full transcript

From the Ted Talk by Brian Olson: How an algorithm can fight election bias so every vote counts

Unscramble the Blue Letters

If you've ever had the suspicion that your vote doesn't really count and the deck might be stacked against you, you might be right. In many places in this country, we don't have a functioning democracy. People might go to the polls, but they might not have a real choice when they get there. In 2010, the people of fdolria were trying to do something about this. They passed a ballot initiative with almost two-thirds of the vote: a new state constitutional amendment requiring that districts be fair and not biased based on race or party. It didn't work. The state legislature sued to try and get out of these new reeieqnrtums, and in subsequent cuort bteatls, the maps they made were found to be racially and partisan biased. Florida is just one example of our national problem with gerrymandering. Gerrymandering is when you take a few people from one plcae and a few people from another place and draw a line around them on the map to ceatre a drisctit with some specific demographic goal. Here's an example world with 25 people: 60% geern people and 40% purple ppoele. If you split that up into five simple districts of five people each, you can preserve that ratio in the outcome and have three districts won by green people and two districts won by purple people. But if you pack enough green people into just two dtricsits, then you can flip that outcome and wind up with three districts where there's a purple majority. Or you can crack the purple people and split them up just right so that they don't have a mairtjoy anywhere. These strategies of packing and cracking are being used in dozens of districts throughout the ctunory. That bright blue district in northeast Florida was found to be rlialcay beasid because it packs too many black people into one district, diminishing their influence elsewhere. That was Florida in 2012, but gerrymandering has been going on for a long time, since at least 1812, when Massachusetts governor Elbridge Gerry sgnied into law a map that was drawn into a political caotron as a monstrous, dragonesque salamander, and thus was born the grmnedyerar. But it's gotten a lot worse lately. The joke is that instead of vortes picking their politicians, politicians are picking their voters. Why is this a problem? When you have too many seats that are gerrymandered to be safe for one party or another, the political process breaks down in some ways. I have one personal example. In 2006, I was a webmaster for a congressional campaign in California. We were in a district that was gerrymandered, say, for the other praty. And the incumbent in that party didn't feel the need to take part in the campaign and wouldn't agree to show up for any debates. He just felt he didn't have to. And our party wouldn't send any help; they figured we were a lost cause. Come election day, the incumbent got the expected 60/40 rlesut. In other plceas, it's even worse. In 2014, there were 32 congressional districts that went uespnopod - no one else on the ballot. That's over 20 miloiln aercnaims with no efvfcteie chcioe over who their representative in Congress is. In some places, the incumbent faces a more eemxrte clnelegahr from within their own party. And whether you have an extremist upstart or a long incumbent safe seat, that paoilciitn might not feel the need to reach out across the aisle and cosoprmmie on anything because they're safe, and the legislative pesorcs grinds to a halt, and voters get more cynical. What if we could have impartial districts? What if we defined what a good district was mathematically and didn't let anyone else's agenda interfere? Florida's map might look something like this. About 10 yraes ago, computers got powerful enough to solve for this kind of map that follows the legal requirements of having equal population per district, cngutuoios districts that are each all one piece, and in this case, solving for compact districts that try to tightly represent one location or region. But I didn't know it would work when I started. Previous work in this area had been on tiny toy maps like the one I showed you earlier, and they didn't think it would scale up to a full state wtorh of data. But I figured I was a pretty good engineer and I'd give it a shot, and I think it worked out pretty well. So, when the 2010 cesuns data starting coimng out, I set my home computer to work, and over the next six months, it came up with 137 maps for state legislature and crnooniegsasl districts all over the country. And I think the results are pretty good. Let's see another one. First, the old way. North Carolina has also been in almost constant legal battles since their maps came out a little over four years ago. Most recently, they were thrown out for racial bias just as primary seoasn was spinning up. New maps were hastily drawn up, and the primary had to be pushed back from mcrah until June. Voters and candidates were left in disarray. That red district in the northeast reaches into and around three other districts. That pink district in the middle pinches down as narrow as possible while reaching out to grab other areas. This is nuts. These are the visual telltales of districts that have been distorted toward some political end. The opposite of a sprawling, non-local gerrymandered map is a compact map, like this. I hope you can see the difference. You can also measure it. I measure compactness as the aevgrae distance per person to the center of their district. In the old North Carolina map, that dcntaise was 38 miles; in my map, it's 25 miles. You can measure how sprawling and non-local a gerrymandered map is and how compact a compact map is. So, it's technically possible. How's the political situation? You might expect that there would be some rcaisestne to this kind of change, and there is, but there is also some damned for it. The republican governor of Maryland has recently called out for national help in overturning his state's democratic gerrymander. That is one of the more contorted messes of tentacled horrors of districts I have seen in any map. (Laughter) I don't know if this is the best map, but I submit that it is a legally viable map, without some of the obvious runaround and daakbwrcs of the old map. There are a lot of states with dvidied gonevemrnt, with the two parties fighting over redistricting. But this shouldn't be something to fight about. Redistricting should be a bureaucratic, brniog process, where you get in new census data, you turn the crnak, and you get out new maps for the next 10 years. In the last few years, California, Arizona, Ohio, and Florida have passed roferm of one kind or another. That sohws that it's possible. Those reforms might not be perfect, and they might still need some tinkering, but we can do it. This is technically possible. Open-source software, free and vlfiaebire, running on home computers that anyone can use can sovle for these kinds of iaaiptrml maps, and the results are pretty good. This is politically possible. People want reform - even some elected officials want it. And the legal mechanisms are achievable. If we could have a chagne now, we could have a big effect on the ftruue of our pltoaicil process. If reform comes to enough places, enough states, we might even be able to get a national sandratd. And a national standard might let us really hold up our core value of equal protection under the law for all. (Applause)

