Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Emily Sappington - Designing for AI

76 views

Published on

Presented at UX Brighton 2019, 1st November 2019

Published in: Design
  • Be the first to comment

Emily Sappington - Designing for AI

  1. 1. Designing for AI Setting User Expectations & Creating Minimum Viable Intelligence Emily Sappington
  2. 2. Designing for AI Setting User Expectations & Creating Minimum Viable Intelligence Emily Sappington
  3. 3. Multi-Modal Voice Assistant Cortana © 2019 Emily SappingtonDesigning for AI
  4. 4. Research aid in the Browser Context Scout © 2019 Emily SappingtonDesigning for AI Multi-Modal Voice Assistant Cortana
  5. 5. Healthcare Assistant Babylon Health © 2019 Emily SappingtonDesigning for AI Research aid in the Browser Context Scout Multi-Modal Voice Assistant Cortana
  6. 6. Healthcare Assistant Babylon Health © 2019 Emily SappingtonDesigning for AI Research aid in the Browser Context Scout Multi-Modal Voice Assistant Cortana 🏢 🌱 🦄
  7. 7. Let’s face it, AI is a buzzword.
  8. 8. “How can we do AI?” –Johnny Founder © 2019 Emily SappingtonDesigning for AI
  9. 9. –Jennifer Founder “Our company needs to start doing AI” © 2019 Emily SappingtonDesigning for AI
  10. 10. They’re thinking of this.
  11. 11. - Bernard Marr “Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way we would consider ‘smart’.” © 2019 Emily SappingtonDesigning for AI
  12. 12. - Bernard Marr “Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way we would consider ‘smart’.” © 2019 Emily SappingtonDesigning for AI
  13. 13. Inspired by human intelligence
  14. 14. What makes someone or something smart?
  15. 15. © 2019 Emily SappingtonDesigning for AI Human Tech
  16. 16. Can users achieve basic tasks with it? Does it wow users? - Appearing human - Saving time - Doing a human task in a fraction of the time Are they alive / awake? Are they competent? Can I trust them? Can they do something better than me? Do they impress me? Does it respond to my input? © 2019 Emily SappingtonDesigning for AI Human Tech
  17. 17. Can users achieve basic tasks with it? Does it wow users? - Appearing human - Saving time - Doing a human task in a fraction of the time Are they alive / awake? Are they competent? Can I trust them? Can they do something better than me? Do they impress me? Does it respond to my input? Human Tech © 2019 Emily SappingtonDesigning for AI
  18. 18. Does it respond to me? First, people need to establish responsiveness as a result of their contact or communication. © 2019 Emily SappingtonDesigning for AI
  19. 19. Can users achieve basic tasks with it? Does it wow users? - Appearing human - Saving time - Doing a human task in a fraction of the time Are they alive / awake? Are they competent? Can I trust them? Can they do something better than me? Do they impress me? Does it respond to my input? © 2019 Emily SappingtonDesigning for AI Human Tech
  20. 20. Evaluating Competence Human Example: Higher GPA Student Higher GPA Student Lower GPA Student © 2019 Emily SappingtonDesigning for AI Students estimate the grades of their peers Those who look more confident, maintaining eye- contact, were assumed to have higher grades
  21. 21. Before achieving intelligence, deliver on the most simple promises of an MVP Discovered a 10-to-win rule for users engaging with Context Scout Tech Example: © 2019 Emily SappingtonDesigning for AI Evaluating Competence Human Example: Students estimate the grades of their peers Those who look more confident, maintaining eye- contact, were assumed to have higher grades
  22. 22. This is when users are testing your product © 2019 Emily SappingtonDesigning for AI Can users achieve basic tasks with it? Does it wow users? - Appearing human - Saving time - Doing a human task in a fraction of the time Are they alive / awake? Are they competent? Can I trust them? Can they do something better than me? Do they impress me? Does it respond to my input? Human Tech
  23. 23. This stage of evaluation has the potential to engage users It can also be when they lose faith in your higher aims Voice Assistant Example: A user who can’t set a reminder with their voice, will not be likely to trust the same voice assistant to take down credit card information and order pizza. Users Testing Competence © 2019 Emily SappingtonDesigning for AI
