Seminar | Socially (IR)Responsible Algorithms: How the internet can betray our privacy (WS 2020/21)
Contact person: Lina Oechsner, M.Sc.
This course takes place in the first and second term of the winter semester.
Course grade: Written research report (60 %), Presentation (40 %)
Please register at the examination office for the early examination period.
Credit points: 12 ECTS (PO BWL 2010)
During the course, please communicate and stay updated via the course page on Learnweb. Announcements, lecture slides and any additional material will be published there.
The password for all lecture materials will be displayed in our showcase at the MCM (Am Stadtgraben 13-15, 1st floor) four weeks prior to the first lecture and will also be given in the first lecture. Email and telephone enquiries regarding the password will not be answered.
Thanks to the rapid development of internet 2.0 technologies as well as AI-powered devices we enjoy a comfortable digital life that allows us to connect in seconds with friends and peers all over the world. Social media allows us to access and share information from and with friends, family, brands, and other entities.
Even though we may enjoy a comfortable digital lifestyle and have technology easing our life at tremendous pace, we rarely reflect on the other side of the coin. Through interacting with social media platforms we leave substantial digital footprints in a more and more inter-connected world.
While leaving few traces in different online places may generally not be harmful, things may rapidly change once entities become able to combine data to paint a clearer picture of consumers. Kosinski et al. (2013) show that by combining data from various online sources with algorithms and sufficient computational power one may be able to understand personal traits of consumers such as marital status, sexual orientation, drug consumption, health, or political orientation.
Meanwhile, digitalization allows companies to offer more and more customized products on a micro-target level. Companies can decide for each consumer checking into a website how to price goods and services (e.g. flight tickets), what kind of advertisement to show to a consumer, or what kind of service to offer or reject (e.g. insurance companies).
Thus – and without consumers yet fully realizing – the costs of digitalization, social media, and online consumption have substantially increased for consumers.
In addition, algorithms still make many bad predictions due to wrong specifications, bad training data and other biases (O'Neil 2016), again increasing costs for unaware consumers who are finally – without knowing and having a chance to react – paying the extra bill.
In this seminar, we want to explore together how combining online single source information of consumers may enable companies to predict personal consumer traits and co-variates.
We aim at replicating the approach by Kosinski et al. (2013) by running together a large-scale web survey to create a sufficiently large training data set that we can then use to predict personal traits with the help of social media behavior. We will then use this data set to explore in groups which types of machine learning algorithms put consumers at higher risks by being more or less prone to miss-classifications.
Beside the technical component, a strong emphasis will be put on the ethical discussion of algorithmic marketing and digital privacy.
The course will require students to apply code in R. R knowledge is not mandatory or required as we will provide students with sufficient interactive learning material that will allow them to get familiar with R, R-Studio and the relevant packages. In addition, we will provide R-tutorials and a high level of supervision.
- Kosinski, Michal, David Stillwell, and Thore Graepel. (2013). Private traits and attributes are predictable from digital records of human behavior, Proceedings of the National Academy of Sciences 110 (15), 5802-5805.
- O'Neil, Cathy (2016). Bomb parts: What is a model?, in: O’Neil, Cathy: Weapons of Math Destruction, Crown, New York, 15-31.
- O'Neil, Cathy (2016). Civilian casualties: Justice in the age of big data, in: O’Neil, Cathy: Weapons of Math Destruction, Crown, New York, 84-104.
- O'Neil, Cathy (2016). The targeted citizen: Civic life, O’Neil, Cathy: Weapons of Math Destruction, Crown, New York, 179-198.
- Prof. Dr. Raoul Kübler (responsible)
- Lina Marie Oechsner (accompanying)