Homework 9: Data Perspectives (Algorithmic Bias)
Beyond teaching you some technical skills around computing and data, CS111 also wants to help you think about some of the broader (societal) issues surrounding data and computing. Along the lines of the moneyball assignment from earlier this semester, here is another reading and reflection exercise that ties into what we will cover in the last week of the course: how poor models can lead to programs producing poor outcomes.
We estimate this assignment will take roughly an hour.
Our Topic: Poor Algorithmic Decision Making
Tasks
You’ll put your answers into a Canvas quiz.
-
Racial bias in algorithmic decision making is (deservedly) getting a lot of attention. Read this Propublica article on racial bias in algorithms used for prison sentencing and recidivism.
-
Read this blog post by Vyacheslav Polonski on Mitigating algorithmic bias in predictive justice: 4 design principles for AI fairness
-
With all of the problems these technologies have yielded, what’s the appeal to using them?
-
Pull up the list of principles and takeaways that you wrote up as part of the Moneyball homework (hwk 5), which focused on a positive use of algorithmic decision making. You put your answers in this Canvas quiz, should you need to find them again.
Extend, revise, or refine your original list in light of these new articles, so that you now have a list that makes sense in the context of both sets of readings.
-
Think of a project or potential project in which you might use computation over data to help answer questions. Describe the project in a couple of sentences, then talk concretely about what you might do to gain trust in the decisions rendered by the program. Your entire answer should be roughly a paragraph, at most two, in length.
For this, you can work with an issue you’ve discussed in another course, something you’ve read about in the media, an issue you are engaged with through volunteer efforts, or so on. Our goal is to have you try to contextualize these readings in a problem that is meaningful to you.
Grading
This assignment will be graded lightly. We will check that your responses:
- are related to the readings,
- have some generalizability to them (rather than be specifics from the articles), while
- clearly arising from these articles (rather than being vague statements that could apply to absolutely any use of data in decision making), and
- are well written, rather than being full of spelling and grammar mistakes.
Handin
Enter your observations in the following Canvas quiz.(Ignore the points values marked on the questions – Canvas just needs us to put points there).
Homework 9: Data Perspectives (Algorithmic Bias)
Beyond teaching you some technical skills around computing and data, CS111 also wants to help you think about some of the broader (societal) issues surrounding data and computing. Along the lines of the moneyball assignment from earlier this semester, here is another reading and reflection exercise that ties into what we will cover in the last week of the course: how poor models can lead to programs producing poor outcomes.
We estimate this assignment will take roughly an hour.
Our Topic: Poor Algorithmic Decision Making
Tasks
You’ll put your answers into a Canvas quiz.
Racial bias in algorithmic decision making is (deservedly) getting a lot of attention. Read this Propublica article on racial bias in algorithms used for prison sentencing and recidivism.
Read this blog post by Vyacheslav Polonski on Mitigating algorithmic bias in predictive justice: 4 design principles for AI fairness
With all of the problems these technologies have yielded, what’s the appeal to using them?
Pull up the list of principles and takeaways that you wrote up as part of the Moneyball homework (hwk 5), which focused on a positive use of algorithmic decision making. You put your answers in this Canvas quiz, should you need to find them again.
Extend, revise, or refine your original list in light of these new articles, so that you now have a list that makes sense in the context of both sets of readings.
Think of a project or potential project in which you might use computation over data to help answer questions. Describe the project in a couple of sentences, then talk concretely about what you might do to gain trust in the decisions rendered by the program. Your entire answer should be roughly a paragraph, at most two, in length.
For this, you can work with an issue you’ve discussed in another course, something you’ve read about in the media, an issue you are engaged with through volunteer efforts, or so on. Our goal is to have you try to contextualize these readings in a problem that is meaningful to you.
Grading
This assignment will be graded lightly. We will check that your responses:
Handin
Enter your observations in the following Canvas quiz.(Ignore the points values marked on the questions – Canvas just needs us to put points there).