5 Whys and Other Lies

"Our words matter. Our words have consequences" - Sidney Dekker

The software has bugs. The systems sometimes fail. People make "mistakes". These are fundamental truths of technology. But the 5 Whys and other lies we've been told about analyzing systems failures are detrimental to actually learning from these failures. This talk will expose these convenient lies we've been told, and offer helpful alternatives.

Transcript

1. Hello there! https://www.allthetalks.org/tickets All the Talks has raised almost $70k for WHO to aid COVID-19 victims! Please donate if you’d like to help!
2. If you were to tell a neighbor about the last time a team in your organization really came together to solve a problem, what would you say? What would you say about how the teams worked well together? - SUSTAIN
3. 5 Whys and Other Lies About Complex System Failures
5. How to Do the 5 Whys, according to some random Medium author Ask Why the problem happens and write the answer down below the problem; If the answer you just provided doesn’t identify the root cause of the problem that you wrote down in Step 1, ask Why again and write that answer down; In the beginning, you usually won’t find the root cause of the problem so loop back to Step 3 until the team agrees that the problem’s root cause has been identified. https://medium.com/productmanagement101/learn-about-the-five-whys-technique-78283d75800f
7. Oh look, a template in Excel
11. The traps
12. Trap #1: Attribution to “human error” Photograph by Sheila Sund (https://flic.kr/p/FzZFb1)
15. Trap #2: Counterfactual reasoning Photograph by Einheit 00 (https://flic.kr/p/oQ5X91)
16. “should have,” “could have,” “would have,” “failed to,” “did not.”
17. Trap #3: Normative language Photograph by Nimish Gogri (https://flic.kr/p/8WXy8B)
18. The problem Decisions are judged based on their outcomes. The outcome is the one piece of information we know wasn’t available to the person who made the decision.
20. Trap #4: Mechanistic reasoning
21. Mechanistic reasoning says “our system would have worked fine… if it hadn’t been for those meddling kids”
22. The problem Identifying one “failed” component isn’t the same as understanding the incident.
23. “The real problem with ‘5 whys’ is...that it so grossly oversimplifies the process of problem exploration that it should not be used at all” https://qualitysafety.bmj.com/content/26/8/671
24. “Delusional clarity” bit.ly/hindsightoinsight
25. It doesn’t apply to the systems we work with today
26. @allspaw
29. • 5 whys is perhaps the most simplistic and it leads to the least amount of learning from events. • Results are not reproducible! Different people will come up with different causes for the same problem! • A single cause is assumed along with one linear path to the event
30. The very act of separating important or contributory events from unimportant ones, by a few decision-makers, then, is an act of construction, of the creation of a story, it is not the reconstruction of a story that was already there, ready to be uncovered - Sidney Dekker
31. “Why?” is the wrong question. Asking “how?” gets you to describe (at least some) of the conditions that allowed an event to take place, and provides rich operational data. From @allspaw’s mouth, to your ears: https://oreilly.com/ideas/the-infinite-hows
33. Sidney Dekker, Field Guide to Human Error Illustration by Dr. Richard Cook “inside the tunnel”
34. Replacement behaviors
35. 1. Run a facilitated post-incident review Photograph by Melody Ayres-Griffiths (https://flic.kr/p/bnT21X)
36. 2. Ask better questions Photograph by Barney Moss (https://flic.kr/p/gmMJ4K)
37. 3. Ask how things went right Photograph by Dave Bleasdale (https://flic.kr/p/H9ZgUq)
38. 4. Keep review and planning meetings separate Photograph by Ian D Keating (https://flic.kr/p/2c9scLA)
40. Parting thoughts
44. "Our words matter. Our words have consequences" Sidney Dekker
45. Community of researchers and practitioners @ri_cook @allspaw @bergstrom_johan @sidneydekkercom @LauraMDMaguire @caseyrosenthal @nora_js @jpaulreed http://learningfromincidents.io
46. Thank you!