Data Morgana - Risks and Limitations of Data Science - Studium Generale - Tilburg University

3 Aufrufe
Published
Data Morgana is a short lecture series, divided into four parts, about the promises, hype, and misconceptions surrounding data science. Watch this third session, where we look at possible dangers and limitations of data science.

It is not much of a shocking statement to say that we are living in a data-driven society. Data science however, and machine learning in particular, can be overly simplistic and sometimes used without sufficient discrimination. This is natural: throwing more data at a simplistic algorithm is easier to get done than rethinking the algorithm or re-collecting better data, so the status quo is the result of taking the path of the least resistance.

Speaker is Dr. Doina Bucur, Assistant Professor in Computer Science at the University of Twente, and part of the faculty's Ethics Committee. She completed her PhD in Computer Science at the University of Aarhus in 2008, after which she worked as a post-doctoral researcher at Oxford University. She works with data which describes how societies of humans (or animals, or objects) link and interact with each other, either on online social networks, or in real life. This is with the aim to understand, and sometimes improve, these societies.

References mentioned on the slides:

Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification http://gendershades.org/ http://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf
(slide 12)

Are We Automating Racism?
https://www.youtube.com/watch?v=Ok5sKLXqynQ
(slides 13-14)

Coded Bias
https://www.codedbias.com/
(slide 15)

Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans https://www.nature.com/articles/s42256-021-00307-0.pdf
(slide 16)

Dissecting racial bias in an algorithm used to manage the health of populations
https://science.sciencemag.org/content/366/6464/447
(slides 19-21)

Characterizing bias in compressed models
https://arxiv.org/abs/2010.03058
(slides 24-26)

Multimodal Neurons in Artificial Neural Networks https://distill.pub/2021/multimodal-neurons/
(slides 27-28)

Preliminary Report Highway HWY18MH010
https://www.ntsb.gov/investigations/AccidentReports/Reports/HWY18MH010-prelim.pdf
(slide 31)

AlgorithmWatch
https://algorithmwatch.org/en/
(slide 33)


---

Other lectures in the Data Morgana series:
March 9: Data Science Milestones - https://youtu.be/C9ZZ6RRdbFg​
March 23: Milestones of Data Science - https://youtu.be/P4wxiklNtHU
April 20: The Future of Data Science

---

For students of Tilburg University this event may count towards the Studium Generale Certificate. More info: https://www.tilburguniversity.edu/campus/studium-generale/certificate-studium-generale

Recorded on Tuesday April 6, online via Zoom.

For more information and future events, follow Studium Generale:
- https://www.tilburguniversity.edu/campus/studium-generale
- Facebook, Instagram & Twitter: @SGTilburg
Kategorien
Corona Virus aktuelle Videos
Kommentare deaktiviert.