Finding a project for an agent

DHUM / DATA 70600 · Class 9

Projects people built with agents

Who	What they built
Jason Heppler, historian	A map of federal lands set aside under a 1937 farm law
Kevin Roose, journalist	An app that suggests school lunches from what is in the fridge
Shawn Graham, historian	A web page that transcribes photos of handwriting
Kevin Schaul, Washington Post reporter	A script that merges federal agencies' spreadsheets on their use of AI
Sarah Bull, historian of print	A map of addresses where pornography was sold in London, 1822 to 1870
Cameron Blevins, historian	A map of how long mail took between US cities in the 1880s
Chris Whong, technologist	A map of public spaces in New York, including plazas owned by private buildings
Benjamin Breen, historian	A game where players live as ordinary people in past societies

Where those projects came from

Sources, processing, presentation

Miriam Posner divides a digital project into three parts

	Sources	Processing	Presentation
Blevins	Mail tables scanned from an 1882 government report	Gemini read about 8,000 travel times from the scans	An interactive map
Bull	Addresses printed in pornographers’ advertisements	Each address placed on a historic map	A map of the shops
Whong	Several city lists of public spaces	Merged into one list	A map app
Schaul	Spreadsheets from each federal agency	Merged into one table	A published story

Past Graduate Center capstones

Data for Power, 2022

A web tool for union organizers that shows which workers no one has talked to yet

Courts of New York, 2026

An atlas of the city’s public basketball courts, made from open data and aerial photos

From Stories to Strategy, 2025

An analysis of oral histories from Jamaica, Queens, written for neighborhood planners

Tracing My Roots, 2025

The author’s family history, told alongside census data on Jewish immigration to New York

The rest are in CUNY Academic Works

Sources of data in New York

NYC Open Data

311 requests, restaurant inspections, street trees, building permits

MTA ridership

Subway and bus riders by day since 2020, by hour since 2025

Municipal Archives

A 1940s tax photo of nearly every building in the city

NYPL Digital Collections

Maps, photos and prints, many in the public domain

Library of Congress

Historic newspapers, photos and maps

US Census

Population, income and rent by zip code and neighborhood

Having Codex interview you

A made-up student

I'm a master's student in data analysis. I don't have a research project, and I'm not sure what I'd use you for. Interview me, one question at a time, about my work, my courses, and what I'm interested in. After a few questions, propose three projects I could build with you in about six weeks.

Codex asked	The student said
What’s one assignment or topic from your courses that you’ve enjoyed?	Mapping farmers markets in a visualization class
What did it make you curious about?	Whether places without markets are where people use food pantries. I work at a pantry network in the Bronx.
What do you do at the pantry network?	I spend two days a month copying visit counts from five pantries into one report.
What tools have you learned?	Excel, some SQL, a little Python
Would your workplace let you use pantry data?	Monthly counts by pantry and zip code, with no names

Three projects Codex proposed

Codex recommended the report builder, because the student can check it against old reports

A monthly report builder

Combines the five spreadsheets into the board report

Pantry visits and farmers markets

A map of pantry visits and farmers markets by zip code

A pantry dashboard

Charts showing which pantries are getting busier

Asking for a wider list

The student then gave Codex the idea skill
The skill has Codex try approaches it wouldn’t start with, such as working backward or borrowing from other fields
Codex came back with 18 ideas, including these

Which pantries and zip codes caused this month’s change
One map of visits shown three ways
Each chart linked to the file it came from
Whether visits grew or reporting got better
A planning game with five made-up pantries
A walk through a few Bronx blocks, with maps and notes

From an interest to a question

These steps come from The Craft of Research, by Wayne Booth, Gregory Colomb and Joseph Williams

Questions to ask about a topic

Kind of question	For the student’s topic
History	How have visits from each zip code changed over the years?
Structure	Which of the five pantries get most of the visits?
Comparison	Do the five pantries draw visitors from different zip codes?
Turned around	Which zip codes have farmers markets and still send many visitors?
What if	What if a market opened next to the busiest pantry?

The book prefers how and why questions to who, what, when and where

The three-part sentence

I am studying pantry visits and farmers markets in Bronx zip codes

because I want to find out whether zip codes with more pantry visits have fewer markets

so that I can help the board understand which neighborhoods a market partnership would reach

Then ask “So what?” until the answer names something the reader cares about

Choosing one idea

Can you check the output yourself?
Are you allowed to use the data?
Can it be done in six weeks?
Will anyone use it, even just you?

The project proposal

Part	The student’s proposal
Question or task	Make the monthly pantry report in an hour instead of two days
Data	Monthly visit counts from five pantries, by zip code, with no names
Course techniques	Saved versions (Class 4), a skill for the monthly run (Class 6), checking each row against its source (Class 7)
Finished thing	A tool that turns the five spreadsheets into the board report and charts
How you’ll check it	Run it on three past months and compare with the reports made by hand

Codex’s criticism of the proposal

The student

I picked the report builder. Here is my proposal. Tell me what’s wrong or missing in it. Don’t rewrite it for me.

Codex

The main gap is that your finished tool starts with five spreadsheets, but your actual work also includes paper sign-in sheets. You need to define whether the one-hour target includes entering those counts.

Codex also asked

Do all five pantries use the same columns and count visits the same way?
Could the hand-made reports have errors too?
How long does the whole monthly run take?

Homework, due in two weeks

Have Codex interview you, then ask for a wider list with the idea skill
Pick one project, yours or Codex’s, and ask Codex what’s wrong with it
Make a one- or two-minute presentation of the project
Slides, a video, a website, drawings, or anything else
State the project in the three-part sentence