Back to Homepage
Credit: Arek Socha
Homework 2: Analyzing a Data Science Case Study
Benjamin Xie & Greg Nelson
The objective of this homework is to practice comprehending the data science process and decision contexts.
For this assignment, you will read through a case study and methodology from ProPublica on machine bias in predicting future criminals.
1. Read the Case Study & methodology
- Read the ProPublica case study.
- Read the accompanying article on their methodology.
2. Decompose the case study
Decompose the case study by answering the following questions:
- Goals: What are they trying to use data science to do?
- Prior Knowledge: What did they know or assume beforehand? Why?
- Roles/ responsibilities in organization: What was the organization's expectations of these data scientists? (you may need to look at the ProPublica website or read about the organization to answer this question)
- Social Relationship: How did they interact with other stakeholders when conducting this analysis?
- Models: How did they represent their objects of analysis?
- Choices: What options exist for the decision they are informing?
- Outcomes: Given these choices, what are the outcomes of deciding on a given choice?
- Missing from analysis: What is missing from this analysis?
3. Synthesizing the analysis
Write up your findings in a ~500-2000 word (excluding any additional citations) paper or report.
This paper/report should read coherently (so do not just answer each question above and submit a bullet-pointed list).
You will submit this assignment by putting a Google Doc or PDF of your write-up in your personal INFO 370 folder.
This assignment is out of 10 points, although there is an option for extra credit.
- Up to 1 point for answering each question listed in Section 2 (max of 8pts)
- 1 point for each question answered completely (answer with explanations/evidence to substantiate it).
- 0.5 points for each incomplete answer (answer with incomplete evidence to substantiate it).
- 0 points for each question left unanswered.
- Up to 2 points for comprehensibility.
- 2 points for a paper that is easily understood, reads coherently and has no major spelling or grammer mistakes which detract from our ability to understand it.
- 0.1-1.9 points for a paper that is not fully coherent or understandable or has major spelling or grammer mistakes which detract from our ability to understand it.
- 0 points for totally ignoring these insturctions and submitting a bullet-pointed list of answers to the questions above.
- Up to 1 point of extra credit for incorporating and referencing another research paper or report which further develops your paper.
- This additional reference must be substantial (e.g. a research publication or government report. A blog post or basic news article likely isn't good enough). You must also cite it!