Meeting with Dan McFarland

Meeting with Dan McFarland Do PLDA. Do seeded topic models. 

The story could be simply, AI is this big field. 

A paper on ARXIV. All the deep learning stuff is on Arxivv, more representative of peoples’ interests. Journals and confereces are more democratic, not so much mob rule.      —> This fills in part of the story. 

NEXT THINGS TO DO:      Seed the topics Londa cares about on PLDA. 

        -> I’ll probably get a tiny topic, and show that it grows in the last period. 

        -> So what’s coming in is likely diverse individuals. 

    Can model gender as logit with the topics.

    Can model Asian/Other as logit with the topics. 

Logits on words

    We need to recruit James and Londa back in.

    Look at using logit for terms and/or topics. 

    This came in from outside. How would we show that? 

        Where did people get their degrees from? 

    PNAS? 

Angele Cristian communication. 

The fairness topic 

The big result is the Asian one.          Performance of diversity. 

        Is there a shift in topics? Is there a shift in prominence?          What kind of response?  

Do we separate the Asian result from the fairness one? 

Cross-tabulate 

Given this concern with minorities and women gaining a foothold, we’ve missed the bigger story.