Indian politics
-
jayakris
- Moderators
- Posts: 38095
- Joined: Mon Dec 16, 2002 7:24 am
- Antispam: No
- Please enter the middle number: 5
- Location: Irvine, CA, USA
- Has thanked: 255 times
- Been thanked: 240 times
- Contact:
Re: Indian politics
Bengal has a new Governor. Mr. Ananda Bose. How perfect a Bengali name! Except, he isn't a Bengali. A full-fledged Malayali from Mannanam, Kottayam, about 25 miles from my home place. Okay, maybe he isn't a Bengali, but being a Malayali, he must at least be a communist... Except, he is no communist either. He is in BJP. How ironic, the whole deal
-
prasen9
- Member
- Posts: 20439
- Joined: Thu Jan 02, 2003 8:49 pm
- Please enter the middle number: 1
- Location: Hamburg, Germany
- Has thanked: 161 times
- Been thanked: 79 times
- Contact:
Re: Indian politics
Dr. Bose seems like a guy who has done something instead of the useless politicos for a change.
-
Kumar
- Authors
- Posts: 7230
- Joined: Sun Jul 27, 2003 12:59 am
- Has thanked: 6 times
- Been thanked: 15 times
Re: Indian politics
Why is there a need to make this numbers available to political parties or politicians? Searchable database to see voting trends by streets
https://www.thehindu.com/data/searchabl ... 0.ece/amp/
https://www.thehindu.com/data/searchabl ... 0.ece/amp/
-
prasen9
- Member
- Posts: 20439
- Joined: Thu Jan 02, 2003 8:49 pm
- Please enter the middle number: 1
- Location: Hamburg, Germany
- Has thanked: 161 times
- Been thanked: 79 times
- Contact:
Re: Indian politics
Because we need politicians to punish streets that did not vote for them by not get it repaired. Others, can also send their pet hooligans and criminals (which may be themselves) to maybe light some houses on fire to celebrate an early Diwali.
But, this raises a question. At what level of granlarity should voting data be released. Obviously not 1. What is the minimum reasonable number? The work on privacy in this area is called k-anonymity. There is seminal work by Latanya Sweeney who found out what cancer Governor Weld of Massachussets by doing some intelligent deduction from anonymized and aggregated health data. Since then there has been work on trying to prevent the release of quasi-identifiers. For example, you release the zipcode as 168xx instead of 16805. Or even, 16xxx until there are at least "k" people in each set/region. This has then led to some quite nice results on differential privacy that the U.S. Census used in the last census.
All that is good. But, here the point may be that politicians are not maybe interested in who voted for you and who did not, but it is pretty good for them if the neighborhood that voted 80% of them or against them is shown up for retaliation against those neighborhoods and streets that did not vote for you.
And, this is the best type of data for gerrymandering algorithms.
Which leads to my question. In political data, the k for k-anonymity needs to be quite large. Has some pol sci researcher/group done research on what is acceptable? I will check but some of you may know the answer or pointers to works already.
But, this raises a question. At what level of granlarity should voting data be released. Obviously not 1. What is the minimum reasonable number? The work on privacy in this area is called k-anonymity. There is seminal work by Latanya Sweeney who found out what cancer Governor Weld of Massachussets by doing some intelligent deduction from anonymized and aggregated health data. Since then there has been work on trying to prevent the release of quasi-identifiers. For example, you release the zipcode as 168xx instead of 16805. Or even, 16xxx until there are at least "k" people in each set/region. This has then led to some quite nice results on differential privacy that the U.S. Census used in the last census.
All that is good. But, here the point may be that politicians are not maybe interested in who voted for you and who did not, but it is pretty good for them if the neighborhood that voted 80% of them or against them is shown up for retaliation against those neighborhoods and streets that did not vote for you.
And, this is the best type of data for gerrymandering algorithms.
Which leads to my question. In political data, the k for k-anonymity needs to be quite large. Has some pol sci researcher/group done research on what is acceptable? I will check but some of you may know the answer or pointers to works already.