Open Cloze

If you've ever had the suspicion that your vote doesn't really count and the deck might be stacked against you, you might be right. In many places in this country, we don't have a functioning democracy. People might go to the polls, but they might not have a real choice when they get there. In 2010, the people of _______ were trying to do something about this. They passed a ballot initiative with almost two-thirds of the vote: a new state constitutional amendment requiring that districts be fair and not biased based on race or party. It didn't work. The state legislature sued to try and get out of these new ____________, and in subsequent _____ _______, the maps they made were found to be racially and partisan biased. Florida is just one example of our national problem with gerrymandering. Gerrymandering is when you take a few people from one _____ and a few people from another place and draw a line around them on the map to ______ a ________ with some specific demographic goal. Here's an example world with 25 people: 60% _____ people and 40% purple ______. If you split that up into five simple districts of five people each, you can preserve that ratio in the outcome and have three districts won by green people and two districts won by purple people. But if you pack enough green people into just two _________, then you can flip that outcome and wind up with three districts where there's a purple majority. Or you can crack the purple people and split them up just right so that they don't have a ________ anywhere. These strategies of packing and cracking are being used in dozens of districts throughout the _______. That bright blue district in northeast Florida was found to be ________ ______ because it packs too many black people into one district, diminishing their influence elsewhere. That was Florida in 2012, but gerrymandering has been going on for a long time, since at least 1812, when Massachusetts governor Elbridge Gerry ______ into law a map that was drawn into a political _______ as a monstrous, dragonesque salamander, and thus was born the ___________. But it's gotten a lot worse lately. The joke is that instead of ______ picking their politicians, politicians are picking their voters. Why is this a problem? When you have too many seats that are gerrymandered to be safe for one party or another, the political process breaks down in some ways. I have one personal example. In 2006, I was a webmaster for a congressional campaign in California. We were in a district that was gerrymandered, say, for the other _____. And the incumbent in that party didn't feel the need to take part in the campaign and wouldn't agree to show up for any debates. He just felt he didn't have to. And our party wouldn't send any help; they figured we were a lost cause. Come election day, the incumbent got the expected 60/40 ______. In other ______, it's even worse. In 2014, there were 32 congressional districts that went _________ - no one else on the ballot. That's over 20 _______ _________ with no _________ ______ over who their representative in Congress is. In some places, the incumbent faces a more _______ __________ from within their own party. And whether you have an extremist upstart or a long incumbent safe seat, that __________ might not feel the need to reach out across the aisle and __________ on anything because they're safe, and the legislative _______ grinds to a halt, and voters get more cynical. What if we could have impartial districts? What if we defined what a good district was mathematically and didn't let anyone else's agenda interfere? Florida's map might look something like this. About 10 _____ ago, computers got powerful enough to solve for this kind of map that follows the legal requirements of having equal population per district, __________ districts that are each all one piece, and in this case, solving for compact districts that try to tightly represent one location or region. But I didn't know it would work when I started. Previous work in this area had been on tiny toy maps like the one I showed you earlier, and they didn't think it would scale up to a full state _____ of data. But I figured I was a pretty good engineer and I'd give it a shot, and I think it worked out pretty well. So, when the 2010 ______ data starting ______ out, I set my home computer to work, and over the next six months, it came up with 137 maps for state legislature and _____________ districts all over the country. And I think the results are pretty good. Let's see another one. First, the old way. North Carolina has also been in almost constant legal battles since their maps came out a little over four years ago. Most recently, they were thrown out for racial bias just as primary ______ was spinning up. New maps were hastily drawn up, and the primary had to be pushed back from _____ until June. Voters and candidates were left in disarray. That red district in the northeast reaches into and around three other districts. That pink district in the middle pinches down as narrow as possible while reaching out to grab other areas. This is nuts. These are the visual telltales of districts that have been distorted toward some political end. The opposite of a sprawling, non-local gerrymandered map is a compact map, like this. I hope you can see the difference. You can also measure it. I measure compactness as the _______ distance per person to the center of their district. In the old North Carolina map, that ________ was 38 miles; in my map, it's 25 miles. You can measure how sprawling and non-local a gerrymandered map is and how compact a compact map is. So, it's technically possible. How's the political situation? You might expect that there would be some __________ to this kind of change, and there is, but there is also some ______ for it. The republican governor of Maryland has recently called out for national help in overturning his state's democratic gerrymander. That is one of the more contorted messes of tentacled horrors of districts I have seen in any map. (Laughter) I don't know if this is the best map, but I submit that it is a legally viable map, without some of the obvious runaround and _________ of the old map. There are a lot of states with _______ __________, with the two parties fighting over redistricting. But this shouldn't be something to fight about. Redistricting should be a bureaucratic, ______ process, where you get in new census data, you turn the _____, and you get out new maps for the next 10 years. In the last few years, California, Arizona, Ohio, and Florida have passed ______ of one kind or another. That _____ that it's possible. Those reforms might not be perfect, and they might still need some tinkering, but we can do it. This is technically possible. Open-source software, free and __________, running on home computers that anyone can use can _____ for these kinds of _________ maps, and the results are pretty good. This is politically possible. People want reform - even some elected officials want it. And the legal mechanisms are achievable. If we could have a ______ now, we could have a big effect on the ______ of our _________ process. If reform comes to enough places, enough states, we might even be able to get a national ________. And a national standard might let us really hold up our core value of equal protection under the law for all. (Applause)

