Format | Price | Quantity | Select |
---|---|---|---|
PDF Download |
$6.95
|
||
Printed Black & White Copy |
$7.25
|
This case describes a Richmond, Virginia, asset-management firm assessing an advertising platform’s content (Scout Guide city guides) to inform its marketing strategy. The firm, Kibet Capital, could manually review each Scout Guide issue to gather demographic statistics—which would take a significant amount of time. Or it could turn to its analytics department and use AI to process the human faces in each city guide with a recognition algorithm and generate a summary statistic for all the guides. While less labor intensive and time consuming than the manual method, one team member questioned whether facial detection and recognition tools would be bias free. Students are encouraged to consider how data science and machine learning can be used to answer quantitative questions as well as address qualitative concerns, such as representation. The material also provides an opportunity to use Python starter code from three Jupyter notebooks to extract images from PDFs, convert images to embeddings, and build deep neural networks for image classification. Students should be very familiar with predictive modeling concepts such as regression and classification techniques and out-of-sample testing. Jupyter notebooks are provided as instructor supplements to this case and will require a Jupyterhub environment to use them.
- To introduce the basics of a neural network - To understand how neural networks modify the representation of features in successive layers - To gain experience with working with unstructured data, particularly images - To understand a typical image processing pipeline for face detection and classification - To practice identifying biases in training data that lead to biased inferences