Paradoxes in Data Science
0 likes256 views
This document discusses several paradoxes that can arise in data science. It begins by discussing modelling and simulations that can be used when data is unavailable. It then outlines Simpson's Paradox, where a trend seen in groups disappears or reverses when the groups are combined. Next, it discusses the accuracy paradox, where a metric stops being useful once it becomes the target. It also discusses the learnability-Godel paradox related to the limitations of mathematics according to Godel's incompleteness theorems. Finally, it discusses the law of unintended consequences as it relates to data science.
Paradoxes in Data Science
- 1. Paradoxes in Data Science Pier Paolo Ippolito
- 2. Hello! ▰ Data Scientist at SAS Institute ▰ Towards Data Science writer ▰ Freelancer ▰ MSc Artificial Intelligence (University of Southampton) 2
- 3. My Work 3
- 4. Paradoxes Paradoxes are a class of phenomena which arise when, although starting from premises known as true, we derive some sort of logically unreasonable result. As Machine Learning models create knowledge from data, this makes them susceptible to possible cognitive paradoxes between training and testing. 4
- 5. Agenda ▰ Data Science without data: Modelling and Simulations ▰ Simpson's Paradox ▰ Accuracy Paradox ▰ Learnability-Godel Paradox ▰ The Law of Unintended Consequences 5
- 6. 1. Modelling and Simulations in Data Science Using Data Science and Machine Learning even when there is no data available.
- 7. “Essentially all models are wrong, but some are useful. 7 George E. P. Box, Statistics for Experimenters, second edition, 2005, page 440
- 9. Modelling Approaches There are two main types of programmable simulation models: ▰ Mathematical Models: make use of mathematical symbols and relationships in order to summarise processes. Compartmental Models in Epidemiology are a typical example of mathematical models (e.g. SIR, SEIR, etc…). ▰ Process Models: are based on a list of steps handcrafted by the designer in order to represent an environment (e.g. Agent-Based Modelling). 9
- 14. Web Application ✔ Streamlit interface ✔ Docker support ✔ Automatic update of COVID-19 stats every 24 hours ✔ International news updated every 2 hours ✔ Docker container hosted on the Azure Container registry, deployed on an Azure Web App 14
- 16. Practical Demonstration ✔ Complete Publication ✔ Extras ✔ Open Source Code ✔ Web Application ✔ Medium Article 16
- 20. Simpson’s Paradox «Simpson's paradox, is a phenomenon in probability and statistics, in which a trend appears in several different groups of data but disappears or reverses when these groups are combined. The paradox can be resolved when causal relations are appropriately addressed in the statistical modelling.» 20
- 25. “When a measure becomes a target, it ceases to be a good measure. 25 Charles Goodhart
- 26. Pareto Principle 26 Image Source
- 28. Learnability-Godel Paradox Kurt Gödel is one of the most famous mathematicians of the last century. Undisputedly, one of it’s most interesting theories are the two Gödel’s Incompleteness Theorems.: ▰ According to these theorems, nowadays Mathematics has some intrinsic limitations which doesn’t allow it to state with certainty if a statement is true or not. The whole field of Data Science is deeply interconnected with mathematical thinking and therefore this leads us to a paradox (Learnability-Godel Paradox). ▰ Depending on if Gödel theory is right or wrong, this would demonstrate to be either possible or not to make extrapolations from a population sample. 28
- 29. 5. The Law of Unintended Consequences
- 30. The Law of Unintended Consequences 30 Image Source
- 31. 5. Conclusion
- 32. Conclusion In this presentation, I introduced some of the main paradoxes related to Data Science. Although, many other common paradoxes could potentially have implications in Data Science and Artificial Intelligence. Some examples are: • Friendship paradox in Network Analysis • Berkson’s Paradox • Braess’s Paradox • Moravec Paradox • Birthday Paradox 32
- 33. Thank you! Questions? Contacts: ▰ LinkedIn ▰ GitHub ▰ Online Portfolio ▰ Towards Data Science 33