Solution

  1. season
  2. resistance
  3. party
  4. people
  5. future
  6. district
  7. challenger
  8. country
  9. americans
  10. effective
  11. census
  12. impartial
  13. average
  14. demand
  15. verifiable
  16. contiguous
  17. government
  18. distance
  19. place
  20. process
  21. requirements
  22. choice
  23. politician
  24. compromise
  25. years
  26. divided
  27. racially
  28. drawbacks
  29. places
  30. biased
  31. standard
  32. crank
  33. solve
  34. create
  35. million
  36. battles
  37. reform
  38. green
  39. court
  40. shows
  41. districts
  42. change
  43. florida
  44. extreme
  45. coming
  46. unopposed
  47. worth
  48. political
  49. gerrymander
  50. cartoon
  51. march
  52. signed
  53. congressional
  54. majority
  55. boring
  56. voters
  57. result

Original Text

If you've ever had the suspicion that your vote doesn't really count and the deck might be stacked against you, you might be right. In many places in this country, we don't have a functioning democracy. People might go to the polls, but they might not have a real choice when they get there. In 2010, the people of Florida were trying to do something about this. They passed a ballot initiative with almost two-thirds of the vote: a new state constitutional amendment requiring that districts be fair and not biased based on race or party. It didn't work. The state legislature sued to try and get out of these new requirements, and in subsequent court battles, the maps they made were found to be racially and partisan biased. Florida is just one example of our national problem with gerrymandering. Gerrymandering is when you take a few people from one place and a few people from another place and draw a line around them on the map to create a district with some specific demographic goal. Here's an example world with 25 people: 60% green people and 40% purple people. If you split that up into five simple districts of five people each, you can preserve that ratio in the outcome and have three districts won by green people and two districts won by purple people. But if you pack enough green people into just two districts, then you can flip that outcome and wind up with three districts where there's a purple majority. Or you can crack the purple people and split them up just right so that they don't have a majority anywhere. These strategies of packing and cracking are being used in dozens of districts throughout the country. That bright blue district in northeast Florida was found to be racially biased because it packs too many black people into one district, diminishing their influence elsewhere. That was Florida in 2012, but gerrymandering has been going on for a long time, since at least 1812, when Massachusetts governor Elbridge Gerry signed into law a map that was drawn into a political cartoon as a monstrous, dragonesque salamander, and thus was born the gerrymander. But it's gotten a lot worse lately. The joke is that instead of voters picking their politicians, politicians are picking their voters. Why is this a problem? When you have too many seats that are gerrymandered to be safe for one party or another, the political process breaks down in some ways. I have one personal example. In 2006, I was a webmaster for a congressional campaign in California. We were in a district that was gerrymandered, say, for the other party. And the incumbent in that party didn't feel the need to take part in the campaign and wouldn't agree to show up for any debates. He just felt he didn't have to. And our party wouldn't send any help; they figured we were a lost cause. Come election day, the incumbent got the expected 60/40 result. In other places, it's even worse. In 2014, there were 32 congressional districts that went unopposed - no one else on the ballot. That's over 20 million Americans with no effective choice over who their representative in Congress is. In some places, the incumbent faces a more extreme challenger from within their own party. And whether you have an extremist upstart or a long incumbent safe seat, that politician might not feel the need to reach out across the aisle and compromise on anything because they're safe, and the legislative process grinds to