Background

This project originated out of curiosity as a way to see if my preferences for the classes I've taken in college are self-consistent. I had a general notion of what classes I liked most and what classes I liked least, but the things in the middle felt like a wash, as if there could easily be a contradiction between (relatively) clear pairwise preferences. I implemented it largely using code I had already written for a separate project (cldf-datasets/mayanphylogeny) that used Z3 to solve for a topological order of quality of a list of sources for linguistic data given some partial rankings between subsets of them.

However in a philosophy class we talked about how some value theorists think that preferences can be incomplete (in the sense that given three classes a, b, c, I prefer b to c but have no preference with a to c or a to b). This struck me as odd, and none of the examples in class were immediately convincing to me. So I extended this project to see if my preferences for classes are incomplete (verdict: it seems they are).

I later expanded the project to include my relative expectations of classes, and then a comparison of whether how good I thought classes were lined up with my expectations (verdict: in general yes, with some notable outliers, like Logic I which I enjoyed much more than I thought I would and 24.951 which I enjoyed less than I thought I would). I'm also not sure how consistent the judgements are from semester to semester, so I've left them rather coarse-grained (i.e. if I have no immediate judgement for a pairing I leave it blank).

Post-graduation update: toyed around with this after every semester and generally found that my perception of classes changes over time (maybe should not be that surprising). Still need to update this after graduation maybe to make sure it all works.

How to run

NOTE: This is probably not up to date. Hopefully by poking around the files you can figure out how to run things if you're interested.

First put your classes into the classes list in the make_classes_csv.py file. This will make a csv file with every single pairwise comparison of classes. Even though not all pairwise comparisons are needed to determine a (complete, never mind partial) order, I figured that giving the opportunity to judge every single pairwise comparison would allow for inconsistencies to more easily show up (harder to "keep track" of what one's judged and let oneself be influenced in this way when one only considers two classes at a time). Fill this file out according to the instructions at the bottom of the make_classes_csv.py file.

After doing that, run the functions in the classes_ranker.py file you're interested in. (Make sure the name of the csv file matches file_name in classes_ranker.py!) There are two functions there you can run:

• find_topological_order finds a topological order given preferences. This solves for consistency.
• print_incomplete_preferences will find all incomplete preferences (i.e., it will check for every combination of three classes, whether there is a preference between two of them but no preference between each of those and the third) and print them.