The devops field guide to cognitive biases

Fad1e9ed293fc5b3ec7d4abdffeb636f?s=47 Lindsay Holmwood
September 24, 2014

The devops field guide to cognitive biases

As devops practitioners we focus on improving the culture of collaboration so that others play nicely with us & we play nicely with others - but what if the biggest thing holding us back from change is our own brains?

Cognitive biases can deeply affect our behaviours towards others by herding us towards mental shortcuts that are optimised for timeliness over accuracy, at the expense of rationalising irrational behaviour.

You are probably pushing these biases onto other people every day but don't even know it. Does that idea make you feel uncomfortable? You are probably experiencing the Semmelweis reflex kicking your confirmation bias right now.

Knowing is half the battle. This talk delves into some of the well-known and less well-known biases that may be affecting your ability to work with your peers, and your team's ability to work constructively with other teams.

Viewers of the talk will come away with an overview of biases they run into every day, how to hack their brains to use these biases to their advantage, and some tips on how to mitigate the effects of the limitations baked into their wetware.

We have met the enemy and he is us.

--

These are the resources used in the presentation:

http://www.flickr.com/photos/williamhook/4742869256
http://www.flickr.com/photos/10268891@N07/856565056
http://www.flickr.com/photos/brenderous/3671106784
http://www.flickr.com/photos/frostnova/2268471558
http://www.flickr.com/photos/buttim/1297081125
http://www.flickr.com/photos/dhdesign/1096464615
http://www.flickr.com/photos/princess_l_88/4948893120
http://www.flickr.com/photos/cphotos/4780625450
http://www.flickr.com/photos/amesis/3777130839
http://www.flickr.com/photos/artotemsco/5465529987
http://www.flickr.com/photos/fish2000/4648423198
http://www.comedycentral.com/press/images/colbertreport/stephencolbert2.jpg
http://althouse.blogspot.com.au/2006/05/who-is-clueless-enough-to-think.html
http://www.alternet.org/story/137918/why_do_conservatives_like_stephen_colbert
http://crooksandliars.com/john-amato/ohio-study-finds-conservatives-believe
http://www.youtube.com/watch?v=Ahg6qcgoay4
http://news.ninemsn.com.au/world/2013/07/10/12/43/asiana-pilot-couldn-t-see-runway-before-crash
http://au.news.yahoo.com/thewest/a/-/world/17940112/asiana-flight-crew-saw-trouble-at-500-feet-u-s-safety-board-says/
http://www.theaustralian.com.au/news/world/crash-pilots-wanted-to-abort-landing-as-boeing-777-approached-airport-too-low-investigators/story-e6frg6so-1226675762698
http://www.theguardian.com.au/story/1628513/inexperienced-pilots-may-be-cause-of-asiana-jet-crash/?cs=5
http://www.news.com.au/world-news/fears-pilots-manual-flying-skills-eroding-after-asiana-airlines-crash-in-san-francisco/story-fndir2ev-1226676769126
http://en.wikipedia.org/wiki/Scientific_method
http://en.wikipedia.org/wiki/Publication_bias
http://en.wikipedia.org/wiki/Hindsight_bias#Judicial_system
http://www.flickr.com/photos/studiogabe/5535916719
http://www.flickr.com/photos/sjliew/1311490235
http://www.flickr.com/photos/pcb85/4258475681
http://www.flickr.com/photos/jjackowski/3812946635
http://library.mpib-berlin.mpg.de/ft/rh/RH_Fluency_2008.pdf
http://www.wrayherbert.com/blog/memory-learning-and-the-fluency-heuristic
http://www.youtube.com/watch?v=IGQmdoK_ZfY
http://www.amazon.com/You-Are-Not-So-Smart/dp/1592407366/
http://www.amazon.com/Sway-Irresistible-Pull-Irrational-Behavior/dp/0385530609/
http://www.amazon.com/Thinking-Fast-Slow-Daniel-Kahneman/dp/0374533555/
http://www.amazon.com/The-Checklist-Manifesto-Things-Right/dp/0312430000/
http://www.amazon.com/Field-Guide-Understanding-Human-Error/dp/0754648265/

Fad1e9ed293fc5b3ec7d4abdffeb636f?s=128

Lindsay Holmwood

September 24, 2014
Tweet

Transcript

  1. devops field guide to understanding cognitive biases (2nd edition)

  2. None
  3. Your current iPhone crashes

  4. Your current iPhone is slower

  5. Time to upgrade!

  6. Why?

  7. Apple has made my device slower!

  8. None
  9. Clustering illusion

  10. See patterns in randomness

  11. See correlations where there are none

  12. Your device has always been that slow

  13. Clustering illusion is a Cognitive Bias

  14. What is a cognitive bias?

  15. “Mental shortcut”