a halt, and voters get more cynical. What if we could have impartial districts? What if we defined what a good district was mathematically and didn't let anyone else's agenda interfere? Florida's map might look something like this. About 10 years ago, computers got powerful enough to solve for this kind of map that follows the legal requirements of having equal population per district, contiguous districts that are each all one piece, and in this case, solving for compact districts that try to tightly represent one location or region. But I didn't know it would work when I started. Previous work in this area had been on tiny toy maps like the one I showed you earlier, and they didn't think it would scale up to a full state worth of data. But I figured I was a pretty good engineer and I'd give it a shot, and I think it worked out pretty well. So, when the 2010 census data starting coming out, I set my home computer to work, and over the next six months, it came up with 137 maps for state legislature and congressional districts all over the country. And I think the results are pretty good. Let's see another one. First, the old way. North Carolina has also been in almost constant legal battles since their maps came out a little over four years ago. Most recently, they were thrown out for racial bias just as primary season was spinning up. New maps were hastily drawn up, and the primary had to be pushed back from March until June. Voters and candidates were left in disarray. That red district in the northeast reaches into and around three other districts. That pink district in the middle pinches down as narrow as possible while reaching out to grab other areas. This is nuts. These are the visual telltales of districts that have been distorted toward some political end. The opposite of a sprawling, non-local gerrymandered map is a compact map, like this. I hope you can see the difference. You can also measure it. I measure compactness as the average distance per person to the center of their district. In the old North Carolina map, that distance was 38 miles; in my map, it's 25 miles. You can measure how sprawling and non-local a gerrymandered map is and how compact a compact map is. So, it's technically possible. How's the political situation? You might expect that there would be some resistance to this kind of change, and there is, but there is also some demand for it. The republican governor of Maryland has recently called out for national help in overturning his state's democratic gerrymander. That is one of the more contorted messes of tentacled horrors of districts I have seen in any map. (Laughter) I don't know if this is the best map, but I submit that it is a legally viable map, without some of the obvious runaround and drawbacks of the old map. There are a lot of states with divided government, with the two parties fighting over redistricting. But this shouldn't be something to fight about. Redistricting should be a bureaucratic, boring process, where you get in new census data, you turn the crank, and you get out new maps for the next 10 years. In the last few years, California, Arizona, Ohio, and Florida have passed reform of one kind or another. That shows that it's possible. Those reforms might not be perfect, and they might still need some tinkering, but we can do it. This is technically possible. Open-source software, free and verifiable, running on home computers that anyone can use can solve for these kinds of impartial maps, and the results are pretty good. This is politically possible. People want reform - even some elected officials want it. And the legal mechanisms are achievable. If we could have a change now, we could have a big effect on the future of our political process. If reform comes to enough places, enough states, we might even be able to get a national standard. And a national standard might let us really hold up our core value of equal protection under the law for all. (Applause)

Frequently Occurring Word Combinations

ngrams of length 2

collocation frequency
green people 3
purple people 3
pretty good 3
state legislature 2
districts won 2
political process 2
congressional districts 2
north carolina 2
gerrymandered map 2
national standard 2

