Feedback for AJ

Feedback for AJ AJ, 

This is really fascinating work. Here are some thoughts: 

First, I think you’d be able to say a lot more if you could associate these findings with outcomes. As I suggested in the meeting, you can get school-level outcomes (https://www.universityofcalifornia.edu/infocenter/admissions-source-school), which you could use to make statistical guesses about individual outcomes, or use the school as your unit of analysis. It would be interesting to create a predictive model that matches admissions decisions, to show that there is a real admission strategy. You could ground truth this by observing the norming process for admission readers. For example, you might be able to use this to show that there really are three good strategies for getting into UC: 

1. signal white, high SES

2. signal model minority

3. signal the right kind of diversity

Using the newer prompts, where there 8 topics to choose from, would only strengthen this claim, if you could show the same kinds of indexical use of language regardless of topic. 

I’ve been developing on a new analytical strategy using word embeddings, which I think could be really interesting for showing how different people use the same words differently. This would be particularly useful for showing that “diversity” means really different things to different people, and that some words provide social capital for some but not others. (Tim Kaine can speak Spanish. Does he want a cookie?) I’m attaching a draft I wrote for Jonathan Rosa (and thereby fulfilling my incomplete from last fall :/ ). If this seems interesting to you, I’d propose we co-write an article using this method of analysis, with you as first author. Regardless of whether that feels like the right way forward, I’d love to talk more about this work. 

Thanks for sharing it!

-Chris