  16. None
  17. Timeliness Accuracy

  18. None
  19. Problem solving

  20. Problem solving Heuristic

  21. Problem solving Heuristic Correct result

  22. Problem solving Heuristic Correct result Rational choice

  23. None
  24. Problem solving

  25. Problem solving Heuristic

  26. Problem solving Heuristic Incorrect result

  27. Problem solving Heuristic Incorrect result Cognitive bias!

  28. Heuristic?

  29. Pattern matching

  30. What helped your ancestors survive!

  31. None
  32. March 27, 1977

  33. None
  34. Control tower

  35. Control tower ✈

  36. Control tower ✈ ✈

  37. Control tower ✈ ✈

  38. Control tower ✈ ✈

  39. Control tower ✈ ✈

  40. Control tower ✈ ✈

  41. None
  42. None
  43. KLM:  234 passengers 16 crew

  44. Pan Am:  326 passengers 9 crew

  45. Frozen in place

  46. Control tower ✈ ✈

  47. Control tower ✈ ✈

  48. Control tower ✈ ✈

  49. Normalcy bias

  50. Before a disaster:

  51. None
  52. • Underestimate:

  53. • Underestimate: • risk

  54. • Underestimate: • risk • effects

  55. • Underestimate: • risk • effects • preparation

  56. “Because something bad has never happened, it never will happen”

  57. During a disaster:

  58. None
  59. • Response:

  60. • Response: • slow reaction

  61. • Response: • slow reaction • seek validation

  62. • Response: • slow reaction • seek validation • optimistic

    interpretation
  63. Reaction steps

  64. None
  65. • Cognition

  66. • Cognition • Perception

  67. • Cognition • Perception • Comprehension

  68. • Cognition • Perception • Comprehension • Decision

  69. • Cognition • Perception • Comprehension • Decision • Implementation

  70. • Cognition • Perception • Comprehension • Decision • Implementation

    • Movement
  71. These are  complex tasks

  72. You cannot skip these tasks

  73. You can practice to make them more automatic

  74. People who don't practice deliberate during the disaster

  75. None
  76. None
  77. • 70% freeze

  78. • 70% freeze • 15% freak out

  79. • 70% freeze • 15% freak out • 15% react

    to situation
  80. No practice == higher MTTR

  81. Counter this behaviour

  82. drill drill drill drill drill drill drill drill drill drill

    drill drill drill drill drill drill drill drill drill drill
  83. Normalcy bias has your head in the sand. Drill to

    react when disaster is at hand.
  84. None
  85. None
  86. None
  87. None
  88. None
  89. None
  90. None
  91. None
  92. see it go over here

  93. None
  94. None
  95. was actually here

  96. Confirmation bias

  97. Seek information that reinforces position

  98. Ignore alternative explanations

  99. Interpret ambiguous information in favour of existing positions!

  100. None
  101. None
  102. • Who is clueless enough to think Colbert is a

    conservative? • Why do conservatives like Stephen Colbert? • Ohio study finds conservatives believe Colbert dislikes liberalism
  103. http://www.youtube.com/watch?v=Ahg6qcgoay4

  104. Debugging code

  105. My code is perfect!

  106. I know the problem is in this other library

  107. I just can't find it!

  108. *30 minutes pass*

  109. The bug was in my code

  110. None
  111. Counter this behaviour

  112. Take opposing viewpoints

  113. Is your conclusion negative? It could be biased.

  114. Confirmation reinforces, You ignore other sources

  115. Provisioning automation @

  116. None
  117. 18 months

  118. None
  119. • Provisioning: • We're being hurt badly by VMware provs!

  120. • Provisioning: • We're being hurt badly by VMware provs!

    • R&D: • We'll automate that now!
  121. 2 weeks later

  122. None
  123. • Provisioning: • We're being hurt badly by load balancer

    provs!
  124. • Provisioning: • We're being hurt badly by load balancer

    provs! • R&D: • We'll automate that now!
  125. 1 week later

  126. None
  127. • Provisioning: • We're being hurt badly by VMware provs!

  128. • Provisioning: • We're being hurt badly by VMware provs!

    • R&D: • We'll automate that now!
  129. 4 weeks later

  130. None
  131. • Provisioning: • We're being hurt badly by EC2 provs!

  132. • Provisioning: • We're being hurt badly by EC2 provs!

    • R&D: • We'll automate that now!
  133. None
  134. 18 months

  135. None
  136. What is  happening?

  137. Working on the thing closest to us that hurts us

    most.
  138. But it's not the biggest problem overall!

