Week 7-8: Start of Second Quarter

Welcome back to my blog. I had a wonderful relaxing break and have begun my second term at WPI. Fall also occurred suddenly in Worcester, which some of the trees on campus took quite literally.

Recap of First Quarter

Last term, I met and worked with both the IQP team and the graduate student mentor, Caitlin, several times a week. Our goal was to become well-acquainted with the project and start some initial designs such that we could start the second term at full speed. I am happy to say that we are all quite familiar with the topic now and have some great drafts of the user interface (please refer to the blog of my teammate, Diana, at diana.matters-creu.com).

One of the difficulties I experienced last term was the sensation of being pulled in two directions. On one side, I was working with the IQP team on cracking the LineUp code and designing the user experience. On the other, I was working one-on-one with Caitlin to define research questions upon which we could formulate a paper. Unfortunately, task-management is one of my weaknesses, and I fell into a trap of un-productivity. 

Like the pair of jeans in the Levi Strauss & Co. brand, I felt like I was being pulled in two directions. Image source: Photobucket.com
I discussed this matter with the faculty advisor, Elke, as well as Caitlin and the IQP team and we decided it would be best if I focused my attentions on working with Caitlin for the research objectives.

Objective for Second Quarter

With all that out of the way, I am excited to start this next term. In the past week, I met twice with Caitlin (in addition to the bi-weekly meetings with the IQP team and faculty advisor) to discuss plans for this term. My objective will be to produce a Python script that takes a dataset and partial ranking information from a user and learns a complete ranking (build-a-rank tool). To motivate this tool, Caitlin and I are considering implementing recently reported metrics that control for bias and fairness. These metrics will be available to a user and will safeguard the user from unintentionally biasing their results. In addition to creating the deliverable script for the RANKit application (under development by the IQP team), Caitlin and I hope to turn the results of our research into a paper.

Bias and Fairness

A 2016 paper by ProPublica caught the attention of the research community when it suggested that the COMPAS score, used by judicial courts to assess a defendant's risk of re-offense, was biased against black defendants. The company in charge of COMPAS, Northpointe, issued a report denying the bias, and a discussion ensued across the research community (read more about it here).

The plan is to incorporate the results of that discussion into our ranking toolkit. RANKit is designed to help users interact with their data and interpret the rankings. In that paradigm, it makes sense for our application to allow users to easily apply the various fairness metrics into their ranking and see how it impacts the results.

In the literature, bias and fairness are almost exclusively considered in classification settings (for example, black defendants versus white defendants). Without constraining the objects we rank into two groups, is there another way to apply the idea of fairness to a ranking? This is the question I hope to answer in my blog next week.

Until next week.

Comments

Post a Comment

Popular posts from this blog

Week 31: Datasets for fairness

This week, Diana and I looked for some datasets on which to evaluate our fairness criteria. This is no simple task. For one, the datasets need to be large enough to be split into training, calibration, and testing sets. Second, there needs to be some protected group attribute like sex, race, or age. Third, there needs to be some outcome ranking as well as some truth value. The truth is often the hardest attribute to find. In this post, I introduce some of the datsets we decided to use. I try to identify each of the necessary features that we need to properly test our fairness criteria. COMPAS The COMPAS dataset is an obvious choice because it is well ingrained in the literature about fairness. It has 18,000 observations (n=18000) and 1000 attributes (k=1000). Protected groups can be sex (81/19) or race (51/49). The outcome attribute can be the recidivism score, the score used in court that represents the risk that the accused will commit another crime. The truth attribute is a li...
Read more

Week 32-33: Wrapping up

Hi all. This is my last blog post for this project. The last few weeks, we spent little time working on the project and instead focused on finishing up our classes. To be honest, we lost some motivation after finishing the paper, and we've been reluctant to start up new tasks so close to the end of the year. In this post, I'd like to talk about some of the things I've learned in this project other than what we found in our research. This post is really more a service to myself; to help me reflect on this past year and what I've learned. When I started this project, I didn't understand how research in computing worked. My exposure to research had been more biology/clinical, in which research involved setting a hypothesis about some protein, designing an experiment with positive and negative controls, and then running the experiment repeatedly until confident in the results. In computing research, my experience has been of two varieties: reading papers and proce...
Read more

Week 30: Returning to Fairness

With the LineUp paper submitted, it's now time to return to the fairness problem we were working on before. As a bit of a refresher for you and for me, I'll review what we've done with fairness so far and what remains to be done. Fairness in Ranking Up until now, the literature regarding fairness has focused on fairness in classification. As an illustrative example, consider fairness in hiring practices with repect to applicant sex. A job opening might receive 100 applicants evenly split into 50 males and 50 females. If males and females are equally capable of doing the job, then you might be concerned to see the hiring company mostly interviewing male applicants. To evaluate fairness, you would compare each applicant's capability (capable of performing duties, not capable) to their outcome (interviewed, not interviewed). In a completely fair scenario, the same proportion of capable applicants from each sex group would receive an interview. Small deviances from this...
Read more