Important Words

  1. achievable
  2. agenda
  3. agree
  4. aisle
  5. amendment
  6. americans
  7. applause
  8. area
  9. areas
  10. arizona
  11. average
  12. ballot
  13. based
  14. battles
  15. bias
  16. biased
  17. big
  18. black
  19. blue
  20. boring
  21. born
  22. breaks
  23. bright
  24. bureaucratic
  25. california
  26. called
  27. campaign
  28. candidates
  29. carolina
  30. cartoon
  31. case
  32. census
  33. center
  34. challenger
  35. change
  36. choice
  37. coming
  38. compact
  39. compactness
  40. compromise
  41. computer
  42. computers
  43. congress
  44. congressional
  45. constant
  46. constitutional
  47. contiguous
  48. contorted
  49. core
  50. count
  51. country
  52. court
  53. crack
  54. cracking
  55. crank
  56. create
  57. cynical
  58. data
  59. day
  60. debates
  61. deck
  62. defined
  63. demand
  64. democracy
  65. democratic
  66. demographic
  67. difference
  68. diminishing
  69. disarray
  70. distance
  71. distorted
  72. district
  73. districts
  74. divided
  75. dozens
  76. dragonesque
  77. draw
  78. drawbacks
  79. drawn
  80. earlier
  81. effect
  82. effective
  83. elbridge
  84. elected
  85. election
  86. engineer
  87. equal
  88. expect
  89. expected
  90. extreme
  91. extremist
  92. faces
  93. fair
  94. feel
  95. felt
  96. fight
  97. fighting
  98. figured
  99. flip
  100. florida
  101. free
  102. full
  103. functioning
  104. future
  105. gerry
  106. gerrymander
  107. gerrymandered
  108. gerrymandering
  109. give
  110. goal
  111. good
  112. government
  113. governor
  114. grab
  115. green
  116. grinds
  117. halt
  118. hastily
  119. hold
  120. home
  121. hope
  122. horrors
  123. impartial
  124. incumbent
  125. influence
  126. initiative
  127. interfere
  128. joke
  129. june
  130. kind
  131. kinds
  132. laughter
  133. law
  134. left
  135. legal
  136. legally
  137. legislative
  138. legislature
  139. line
  140. location
  141. long
  142. lost
  143. lot
  144. majority
  145. map
  146. maps
  147. march
  148. maryland
  149. massachusetts
  150. mathematically
  151. measure
  152. mechanisms
  153. messes
  154. middle
  155. miles
  156. million
  157. monstrous
  158. months
  159. narrow
  160. national
  161. north
  162. northeast
  163. nuts
  164. obvious
  165. officials
  166. ohio
  167. outcome
  168. overturning
  169. pack
  170. packing
  171. packs
  172. part
  173. parties
  174. partisan
  175. party
  176. passed
  177. people
  178. perfect
  179. person
  180. personal
  181. picking
  182. piece
  183. pinches
  184. pink
  185. place
  186. places
  187. political
  188. politically
  189. politician
  190. politicians
  191. polls
  192. population
  193. powerful
  194. preserve
  195. pretty
  196. previous
  197. primary
  198. problem
  199. process
  200. protection
  201. purple
  202. pushed
  203. race
  204. racial
  205. racially
  206. ratio
  207. reach
  208. reaches
  209. reaching
  210. real
  211. red
  212. redistricting
  213. reform
  214. reforms
  215. region
  216. represent
  217. representative
  218. republican
  219. requirements
  220. requiring
  221. resistance
  222. result
  223. results
  224. runaround
  225. running
  226. safe
  227. salamander
  228. scale
  229. season
  230. seat
  231. seats
  232. send
  233. set
  234. shot
  235. show
  236. showed
  237. shows
  238. signed
  239. simple
  240. situation
  241. software
  242. solve
  243. solving
  244. specific
  245. spinning
  246. split
  247. sprawling
  248. stacked
  249. standard
  250. started
  251. starting
  252. state
  253. states
  254. strategies
  255. submit
  256. subsequent
  257. sued
  258. suspicion
  259. technically
  260. telltales
  261. tentacled
  262. thrown
  263. tightly
  264. time
  265. tinkering
  266. tiny
  267. toy
  268. turn
  269. unopposed
  270. upstart
  271. verifiable
  272. viable
  273. visual
  274. vote
  275. voters
  276. ways
  277. webmaster
  278. wind
  279. won
  280. work
  281. worked
  282. world
  283. worse
  284. worth
  285. years