Case: "AI-Powered Analytics: Navigating Boston's 311 Data on Your First Day"

Opening:

John is a recent graduate of a public policy program and has joined the staff at the Mayor’s office for the City of Boston. In graduate school, John was known for his poor Excel and his even worse PowerPoint skills. Unfortunately for John, his first assignment on his first day of work is to analyze 311 data, which is in CSV format. It is one year of 2023 data (the file has about 5,000 records). Since he is not good with Excel or PowerPoint, the Mayor’s Chief of Staff has suggested he use ChatGPT4.0 (the plus version available by subscription) to do the work

John’s instructions from the Mayor's Chief of Staff are to provide a summary of the data, identify trends, and highlight any potential problems (e.g., what is going wrong, what can be done better).

After some discussion with the Chief of Staff, John has decided these are some of the potential things he can look at:

• Summary statistics of the data, for example:
  • Total number of cases and distribution of cases by different attributes:
    • department,
    • case type;
• 2-Monthly Trend of 311 Requests: A line chart showing the total number of 311 requests over each month of 2023.
  • Number of Cases by Day of the Week (bar chart) ;
• 3-Top 10 Request Types by Volume: A bar chart displaying the most common request types (from the type column) and their frequencies.
• 4-Request Resolution Time Analysis: How long does it take to close different types of cases
• 5-Geographic Distribution of Requests by City Council District Boston City Council Districts should have approximately equal populations
  • Any recurring issues within specific city council districts
• The Chief of Staff also suggests that you use ChatGPT to prepare a bullet point executive summary of what he finds

<aside> 💡 The data for the assignment is in the appendix below

</aside>

Context:

• Explanation of 311 System and more generally open data:
• Overview of Boston:
• Overview of Mayor Michelle Wu:

Writing a high impact prompt to get the AI to do the work

Conclusion

Appendices