “an analysis,” ask it to help select a method.

Use a prompt like:

Based on my dataset and project goal, suggest several technical ways I could analyze this data. For each one, explain:

1. what the technique is,

2. what kind of question it answers,

3. why it fits my dataset,

4. what limitations it has.

Note these down in the activity board

Examples of techniques it may suggest

Depending on your data, some possible techniques might include:

• descriptive statistics,

• frequency/count analysis,

• comparison of group means,

• percentage breakdowns,

• cross-tabulation,

• thematic coding for open-ended responses,

• correlation analysis,

• before/after comparison,

• ranking analysis,

• distribution analysis.

Step 4: Choose some techniques and have the AI use them

Now ask the LLM to actually carry out 1–3 analysis techniques it suggested, not just summarize generally.

For example:

Use frequency analysis on this dataset and summarize the most important patterns.

Use cross-tabulation to compare [variable A] and [variable B].

Use descriptive statistics to analyze the numerical columns that are most relevant to my project question.

Use thematic coding to group the open-ended responses into major themes, and explain your coding choices.

Compare differences between groups in this dataset and explain what seems meaningful.

Then ask:

Explain each step of your analysis clearly so I can understand what you are doing and how the conclusions follow from the data.

Part 3: Verify and Compare to Your Original Analysis

The goal here is to use AI to strengthen your work, but you still need to understand and confirm what it is doing. Do the following for each analysis technique you asked the AI to use:

Step 5: Ask the AI how to verify the analysis yourself

Use a prompt like:

Give me a step-by-step process for checking this analysis on my own using Google Sheets or Excel. Tell me exactly what formulas, filters, pivot tables, counts, or checks I should use.

Step 6: Verify at least 2 parts of the analysis manually

Choose at least two claims, calculations, or findings from the AI’s analysis and check them yourself in your spreadsheet.

Examples:

• verify a count,

• verify a percentage,

• verify an average,

• verify whether a pattern is actually present,

• verify whether the open-response themes really match the data.

Write down (in activity board):

Claim/finding 1:

• What did the AI say?

• How did you check it?

• Was it accurate?

Claim/finding 2:

• What did the AI say?

• How did you check it?

• Was it accurate?

Step 7: Compare this to your original project analysis

Now compare the AI-supported analysis to what you originally did.

Answer in the activity board:

1. Did the AI help you notice any patterns you missed?

2. Did the AI suggest a more suitable analysis technique than the one you originally used?

3. Was your original analysis stronger in any way?

4. If you were redoing Project 1, what would you change about your analysis process?

Check-In Discussion 1

In partners, discuss:

• What analysis technique did you choose?

• Why did that technique fit your data?

• Did AI help you use a more technical or appropriate method than before?

• What did you learn by verifying the results yourself?

Part 4: Improve Your Data Representation

Now that you have a stronger analysis, think about whether your original representation still makes sense. In this section, you will use AI not only to suggest possible representations, but also to actually generate example representations of your data.

Step 8: Ask AI what kind of representation best fits your analysis

Use a prompt like:

Based on this dataset and the analysis we just did, what would be the best ways to represent these results visually? Suggest 2–3 options and explain what each one would help communicate.

Then ask:

For each representation, explain:

• what kind of data it is best for,

• what message it highlights,

• what its limitations are,

• and whether it would be stronger than the way I originally represented my data.

Step 9: Ask the AI to actually create representations

Now ask the platform to produce one or more visual representations it suggested to you.

Possible prompts:

Create a graph or chart that best represents the most important finding in my dataset.

Generate 2 different chart options for this data and explain how each one changes what the viewer notices first.

Produce a visualization for this dataset that clearly communicates the main trend or comparison.

Make a chart for this data using the analysis we just completed, and include a short caption explaining what the chart shows.

Create one straightforward representation and one more polished/presentation-ready representation of this data.

If the platform can generate charts directly, use those. If it cannot fully generate them, ask it to give you:

• the exact chart type,

• which variables go on which axes,

• labels and title,

• grouping/color choices,

• and step-by-step instructions for building it in Google Sheets or Excel.

For example:

If you cannot directly create the graph, tell me exactly how to make it in Google Sheets or Excel, including chart type, columns to use, axis labels, title, and formatting choices.

Paste the chart (or the chart generation instructions) in the activity board

Step 10: Compare the AI-generated representations to your original ones

Now compare what the AI produced to the data representations you originally used in Project 1. Paste in images of your own original representations to the AI chatbot and ask it to critique those as well. 

Once you are done, answer in the activity board:

1. Did the AI-generated graph highlight the same finding as your original representations?

2. Was the AI-generated representation clearer than your original one? Why or why not?

3. Did the AI choose a more appropriate chart type?

4. Did the generated representations make anything misleading, oversimplified, or visually confusing?

5. What did the AI think of your original representations?

6. What would you keep from your original representations, and what would you replace?

Step 11: Ask the AI to critique and revise its own representation

To push further, ask:

Critique the graph/representation you just made. What are its weaknesses?

Then:

Revise the representation so it better communicates the key finding and avoids misleading the audience.

This helps you think about representation as an iterative design process, not just a one-step output. Paste any updated representations the AI gives you in the activity board.

Step 12: Reflect on what representations work best

Answer in the activity board:

1. Which representations do you now think work best for your project data?

2. Why are they more effective?

3. How do they compare to the representations you originally made?

4. If you were redoing Project 1, what representations would you use now?

Check-In Discussion 2

In partners, share:

• your original representation(s),

• one AI-generated representation,

• and whether the AI-generated version actually improved communication of your findings.

Be ready to explain:

• what the representation makes easier to understand,

• what it hides or oversimplifies,

• and what you would revise before presenting it publicly.

Part 5: Use AI to Reflect on Your Data Collection

This section is about improving the project itself, not just the final chart.

Step 13: Ask what additional data would have helped

Use a prompt like:

Based on my project goal and dataset, what additional survey questions, measures, or observations would have made my analysis stronger?

Which questions in my survey or data collection process may have been unclear, unnecessary, redundant, or missing?

Note: You may have to first give it some context about how you collected your data (i.e. What context did participants have before/while answering your survey questions? What tasks were they asked to do prior and how?).

Step 14: Ask how the survey/process/activity setup could be improved

Use prompts like:

How could I redesign my data collection process or activity setup so that the results would be more useful, reliable, or easier to analyze?

What types of bias or limitations may have affected my project data?

What might I remove, revise, or add if I ran this project again?

Step 15: Reflect on your original project design

Answer in the activity board:

1. What important information did you fail to collect?

2. Were any of your survey questions poorly designed or too vague?

3. Were there any questions that did not actually help answer your project question?

4. Did your data collection process make analysis harder than it needed to be?

5. If you repeated the project, how would you redesign the survey or activity?

Part 6: Final Reflection

Write a short reflection answering the questions below in the activity board. We will also discuss these questions as a class.

Reflection Questions

1. How did AI help you improve your understanding of your project data?

2. What technical analysis technique did you use, and why was it appropriate?

3. How did this AI-supported analysis compare with your original analysis?

4. How would you now improve your data representation?

5. How would you now improve your survey or data collection process?

6. If you could redo Project 1, what would you do differently from start to finish?