  139. Would you work like this?

  140. 18 months

  141. Outcome bias

  142. Hindsight bias

  143. "knew it all along" effect

  144. memory distortion distortion

  145. Judgement

  146. “You were capable  of preventing  a bad outcome”

  147. Failure!

  148. Talk about events  in the context  of the

    outcome Failure!
  149. Bad outcomes come from  bad decisions

  150. None
  151. • MH17 tragedy reveals need for flight path overhaul •

    Flight MH17 took abandoned flight path • MH17 flightpath questioned as Malaysia mourns second airline tragedy • Flight MH17 ‘could have avoided Ukrainian airspace for $66 per passenger’ • Why was MH17’s flight path over a conflict zone?
  152. • news.com.au • Sky News • ABC • Daily Mail

    • PBS
  153. Taints all interactions

  154. Hindsight bias  is your culture killer

  155. None
  156. Monitoring thresholds

  157. Backend HTTP Port 81

  158. “I don’t want that thing  that woke me last

    night  to wake me again”
  159. “I’m going to make that problem go away”

  160. But...

  161. How often does  that alert fire?

  162. What percentage  of all alerts is that?

  163. Are there other alerts that are more frequent?

  164. What makes this alert the most important to fix?

  165. +----------------------------+---------------+-----------------+---------------------+ | 2013-07-01T00:00:00+09:30 - 2013-07-31T23:59:59+09:30 | +----------------------------+---------------+-----------------+---------------------+ | Check |

    # of entities | Critical | Total | +----------------------------+---------------+-----------------+---------------------+ | Memory Utilisation | 85 | 9h 44m 31s | 2mo 5d 8h 2m 5s | | Load | 44 | 22d 2h 38m 50s | 1mo 22d 15h 51m 42s | | ** DISK Utilisation (Win) | 41 | 5m 0s | 1mo 19d 16h 2m 7s | | ** Disk Utilisation (Unix) | 43 | 3h 34m 2s | 1mo 1d 18h 24m 28s | | PING | 35 | 1mo 16h 16m 43s | 1mo 16h 16m 43s | | HOST | 37 | 1mo 2h 25m 52s | 1mo 2h 25m 52s | | SSH | 14 | 26d 21h 8m 33s | 26d 21h 8m 33s | | Flapper | 1 | 15d 11h 42m 16s | 15d 11h 42m 16s | | CPU Load | 5 | 8d 12h 21m 31s | 8d 13h 36m 22s | | CPU Utilisation | 30 | 1h 35m 18s | 4d 3h 17m 37s | | ** check_tcp Port | 56 | 2d 19h 25m 47s | 2d 19h 25m 47s | | RDP | 6 | 6h 34m 39s | 6h 34m 39s | | Swap Utilisation | 18 | 5h 43m 44s | 5h 43m 44s | | Frontend HTTP Port 80 | 4 | 4h 22m 14s | 4h 22m 14s | | Backend HTTPS Port 444 | 4 | 4h 20m 19s | 4h 20m 19s | | Frontend HTTPS Port 443 | 4 | 4h 20m 9s | 4h 20m 9s | | Backend HTTP Port 81 | 4 | 4h 17m 41s | 4h 17m 41s | | Disk Utilisation | 2 | 1h 45m 0s | 3h 9m 0s | | ** HTTP Regex | 5 | 11m 53s | 11m 53s | +----------------------------+---------------+-----------------+---------------------+
  166. +----------------------------+---------------+-----------------+---------------------+ | 2013-07-01T00:00:00+09:30 - 2013-07-31T23:59:59+09:30 | +----------------------------+---------------+-----------------+---------------------+ | Check |

    # of entities | Critical | Total | +----------------------------+---------------+-----------------+---------------------+ | Memory Utilisation | 85 | 9h 44m 31s | 2mo 5d 8h 2m 5s | | Load | 44 | 22d 2h 38m 50s | 1mo 22d 15h 51m 42s | | ** DISK Utilisation (Win) | 41 | 5m 0s | 1mo 19d 16h 2m 7s | | ** Disk Utilisation (Unix) | 43 | 3h 34m 2s | 1mo 1d 18h 24m 28s | | PING | 35 | 1mo 16h 16m 43s | 1mo 16h 16m 43s | | HOST | 37 | 1mo 2h 25m 52s | 1mo 2h 25m 52s | | SSH | 14 | 26d 21h 8m 33s | 26d 21h 8m 33s | | Flapper | 1 | 15d 11h 42m 16s | 15d 11h 42m 16s | | CPU Load | 5 | 8d 12h 21m 31s | 8d 13h 36m 22s | | CPU Utilisation | 30 | 1h 35m 18s | 4d 3h 17m 37s | | ** check_tcp Port | 56 | 2d 19h 25m 47s | 2d 19h 25m 47s | | RDP | 6 | 6h 34m 39s | 6h 34m 39s | | Swap Utilisation | 18 | 5h 43m 44s | 5h 43m 44s | | Frontend HTTP Port 80 | 4 | 4h 22m 14s | 4h 22m 14s | | Backend HTTPS Port 444 | 4 | 4h 20m 19s | 4h 20m 19s | | Frontend HTTPS Port 443 | 4 | 4h 20m 9s | 4h 20m 9s | | Backend HTTP Port 81 | 4 | 4h 17m 41s | 4h 17m 41s | | Disk Utilisation | 2 | 1h 45m 0s | 3h 9m 0s | | ** HTTP Regex | 5 | 11m 53s | 11m 53s | +----------------------------+---------------+-----------------+---------------------+ | Backend HTTP Port 81 | 4 | 4h 17m 41s | 4h 17m 41s |
  167. Harness this behaviour

