full transcript

From the Ted Talk by Margaret Gould Stewart: How giant websites design for you (and a billion others, too)

Unscramble the Blue Letters

Now, the next thing that you need to understand is how to degisn with data. Now, when you're wrnoikg on products like this, you have incredible anmuots of information about how people are using your product that you can then use to influence your design decisions, but it's not just as simple as following the numbers. Let me give you an example so that you can understand what I mean. Facebook has had a tool for a long time that allowed people to report photos that may be in violation of our community sdnaadrts, things like spam and abuse. And there were a ton of photos reported, but as it truns out, only a slmal percentage were actually in vaoiioltn of those cumimotny standards. Most of them were just your typical party photo. Now, to give you a sicpfeic hypothetical example, let's say my friend Laura hypothetically uploads a picture of me from a dnurken night of karaoke. This is purely hypothetical, I can assure you. (Laughter) Now, incidentally, you know how some people are kind of worried that their boss or employee is going to discover embarrassing photos of them on Facebook? Do you know how hard that is to avoid when you actually work at Facebook? So anyway, there are lots of these photos being erroneously rotreepd as spam and abuse, and one of the einnreges on the team had a hunch. He really thhguot there was something else going on and he was right, because when he looked through a bunch of the cases, he found that most of them were from people who were requesting the takedown of a photo of themselves. Now this was a scenario that the team never even took into account before. So they adedd a new feature that allowed peploe to message their friend to ask them to take the potho down. But it didn't work. Only 20 percent of people sent the message to their friend. So the team went back at it. They consulted with experts in conflict resolution. They even studied the universal principles of piotle language, which I didn't even actually know existed until this research happened. And they found something really interesting. They had to go beyond just helping people ask their friend to take the photo down. They had to help people express to their friend how the photo made them feel.

Open Cloze

Now, the next thing that you need to understand is how to ______ with data. Now, when you're _______ on products like this, you have incredible _______ of information about how people are using your product that you can then use to influence your design decisions, but it's not just as simple as following the numbers. Let me give you an example so that you can understand what I mean. Facebook has had a tool for a long time that allowed people to report photos that may be in violation of our community _________, things like spam and abuse. And there were a ton of photos reported, but as it _____ out, only a _____ percentage were actually in _________ of those _________ standards. Most of them were just your typical party photo. Now, to give you a ________ hypothetical example, let's say my friend Laura hypothetically uploads a picture of me from a _______ night of karaoke. This is purely hypothetical, I can assure you. (Laughter) Now, incidentally, you know how some people are kind of worried that their boss or employee is going to discover embarrassing photos of them on Facebook? Do you know how hard that is to avoid when you actually work at Facebook? So anyway, there are lots of these photos being erroneously ________ as spam and abuse, and one of the _________ on the team had a hunch. He really _______ there was something else going on and he was right, because when he looked through a bunch of the cases, he found that most of them were from people who were requesting the takedown of a photo of themselves. Now this was a scenario that the team never even took into account before. So they _____ a new feature that allowed ______ to message their friend to ask them to take the _____ down. But it didn't work. Only 20 percent of people sent the message to their friend. So the team went back at it. They consulted with experts in conflict resolution. They even studied the universal principles of ______ language, which I didn't even actually know existed until this research happened. And they found something really interesting. They had to go beyond just helping people ask their friend to take the photo down. They had to help people express to their friend how the photo made them feel.

Solution

  1. polite
  2. photo
  3. engineers
  4. community
  5. working
  6. amounts
  7. violation
  8. people
  9. design
  10. small
  11. reported
  12. thought
  13. added
  14. standards
  15. drunken
  16. specific
  17. turns

Original Text

Now, the next thing that you need to understand is how to design with data. Now, when you're working on products like this, you have incredible amounts of information about how people are using your product that you can then use to influence your design decisions, but it's not just as simple as following the numbers. Let me give you an example so that you can understand what I mean. Facebook has had a tool for a long time that allowed people to report photos that may be in violation of our community standards, things like spam and abuse. And there were a ton of photos reported, but as it turns out, only a small percentage were actually in violation of those community standards. Most of them were just your typical party photo. Now, to give you a specific hypothetical example, let's say my friend Laura hypothetically uploads a picture of me from a drunken night of karaoke. This is purely hypothetical, I can assure you. (Laughter) Now, incidentally, you know how some people are kind of worried that their boss or employee is going to discover embarrassing photos of them on Facebook? Do you know how hard that is to avoid when you actually work at Facebook? So anyway, there are lots of these photos being erroneously reported as spam and abuse, and one of the engineers on the team had a hunch. He really thought there was something else going on and he was right, because when he looked through a bunch of the cases, he found that most of them were from people who were requesting the takedown of a photo of themselves. Now this was a scenario that the team never even took into account before. So they added a new feature that allowed people to message their friend to ask them to take the photo down. But it didn't work. Only 20 percent of people sent the message to their friend. So the team went back at it. They consulted with experts in conflict resolution. They even studied the universal principles of polite language, which I didn't even actually know existed until this research happened. And they found something really interesting. They had to go beyond just helping people ask their friend to take the photo down. They had to help people express to their friend how the photo made them feel.

Frequently Occurring Word Combinations

ngrams of length 2

collocation frequency
billion people 2
allowed people 2

Important Words

  1. abuse
  2. account
  3. added
  4. allowed
  5. amounts
  6. assure
  7. avoid
  8. boss
  9. bunch
  10. cases
  11. community
  12. conflict
  13. consulted
  14. data
  15. decisions
  16. design
  17. discover
  18. drunken
  19. embarrassing
  20. employee
  21. engineers
  22. erroneously
  23. existed
  24. experts
  25. express
  26. facebook
  27. feature
  28. feel
  29. friend
  30. give
  31. happened
  32. hard
  33. helping
  34. hunch
  35. hypothetical
  36. hypothetically
  37. incidentally
  38. incredible
  39. influence
  40. information
  41. interesting
  42. karaoke
  43. kind
  44. language
  45. laughter
  46. laura
  47. long
  48. looked
  49. lots
  50. message
  51. night
  52. numbers
  53. party
  54. people
  55. percent
  56. percentage
  57. photo
  58. photos
  59. picture
  60. polite
  61. principles
  62. product
  63. products
  64. purely
  65. report
  66. reported
  67. requesting
  68. research
  69. resolution
  70. scenario
  71. simple
  72. small
  73. spam
  74. specific
  75. standards
  76. studied
  77. takedown
  78. team
  79. thought
  80. time
  81. ton
  82. tool
  83. turns
  84. typical
  85. understand
  86. universal
  87. uploads
  88. violation
  89. work
  90. working
  91. worried