/
Homework 1 - Team BTFD

Homework 1 - Team BTFD

Group Members: Melanie Grudinschi, Cameron Snyder, Nicholas Stout, Raul Ochoa, Ryan Papetti, Chosen Obih

 

Link to Slides: https://docs.google.com/presentation/d/1eWCkBvxsDYUEiD0PXZ5V1VehOl6GJIX-AAJROROfUeI/edit?usp=sharing

 

Cassie:

In the field of modern data science, there are few voices more well-respected than Cassie Kozyrkov. Beginning as a brilliant statistician for Google, she developed a unique acumen to use statistics to compare and evaluate different methods of decision making. Her work has launched her to the Chief Decision Scientist post at Google, opening a new frontier of data science. What differentiates her work from those of many other prominent data scientists is her emphasis on melding the fields of data science, behavioral psychology, and game theory to better understand the implications of “data-driven” decisions. In one of her Harvard Business Review essays, she points out that many times when we believe our decisions are “data-driven”, they are actually “data-inspired” and, at best, confirming our unconscious biases about the context/data. According to her, the best way to circumvent this fallacious thinking is to set your default decision/hypotheses beforehand, something she says is taught well in the social sciences but not in data science curriculums. For her contributions to the emerging field of Decision Intelligence, Cassie has been honored as a guest speaker at many popular data science conferences, a cover on a special data science issue of Forbes, and as the top voice in Data Science & Analytics on LinkedIn in 2019. Her voice has inspired many data scientists to begin to think beyond the walls of their models and to avoid confirmation bias as best as they can, invaluable concepts for this course.

Referenced Essay (has paywall): The First Thing Great Decision Makers Do

Judea:

Judea Pearl is recognized as one of the people whose work helped to advance artificial intelligence (AI). He is known globally for his pioneering work in probabilistic approach to AI and also for the development of Bayesian networks. Through the development of the Bayesian networks, Judea was able to solve one the difficult tasks in AI research back in the 1980s, which was to train machines to associate potential cause to a set of observable conditions. Judea is a professor of computer science and statistics at UCLA and also the director of the institution's cognitive systems laboratory. Despite being one of the pioneers in AI research, Judea has in recent years become of the strong critics of the filed. In an interview with Kevin Hartnett and Quanta Magazine of The Atlantic, he said that, “All the impressive achievements of deep learning amount to just curve fitting.” He believes that that some of the current research in AI is just a pumped version of what machines could do decades ago. Also, in his latest book ‘The Book of Why: The New Science of Cause and Effect’ with Dana Mackenzie, Judea presents an argument that artificial intelligence has been impeded by an incomplete understanding of what intelligence really is. In addition, he pointed out that a possible solution to improving how intelligent machines would think, would be to replace reasoning by association with causal reasoning. This is something to keep in mind as we proceed in this class.

 

Retrospective Analysis of Group Work

  • The good: We were all able to communicate effectively through email to set up a time to meet and work on homework 1 via zoom. Once we met via zoom we were able to complete the assignment relatively easily. Everyone was responsive in the google chat.

  •  The bad: Most of the members in the group have really busy schedules so find a time that works for each person was kinda tricky. There was not a ton of time to get the team together before the assignment was due.

  •  The ugly: Having a team of 6 group members didn't really allow for students to work on a particular aspect of the homework. Lots of information was coming all at once and slides were done with multiple people working on one slide at a time.

  •  What would be done differently (next time): Try and give group members a certain task to do. Do not have a ton of group members work on one slide at a time. Break up which slide each person works on. Practice giving the oral presentation beforehand.