TL;DR

• Easy
• 7 hours per week
• Interesting

What is Ethics?

“Ethics is a real discussion of the competing conceptions of the good.
This is just the corporate anti-shoplifting rules.”

Oscar martinez

Okay what really is ethics?
It’s basically how to treat all users and people fairly

Graded course material

• 5 Homework Projects
  • Some programming involved
  • 8% each
• 2 Written Crituques
  • Short 1-2 page essays
  • 5% each
• 6 Case Studies
  • Quick survey-type responses
  • ~1% each
• 4 Exercises
  • Simple 1-paragraph responses
  • ~2% each
• 2 Exams
  • 1 online, 1 take-home
  • 10% each

Homework Projects

Facebook

In this assignment you will pull the advertising data that Facebook (or any other social media platform) has on you personally
You will then categorize the different companies into things like “Clothing and Apparel”, “Automotive” etc.
You then create a data flow graph of the different accuracies of the advertising for you personally

You do not need a Facebook account for this
But it definitely makes things easier (I used my wife’s account LOL)

Stats 101

Given a dataset, you will analyze the statistics of it by separating out the “protected class variables” and make some graphs and charts about them
You will be using:

• python (most likely)
• matploblib
• pandas

The graphs can be tricky to get right on matplotlib but other than that the assignment is pretty straightforward

AI/ML Parts 1 & 2

In this project (2 parts) you will first:

• Download a “toxicity” in comments datasets, actually really interesting
• Try to find trends in the data, like gender, age, race etc.

In part 2 you’ll:

• Use word-embeddings to find bias in words
• Word embeddings are cool! Read more about them here: The Illustrated Word2vec

Imagine you can have words, and add them together, to create new meanings
Such as female + royalty = queen
You can literally do this with word embeddings
Problem is there are biased meanings in words today, so you’ll investigate that

Fairness and Bias

You will find a dataset that has to do with a protected class
You will then split the data into “privileged” vs “unprivileged” groups and analyze using:

• Disparate impact
• Statistical difference parity

Matplotlib is your friend!
It will help you render all the graphs you need
One hack a student found in my semester was to inspect the website for the AI 360 fairness assignment to make your graph

Final Project

Here you will take a dataset of your choice and do a bunch of analysis on it
It does take a while actually, maybe like 15-20 hours
It is similar to the Fairness project, but lengthier

• You could always use the UCI Machine Learning Repository to get an idea for a dataset
• Start early!

Written Critiques

These are very short 1-2 page write-ups, about a certain topic in AI:

• You will use the “Ethical Autonomous Vehicle” tool to experience some self-driving car scenarios
• And also use Google’s “What-If” tool to experiment with some datasets

These are very easy assignments, about 30-45 minutes to complete

Case studies

Case studies are just paragraph responses to prompts, that you comment on for your peers
Super simple and quick

Exercises

Even simpler than case studies, just a paragraph or two about a topic in AI

Exams

• Midterm
  • Canvas
  • Open-note
• Final
  • Take-home
  • 15-20 hours so start early
  • Programming
  • Similar to final project

Grade breakdown

• Projects
  • Facebook – 100%
  • Stats 101 – 18% (didn’t do it)
  • AI/ML – 100%
  • Fairness and Bias – 98%
  • Final project – 100%
• Written critiques
  • Ethical autonomous vehicles – 85%
  • What-if – 100%
• Case studies – 95%
• Exercises – 95%
• Exams
  • Midterm – 82%
  • Final – 100%!

Final score: 89.16% or a B 😋