Ethics for the AI Age

[As presented at Interaction ’17, NYC.]

Over the next two decades, connected products will demand an unprecedented amount of user trust. Technologists and designers will ask the public for yet more of their attention, more of their data, more of their lives. AIs will know users’ deepest secrets. Co-operating devices will automate security and safety. Autonomous vehicles will even make life-or-death decisions for passengers.

But ours is an industry still unwilling to grapple with the ethical, social, and political angles of this future. We mistakenly believe that technology is neutral; that mere objects cannot have moral relevance. And so we make embarrassing blunders – racist chatbots, manipulative research, privacy violations – that undermine trust and harm those we should help.

This is a dangerous trajectory. We urgently need a deeper ethical dialogue about emerging technology, and interaction design’s role within it.

Transcript

1. @CENNYDD ETHICS
 FOR THE
 AI AGE
2. The neutrality fallacy.
3. Technology
 is not neutral.
4. bbc.co.uk · 25 march 2016 theguardian.com · 2 october 2014 azcentral.com · 31 October 2016 cnn.com · 20 april 2016
5. The industry hasn’t earned trust.
6. Design is
 applied ethics.
7. Progress deserves ethical accompaniment.
8. ETHICAL CROSSROADS
9. The promise of automation & AI.
10. AI inherits bias.
12. Agency vs. concealment.
13. “We have to demand friction and truth from the products we use.” —Kelsey Campbell-Dollaghan
14. Physical threat.
16. “The answer is almost always ‘slam on the brakes’.” —Andrew Chatham
17. The trolley problem isn’t.
19. Control,
 handover,
 & the deadly seams.
20. Level 0 Level 1 Level 2 Level 3 Level 4 Level 5 Light automation of specific functions e.g. cruise/lane control Autonomous steering, acceleration, braking. Driver disengaged but ready. Autonomous in limited environments e.g. freeways. Driver can relax. Fully autonomous except e.g. severe weather. Driver attention not required. Full autonomy to any legal location in all conditions. No autonomy, just warnings
21. What limits to automation?
23. Automation, labour, and capital.
24. THE ETHICAL DESIGNER
25. We become more moral by
 working at it.
26. Four ethical tests:
27. What if everyone did what I’m about to do? (DEONTOLOGICAL ETHICS) ETHICAL TEST #1
28. Am I treating people as ends or means? ETHICAL TEST #2 (DEONTOLOGICAL ETHICS)
29. Am I maximising happiness for the greatest number? (UTILITARIANISM / CONSEQUENTIALIST ETHICS) ETHICAL TEST #3
30. Would I be happy for this to be published in tomorrow’s papers? (VIRTUE ETHICS) ETHICAL TEST #4
31. Choosing an ethical lens.
32. Are some verticals unethical?
33. Accessibility is an ethical litmus.
34. You can’t change it from the inside.
35. THE ETHICAL COMPANY
36. Business case: revenue, cost, risk.
37. Build ethical infrastructure.
38. Appoint a designated dissenter.
39. Diversity is an ethical early-warning system.
40. Research has several ethical benefits.
41. Policies and docs sometimes help.
42. Policies and docs sometimes help.
43. But culture dictates.
44. THE ETHICAL COMMUNITY
45. The politics of tech:
46. Technological determinism.
47. Internet exceptionalism.
48. Data as ideology.
49. Public vs.
 private morality.
50. Disregarding the politics of our work is itself a political act.
51. Regulation won’t save us.
52. Nor will
 codes of ethics.
53. It’s down to us.
54. FURTHER READING Moralizing Technology Peter-Paul Verbeek Design for Real Life Eric Meyer & Sara Wachter-Boettcher The Little Book of Design Research Ethics IDEO The Republic Plato Groundwork of the Metaphysic of Morals Immanuel Kant Utilitarianism JS Mill Introducing Ethics Dave Robinson, Chris Garratt Authority and American Usage David Foster Wallace Four Futures: Life After Capitalism Peter Frase The Politics of Bitcoin: Software as Right-Wing Extremism David Golumbia
55. @CENNYDD THANK YOU.