-
Notifications
You must be signed in to change notification settings - Fork 7
Home
The Year of Data and Society at the University of Pittsburgh has supported the development of this new engineering technical elective, called “Data for Social Good”, listed under the industrial engineering curriculum as IE-1171. The new curriculum engages students in a critical study of contemporary topics at the intersection of AI, data and society from three perspectives: (i) essence of data, (ii) AI in fabrics of society, and (iii) algorithms in the wild. The course also covers topics in media, economy, law and ethics. Modules on algorithmic bias, data privacy, and data ethics are openly licensed and available on GitHub for adoption and adaptation:
- Rahimian, A., Kamara, N., & Xing, Z. (2022). Data for Social Good: http://d-scholarship.pitt.edu/43094/
- https://github.com/aminrahimian/data-for-good/wiki
In this course, you will learn to apply machine learning (ML) techniques for data-driven decision making in real-world contexts. We will walk you through modules from easy to more advanced ones and think through the ethical and societal consequences of ML-based decision making in each context. If you are not familiar with machine learning, don't panic! We will learn as we go. The tutorial will guide you to master ML in practice, instead of focusing on the mathematical foundations behind it!
Our two main textbooks for the course are “An Introduction to Statistical Learning with Applications in Python” and “The Ethical Algorithm: The Science of Socially Aware Algorithm Design”. The first book provides a perfect introduction to data processing and all sorts of useful ML models with their algorithms and applications. The second one introduces ethical issues that arise in ML applications. Since machine learning tools can be used to leverage both beneficial and harmful societal outcomes, it is essential for engineers to be cognizant of consequences of deploying ML-based systems when dealing with people's lives and livelihoods. We will encounter sensitive topics such as discrimination, gender bias, privacy and equity.
Doing data for social good is all about appreciating and respecting the social contexts of the models that we make and thinking about how they are going to be applied in those contexts. Sometimes you may end up concluding that no form of predictive modeling can do any good or the risk for harm and misuse is greater than the model’s benefits. We have barely tapped the potential of data science, machine learning and AI technologies to transform our economies and industries. It is natural (and rational) to have reservations towards these transformations and a good way to help overcome the trust barriers is by appreciating the social context of their applications. The gravity of the social contexts (e.g., when dealing with homelessness, substance abuse and suicide prevention) just makes the same principles more poignant and prominent. In data for social good we discuss how predictive models can be made privacy-preserving and fair, and discuss design principles such as explainability and interpretability for building such models. These are all ways to mitigate risk and to ensure that the potential of these technologies are put to use for the good of humanity.
It is time to START YOUR JOURNEY HERE! We hope you find the tutorials inspiring and gain confidence in your ability to analyze the ethical and societal dimensions of predictive modeling and data science.