MSc.

Course Description

Data science has experienced rapid growth in recent years, driven largely by the progress in machine learning. This development has opened up new opportunities in a wide range of social, scientific, and technological fields. However, it has become increasingly clear that focusing solely on the statistical and numerical aspects of data science often overlooks social nuances and ethical considerations. The field of Human-Centered Data Science (HCDS) is emerging to fill this gap, combining elements of human-computer interaction, social science, statistics, and computational techniques.

HCDS emphasizes the fundamental principles of data science and its human implications. These include research ethics, privacy, legal frameworks, algorithmic bias, transparency, fairness, accountability, data provenance, reproducibility, user experience design, human computation, and the societal impact of data science.

By the end of this course, students will be expected to

• Apply human-centered design methods to data science practice, taking into account ethical concerns and privacy requirements.
• Construct a reproducible data science workflow.
• Understand and differentiate key terms such as bias, fairness, accountability, transparency, and interpretability.
• Apply measures, techniques, and frameworks to make their data science results interpretable in the context of human-centered explainable AI (HC-XAI).
• Enhance data science workflows with qualitative research approaches.
• Be aware of the existing measures, techniques, and approaches that help to reflect on data science practices.

Students will not only understand the core concepts, theories, and practices of HCDS, but also the multiple perspectives from which data can be collected and processed. In addition, students will gain insight into the potential societal implications of current technological advances. This course aims to equip students with the ability to apply data science techniques in a mindful and conscientious manner, taking into account human and societal contexts, resulting in more ethical, inclusive, and meaningful data-driven solutions.

Here you can find our Code of Conduct.

Projects

This term, students in the “Human-Centered Data Science” course tackled a project aimed at making AI more accessible and useful for decision makers. Working in small groups, they created an interactive explanation interface designed to help people understand and make informed decisions based on explainable AI. Throughout the project, students analyzed a dataset, trained models, designed explanations for them, and built an easy-to-use interface to answer important questions for their target audience. Their goal was to combine data science with human-centered design to uncover meaningful insights that can impact real-world decisions.

Link to the GitHub repository: https://github.com/FUB-HCC/hcds-summer-2024

The NPHA Doctor Visits project aims to provide valuable insights into healthcare utilization among older adults in the United States. Using data from the University of Michigan’s National Poll on Healthy Aging, the project analyzes factors influencing doctor visits, such as physical and mental health, employment status, and demographic variables. The insights generated are intended to assist healthcare policymakers and insurance providers in making informed decisions to improve the well-being of older Americans. Additionally, the project addresses biases in the dataset and evaluates the impact of excluding sensitive features like race and gender on healthcare predictions.

GitHub: https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_1
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_1/Report.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_1/slides.pdf
Screencast: Coming soon

This project involves the development of an interactive application designed to predict heart disease using machine learning models. It allows users to explore data, compare model performances, and understand model decisions through explanations. The app emphasizes transparency and fairness by assessing biases and using techniques like SHAP values. It serves as both a predictive tool and an educational resource for understanding AI in medical research.

GitHub: https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_2
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_2/Project%20Description.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_2/HCDS%20Final%20Presentation.pdf
Screencast: Coming soon

This project focuses on developing an application to help medical staff assess maternal health risks using data-driven insights. The app provides predictions of risk levels and allows users to simulate how changes in health indicators affect outcomes. It features an intuitive interface with traffic light color coding to highlight risk levels and offers detailed explanations of predictions. Developed for educational purposes, it emphasizes ethical data science and human-centered design.

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_3
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_3/Project%20Description.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_3/Slides.pdf
Screencast: Coming soon

This project focuses on creating an explanation interface for a Human Activity Recognition system using data from wearable devices like Apple Watch and Fitbit. It aims to help users understand physical activity patterns by providing clear, accessible insights based on machine learning predictions. The interface is designed for various users, including health professionals and researchers, to easily navigate and analyze data. Emphasizing user needs, the project addresses challenges in integrating technical models with intuitive design.

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_4
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_4/Human%20Activity%20Recognition%20XAI-%20PD.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_4/G4_Human%20Activity%20Recognition%20XAI.pdf
Screencast: Coming soon

This project developed an interactive web application to help doctors predict a patient's risk of experiencing a stroke. By entering patient data, such as age and medical history, the tool provides predictions using a machine learning model. It also offers insights and visual explanations to help physicians understand the factors influencing the risk. Designed for ease of use, the application supports doctors in making informed clinical decisions

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_5
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_5/Project%20Description.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_5/Stroke%20Prediction.pdf
Screencast: Coming soon

This project focused on developing an intuitive interface for understanding vaccination data. Targeting pharmaceutical companies, the tool helps identify populations likely to receive vaccines, using data from a U.S. survey on flu vaccinations. The interface is designed for users with a technical background in computer science but not necessarily data science experts.

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_6
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_6/project_description.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_6/project_presentation.pdf
Screencast: Coming soon

This project developed an Asthma Diagnosis Aid Tool aimed at helping doctors diagnose and treat asthma more effectively. The tool uses a model trained on patient data, including lifestyle, medical history, and symptoms, to predict the likelihood of an asthma diagnosis. Designed with a user-friendly interface, the tool provides detailed explanations of how predictions are made, ensuring that doctors can trust and understand the results even without a data science background. The project also emphasizes fairness and transparency, allowing doctors to assess the reliability of the tool’s predictions across different demographic groups.

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_7
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_7/project_description.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_7/project_presentation.pdf
Screencast: Coming soon

The Obesity Level Prediction Dashboard is an AI-driven tool designed to help users understand and manage their obesity risk. The dashboard uses machine learning to analyze factors like age, diet, and physical activity, providing personalized risk assessments and health recommendations. The user-friendly interface makes complex data insights accessible to both the general public and health professionals. This project highlights the potential of AI in promoting healthier lifestyles and informed decision-making.

GitHub:  https://github.com/FUB-HCC/hcds-summer-2024/tree/main/project_8
Project description: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_8/HCDS_PROJECT_DOCUMENTATION.pdf
Slides: https://github.com/FUB-HCC/hcds-summer-2024/blob/main/project_8/HCDS%20presentation.pdf
Screencast: Coming soon

Literature

Aragon, C., Guha, S., Kogan, M., Muller, M., & Neff, G. (2022). "Human-centered data science: An introduction." MIT Press.

Baumer, Eric PS. “Toward Human-Centered Algorithm Design.” Big Data & Society, 4(2), Dec. 2017. http://doi.org/10.1177/2053951717718854.

Aragon, Cecilia, et al. "Developing a research agenda for human-centered data science." Proceedings of the 19th ACM Conference on Computer Supported Cooperative Work and Social Computing Companion. 2016. http://doi.org/10.1145/2818052.2855518

Kogan, M., Halfaker, A., Guha, S., Aragon, C., Muller, M., & Geiger, S. (2020, January). Mapping out human-centered data science: Methods, approaches, and best practices. In Companion of the 2020 ACM International Conference on Supporting Group Work (pp. 151-156).