Using AI to improve the environmental and nutritional efficacy of food subsidies

Mentor Areas

Value Proposition for Undergraduate students:

Learning Opportunities: A chance to work with sophisticated data, understand the intricacies of field data collection, and potentially delve deep into building machine learning and optimization models.

Real-world Application: Exposure to a project with tangible implications for improving health and sustainability outcomes in developing countries.

Mentorship and hands-on field research experience: Working alongside experienced researchers, especially in executing field data collection processes.

Possibility of Extension: Students who demonstrate exceptional commitment and skill may be offered the chance to extend their involvement, potentially assisting in algorithm development and other advanced tasks.

Description:

This study involves extensive data collection processes, alongside coding and statistical analysis. We believe involving undergraduate researchers with strong organizational and technical skills will add tremendous value to the research. Concurrently, this collaboration offers undergraduates a unique research experience. This goes beyond some of the more common tasks such as managing bibliographies or editing tables. These tasks involve:

Research on Developing Carbon Footprint Scores:

We have established a list of about 15,000 SKUs of food products in our database. In an earlier research project, we identified the macro and micro nutritional values of each product. This project involves collaboration in the development of a comprehensive carbon footprint score database for these 15,000 unique food products. This task will involve using multiple sources such as i) using existing databanks for gathering estimates of each food ingredient's carbon footprint, then aggregating these ingredients for packaged items, ii) using carbon footprint calculators, and iii) collecting supplementary data on the field.

Remote Learning Opportunity to Conducting Hands-on Field Experiments:

Several stages of this research project will involve the undergraduate research assistants to build partnerships and develop, organize, and manage field projects with our local team in the field. These projects will entail leading a team to gather additional data essential for data collection and analysis. For example, to obtain a more accurate carbon footprint estimate of certain products, the research assistant may select a group of products, design a data entry platform (often using Qualtrics or Knack.com), and formulate a data collection protocol for the field team to exercise. The exact nature of these tasks will depend on the course of the data collection process but some examples are as follows:

Having our field team take photographs of product packages to glean more product details, then enter these details into the platform designed by the undergraduate RA. Another task might be to conduct data-quality assessments using methods like deploying mystery shoppers. The research assistants, in collaboration with our field team, will pinpoint which stores to monitor closely, send out shoppers with predefined shopping lists, and verify the accuracy of the data collected on-site. This procedure is regularly repeated to maintain the integrity of our data collection.

Data Collection and Cleaning:

Assisting in gathering transaction data, survey responses, and other relevant information. Cleaning and preparing the data for analysis, ensuring its accuracy and reliability.

Data Analysis:

Assisting in analyzing the data using statistical tools and software (Python or STATA). Collaborating with the team to decipher patterns, derive insights, and asses the efficacy of the proposed food subsidy programs.

Note that the data analysis in this project involves building an AI-backed algorithm to predict the nutritional and environmental effects of food subsidies and develop an optimization model to identify the optimal subsidy program. Depending on the technical skills of the undergraduate RA and the progression of their involvement, there is an opportunity to be involved in advanced mathematical modeling and programming.

Preferred Qualifications

Coding in Python (and/or Stata). Main tasks include data cleaning, analysis (running linear regression models and statistical tests), graph and table preparation.

Basic understanding of statistics and/or econometrics. 

Students from the computer science, statistics, economics are a good fit, but we also encourage applicants from other engineering departments, public policy/political science and Wharton students with strong coding background.