  168. Counterfactuals

  169. *we’re about to launch* “What do you think will break?”

  170. versus

  171. *imagine 5 minutes post-launch* “What just broke?”

  172. Counter this behaviour

  173. During

  174. None
  175. • 1. collect data

  176. • 1. collect data • 2. analyse that data

  177. • 1. collect data • 2. analyse that data •

    3. make decisions based on that data
  178. • 1. collect data • 2. analyse that data •

    3. make decisions based on that data • 4. act on data, not feelings
  179. • 1. collect data • 2. analyse that data •

    3. make decisions based on that data • 4. act on data, and feelings
  180. After:

  181. Explain in terms of foresight

  182. Local rationality

  183. “People make what they think are best decisions based on

    the information at hand”
  184. None
  185. Hindsight affords us global rationality

  186. None
  187. Hindsight blinds, Foresight finds

  188. None
  189. Normalcy bias

  190. Confirmation bias

  191. Hindsight bias

  192. None
  193. How vulnerable to these biases are you?

  194. – you, right now “I’m aware of my biases.”

  195. – you, right now “These don’t affect me.”

  196. – you, right now “I am better than average.”

  197. Better-than-average effect & Illusory superiority

  198. Bias blind spot

  199. None
  200. • Positive attributes

  201. • Positive attributes • Above average

  202. None
  203. • Negative attributes

  204. • Negative attributes • Below average

  205. Why?

  206. Self-esteem

  207. Control

  208. None
  209. Normalcy bias has your head in the sand. Drill to

    react when disaster is at hand.
  210. Confirmation reinforces, You ignore other sources

  211. Hindsight blinds, Foresight finds

  212. Rhyme-as-reason effect

  213. Aesthetics affect truth

  214. None
  215. • Fluent, Fast, Smooth

  216. • Fluent, Fast, Smooth • Higher value

  217. This is a test

  218. None
  219. 1. Read the statement once

  220. 1. Read the statement once 2. Count the letter F

  221. None
  222. FINAL FOLIOS SEEM TO RESULT FROM YEARS OF DUTIFUL STUDY

    OF TEXTS ALONG WITH YEARS OF SCIENTIFIC EXPERIENCE.
  223. None
  224. How many  did you see?

  225. How many  did you see? The answer is 8

  226. Fluency heuristic

  227. FINAL FOLIOS SEEM TO RESULT FROM YEARS OF DUTIFUL STUDY

    OF TEXTS ALONG WITH YEARS OF SCIENTIFIC EXPERIENCE.
  228. Brain expects pattern to continue

  229. Brain skips  other information

  230. Harness this behaviour

  231. Rhyme your mantras

  232. None
  233. Have you learnt anything?

  234. https://www.youtube.com/watch?v=IGQmdoK_ZfY

  235. How many passes?

  236. What did the  gorilla do?

  237. Did you see the curtains change colour?

  238. Or one of the  players leave?

  239. https://www.youtube.com/watch?v=IGQmdoK_ZfY

  240. – Mark Twain “It ain’t what you don’t know that

    gets you into trouble. It’s what you know for sure that just ain’t so.”
  241. Culture is made from people.

  242. Everyone has biases.

  243. Understand how  you think.

  244. Become more rational.

  245. Become less rationalising.

  246. Be aware of bias.

  247. Improve ourselves.

  248. Culture will flow.

  249. I’m Lindsay @auxesis

  250. Thank you!

  251. Thank you! Liked the talk? Let @auxesis know.

  252. None
  253. • Further reading • You Are Not So Smart –

    David McRaney • The field guide to understanding 
 human error – Sidney Dekker • Sway: The irresistible pull of irrational behavior – Ori Brafman • Thinking, Fast and Slow – Daniel Kahneman