Home
> Uncategorized > Stuff’s going on! Some of it’s progress!
Stuff’s going on! Some of it’s progress!
August 2, 2016
Stuff’s going on, peoples, and some of it’s actually really great. I am so happy to tell you about it now that I’m back from vacation.
- The Tampon Tax is gone from New York State. This is actually old news but I somehow forgot to blog it. As my friend Josh says, we have to remember to celebrate our victories!!
- Next stop, Menstrual Equality! Jennifer Weiss-Wolf is a force of nature and she won’t stop until everyone has a free tampon in their… near vicinity.
- There’s a new “bail” algorithm in San Francisco, built by the Arnold Foundation. The good news is, they aren’t using educational background and other race and class proxies in the algorithm. The bad news is, they’re marketing it just like all the other problematic WMD algorithms out there. According to Arnold Foundation vice president of criminal justice Matt Alsdorf, “The idea is to provide judges with objective, data-driven, consistent information that can inform the decisions they make.” I believe the consistent part, but I’d like to see some data about the claim of objectivity. At the very least, Arnold Foundation, can you promise a transparent auditing process of your bail algorithms?
- In very very related news, Julia Angwin calls for algorithmic accountability.
- There’s a new method to de-bias sexist word corpora using vector algebra and Mechanical Turks. Cool! I might try to understand the math here and tell you more about it at a later date.
- Speaking of Mechanical Turk, are we paying them enough? The answer is no. Let’s require a reasonable hourly minimum wage for academic work. NSF?
Categories: Uncategorized
mathbabe 
Thanks for this roundup. I’m at ICERM this week, and next week there’s a workshop on “Predictive Policing”: https://icerm.brown.edu/topical_workshops/tw16-7-pp/. Do you know the folks involved? I’m disappointed to see that the workshop abstract does not have even a perfunctory mention of social justice issues or acknowledgment of the widespread issues in predictive models, though I hope both will be discussed during the workshop.
LikeLike
Yeah, sheesh, what’s up with that? A perfect opportunity to bring in obvious questions and discussions.
Maybe we can contact the organizers?
LikeLike
I’ll see if I can talk to Brendan Hassett, and see what the actual plan for the workshop is. They don’t have a schedule up yet.
LikeLike
I assumed that the main goal of that workshop was for Jeffrey Brantingham (UCLA Anthropology) to expand his PredPol business. I am not optimistic that there will be any strong sociological or legal (e.g., 4th Amendment rights) perspectives presented, let alone taken into account by the working groups.
LikeLike
Sigh. I had a brief conversation with Jill Pipher, ICERM founder and ex-director, where she expressed hope that issues of fairness and social justice would be discussed, but looking at PredPol makes me deeply pessimistic. It would be good if as mathematicians we developed a code of ethics that would at least make us think hard about participating in events like this.
LikeLike
Cathy, what did you think about the NYT article? I thought it could’ve been written by you (I even looked for a citation or quote!). My two cents are:
– biased estimators (such as most Bayesian ones) generally outperform unbiased, but we really do have to be careful which factors we choose, for a variety of reasons, including legal and optics. Sometimes we should make do with a proxy, and sometimes by working a bit harder we can find a better explanatory predictive barrier with less noise. (I’m alluding to things like age, race, gender – those are often powerful predictors, but in some situations we shouldn’t use them. For example in setting credit limits or mortgage underwriting. But in other exercises they might be useful and not illegal or wrong (such as choosing which advertisement to run during (a) a Tyler Perry movie or (b) a Katy Perry concert.
– it’s unfortunate that the word bias has a specific meaning in statistics that is different than the colloquial definition, and therefore many people are, ahem, biased against bias (statistics) when they are really against bias (social/political/cultural).
LikeLike
I’m really not sure how to categorize low wages for filling out surveys. I sometimes get telemarketing calls from, say, political pollsters, and while I don’t often participate, I do sometimes participate. Despite the difference in format, these are unquestionably surveys, and I don’t get paid at all for this work.
I also participate in Google Opinion Rewards, which is in principle identical to Mechanical Turk, except that it only has surveys (not other tasks), and the remuneration is not cash, but Play Store credit. In practice, Google’s surveys only take a few seconds to minutes to complete (compared to Turk’s hours-long surveys), but the “pay” is only a few cents of credit at a time. I’m pretty sure the effective pay rate is lower than minimum wage, before even getting to the issue of whether store credit counts as wages.
So where do you draw the line between volunteer survey participation (which has always been a thing) and paid work morally requiring a minimum wage?
LikeLike
After reading your post I did some searching and found that our Australian goods and services tax applies to tampons despite some vigorous attempts over many years to have it removed. I had forgotten about the original mechanical Turk until reading your last two points (I vaguely recall reading Poe’s story) but had not heard of amazons modern version. It is disturbing and shows how inventive corporations are in creating ways to push down wages.
LikeLiked by 2 people
Reblogged this on Matthews' Blog.
LikeLike
If a friend of mine were to question whether the ends justify the means, would they be guilty of class reductionism? I’m asking on behalf of a friend.
LikeLike