Sunday, February 17, 2019

Glassdoor, uMap, morality, privacy, Pandas

Glassdoor ratings widely gamed; another instance of Pearson's Law (or Drucker's or Kelvin's or whatever): "what gets measured, gets managed." I know that Glassdoor got kinda useless for me when picking a job. Oh well. Maybe we should call this "Yelp's Law," after the review site that has similarly become useless.

Anyone used uMap? I so much want a very simple "put a few dots and maybe polygons on a map and share it" tool. I've used Google My Maps before, but: A. it's such a UI mess; is it part of google maps or not? how do I just *get to my map that I made*?, B. it's on google, C. it loads slow.

7 parts of morality from a study by Oliver Scott Curry and Et Al, cross culturally: family values, group loyalty, reciprocity, bravery, respect, fairness, property rights. Compare to Haidt's care, fairness, loyalty, authority, sanctity. I guess we have (Haidt on left, Curry on right):
care (harm reduction) = ?
fairness = fairness + reciprocity + property rights
loyalty = group loyalty + respect
authority = family values + respect + property rights
sanctity = ?
? = bravery
Now, they're probably looking at subtly different things. Also, I'm disinclined to take either of them too seriously.

Privacy is a commons. YES. I do feel worried by ideas like Newsom/Steyer's idea that you should get a cut of the proceeds from Facegoogapplezon - feels like, if that comes to pass, Facebook will pay $7/user (so, $7 or 14B; vs Facebook's $460B market cap, it's a chunk but not a company-killing one) and just keep doing the same ol' nonsense. The analogy to votes is spot on. We don't let you sell your vote, because that corrupts democracy for all of us; we shouldn't let you "sell your data", because that corrupts privacy for all of us. (Also, what the hell does "sell your data" even mean!)

Minimally sufficient Pandas. I feel like especially deduping "pivot", "pivot_table", "crosstab", "melt", "stack", "unpivot", "unstack", "reshape" would be helpful to my state of mind and fluency when dealing with all these things. I often think "I need to make my data look like this: ___" and then google around for a while and settle on one of the above verbs.
I was getting closer to this with dplyr in R! That ecosystem seems like it will get you to think in sane ways. But then I mostly quit using R :-/

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.