DS817: Algorithms we live by

In this course we will examine human and artificial intelligence comparatively. We will look at state-of-the-art cognitive models that describe how humans cope with a wide range of demanding tasks. Then we will use that knowledge as a lever to understand machine learning models for dealing with the same problems (or vice versa). The topics discussed will include, among others, categorization and forecasting, recommendation, ranking and search, optimization and planning, creativity, as well as different flavors of multi-armed bandits and reinforcement learning. We will examine scenarios where humans are assisted by artificial intelligence, and problems where humans compete against algorithms. Further, we will discuss whether human cognitive processes can be recovered from the behavioral traces that people leave behind on the Internet; Finally, we will look at problems where human intelligence can provide insights and inspiration for the development of new methods in the quest for better AI agents.

In relation to the competence profile of the degree it is the explicit focus of the course to: 

The students will become familiar with seminal algorithms that are deployed in online interfaces and govern our every-day life and will learn how people adapt their decision-making processes to these algorithms. We will discuss problems commonly encountered by start-up companies (e.g. in the e-commerce or media domains) but also by individual decision makers and organizations more broadly. The acquired competencies will be useful for people planning to work as product managers, data analysts or data scientists in the industry. 

The learning objective of the course is that the student demonstrates the ability to:

By the end of the course students are expected to develop a better understanding machine learning algorithms and human cognition.They should be able to have a good grasp of cognitive models that describe human behavior and machine learning models for solving practical problems encountered by individuals and organizations such as categorization, estimation, ranking, clustering etc. 

The exploration-exploitation trade-off: when should people or algorithms try new things and when should they settle for tested courses of action?

Supervised learning and the bias/variance dilemma: what are the main approaches in estimation and categorization in machine learning? What does it mean for a model to overfit the data? What are good ways to get around this problem? 

Unsupervised learning and natural language processing: Can AI learn without a teacher telling it what is right or wrong? How can this be achieved? 

Humans collaborating and competing with AI: What techniques where used in Deep Blue when it won Kasparov? What techniques did IBM use to win in Jeopardy? Similarly, what type of AI algorithms were used by Deepmind to win human champions in the game of Go? How can humans and machines effectively combine their distinct intelligences?  

Human behavior and big data: What can big data teach us about human behavior? How are our data used by companies to personalize their products to us? 

The teaching method is based on three phase model.

Activities during the study phase: