Download to read offline
Uploaded byRobert Pickering
Science and software development
This document discusses science, software development, and the scientific method. It provides examples of cognitive biases like confirmation bias and availability bias that can impact scientific studies. Randomized controlled trials are presented as a way to reduce biases. The importance of practicing evidence-based decision making and maintaining close relationships with academia to stay informed on new research is emphasized. Disclosure of interests is also highlighted to ensure objectivity.
Science and software development
- 1. Science and Software Development
- 2.
- 5. It reveals thatthe length of our heels, or the distance between our ankle joint and the back of the Achilles tendon, seems to have an impact on the development of the size of our calf muscles
- 6. Why Science and SoftwareDevelopment?
- 7. But first … …a little memory test, try to remember the names, without writing them down
- 8. sci·ence (ˈsīəns/) noun. A systematically organized body of knowledge on a particular subject
- 9. en·gi·neer·ing (enjəˈni(ə)riNG) noun. The branch of science and technology concerned with the design, building, and use of engines, machines, and structures
- 10.
- 11. If we’re scientist we’re nolonger craftsmen?
- 12. Hear Surgeons and LaceMakers
- 13. Scientific Method Theory Evidence Experiment
- 14. Science is away of try not to fool yourself ... and you are the easiest person to fool. - Richard Feynman
- 15. A study showson average people who consult psychics more often live longer. Can you explain this result?
- 16. On average: • Womenlive longer that men • Women are more likely to consult a psychic
- 17. Why women live loverthan men
- 19.
- 20. Name some people fromthe earlier list.
- 21. In the listread out earlier, was there more men or women?
- 22. There were more womenthan men
- 23. Availability bias: The overestimationof the importance of anything memorable or vivid
- 24. Example triplet: 8 1012 Guess the hidden rule by testing other triplets.
- 25. Hidden rule: Any threepositive integers
- 26. Confirmation Bias: Tendency ofpeople to favor information that confirms their beliefs
- 27.
- 28. Overconfidence: A survey found95% of British drivers thought they were above average motorist
- 29.
- 30.
- 31.
- 32. In medicine is considerimportant that people teaching and make policy decisions are still practicing regularly
- 33. What can youdo? Use experiments/Be aware of bias: • View testing as an experiment • Gather data and use as decision making aid • Encourage your organisition to use randomized controlled trails
- 34. What can youdo? Encourage closer relations with acadeamia: • Stay in touch with your lecturers • Read journals • Join university mentor schemes • Join professional organizations
- 35. What can youdo? Next time someone tells you something, ask where the evidence is!
- 36.
- 37.
- 38.
Editor's Notes
- #7 Currently little use of experiments and the scientific method in the software industryThe aim of this talk is:Explain what scientific method isShow that it is needed in software developmentExplain how other industries use scientific method
- #11 Science and Engineer are not completely separate entities, they cannot exist without each otherAll scientist need engineering to some extent as it is necessary for good experiment design
- #13 - From Guardian Weekly Science Podcast 06 Feb 2012: http://www.guardian.co.uk/science/audio/2012/feb/06/medical-research-anthropology
- #14 The software industry tends to believe theories can be validated by reasoning aloneHalf of all business fail why should software methodologies do any better?
- #17 - Based on experiment used by Susan Blackmore, see page P138 the Geek Manifesto by Mark Henderson
- #24 Definition was taken from WikipediaUsed extensively by Ad men, not least in pricing things 4.99 € rather than 5.00 €See p11-24 of Irrationality by Stuart Sutherland
- #27 Definition was taken from WikipediaPeople have a tendency to look for data that fits existing theories rather than look for counter examplesSee P98-103, 109 Irrationality by Stuart Sutherland
- #28 Red placebos are more effective than blue onesTwo placebos a day is more effective than onePlacebo saline injection is more effective than placebo pillsPeople change their behavior when they are observedSource: Bad Science by Ben Goldarce
- #29 P175 Irrationality by Stuart Sutherland
- #30 Studies include thing being tested and a control groupGroups allocated randomlyIncludes blinding where possibleSeek to minimize bias though safe guardsP128-132 The Geek Manifesto by Mark Henderson
- #31 - We have come to expect self-interest/bias as the norm!
- #32 Wisdom teeth have a habit of growing at funny angles, “impacted” in the jargonUp to the mid-1990s impacted teeth were extracted routinely, even if there was no pain or other symptomsJohnathan Shepard began a review to see if this policy was correct, he found no evidence that removing teeth where patients were not in pain had any benefitAs a result the policy was changed and the number of wisdom teeth removal were vastly reducedKey to this policy change was the fact that Johnathan Shepard practiced dentistry as well as teaching and researching the subjectSee p135 The Geek Manifesto by Mark Henderson