  24. 24. Setting expectations is key.
  25. 25. Crawl Walk Run © 2019 Emily SappingtonDesigning for AI
  26. 26. Set expectations all, along, the way. I can do a 5k! © 2019 Emily SappingtonDesigning for AI Crawl Walk Run
  27. 27. No you can’t © 2019 Emily SappingtonDesigning for AI Set expectations all, along, the way. Crawl Walk Run
  28. 28. Babylon Health Symptom Checker Set expectations early, and make them clear. © 2019 Emily SappingtonDesigning for AI
  29. 29. Be careful not to over- promise. Users will test your system’s intelligence. Cortana v1 for Windows © 2019 Emily SappingtonDesigning for AI
  30. 30. Expectations Case Study: Context Scout AI for the Browser © 2019 Emily SappingtonDesigning for AI
  31. 31. © 2019 Emily SappingtonDesigning for AI
  32. 32. Context Scout No rigid menu Things the AI can’t find aren’t called out Only show the positive The click-target shows discovered categories as the AI finds them © 2019 Emily SappingtonDesigning for AI
  33. 33. Delightfully intelligent What’s shown on the page is what we have to offer, and not our shortcomings. © 2019 Emily SappingtonDesigning for AI
  34. 34. Are they alive / awake? Does it respond to my input? Can users achieve basic tasks with it? Does it wow users? - Appearing human - Saving time - Doing a human task in a fraction of the time Are they competent? Can I trust them? Can they do something better than me? Do they impress me? Human Tech © 2019 Emily SappingtonDesigning for AI
  35. 35. Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart” © 2019 Emily SappingtonDesigning for AI
  36. 36. © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  37. 37. © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  38. 38. © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  39. 39. Productivity / Personal management © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  40. 40. Productivity / Personal management © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  41. 41. Productivity / Personal management © 2019 Emily SappingtonDesigning for AI Imitating Human Behavior Providing Unique Value Users’ perceptions of tech being “Smart”
  42. 42. Winning moments can be both big and small Healthcare is an uncertain place. When technology can provide clarity and assurance, a product has the potential to be an intelligent comfort to users. © 2019 Emily SappingtonDesigning for AI
  43. 43. Whether a delightfully natural voice experience, or just saving a worker having to skim a page… AI’s magic is in the eyes of the user. Smart moments are in the eye of the beholder © 2019 Emily SappingtonDesigning for AI
  44. 44. So, how do you handle the question: “How can we do AI?”
  45. 45. • Set expectation appropriately • Claiming intelligence, you need to go above & beyond user expectations • Do a few things well Lessons for large & small companies © 2019 Emily SappingtonDesigning for AI
  46. 46. Use Voice only if your scenario, devices & environment all support it Personify only if it's helpful Example At Babylon we’re asking people to open up about their personal health Startups: Get scenario-focused © 2019 Emily SappingtonDesigning for AI
  47. 47. Personality in Writing © 2019 Emily SappingtonDesigning for AI
  48. 48. Sometimes that means saying no to engineers If it’s too early to launch a feature, sometimes design needs to step in. © 2019 Emily SappingtonDesigning for AI
  49. 49. …or building rails into your experience The Babylon Health Symptom Checker doesn’t support full natural language dialog, but uses multiple-choice instead to gather patient symptoms. © 2019 Emily SappingtonDesigning for AI
  50. 50. Don’t be this guy I know more than you.
  51. 51. What kind of “intelligence” are you creating? Good intelligence: • Makes users feel smart & competent • Saves them time & effort • Lets them use their brains for what they do best Bad intelligence: • Makes users feel stupid • Doesn’t save any time © 2019 Emily SappingtonDesigning for AI
  52. 52. © 2019 Emily SappingtonDesigning for AI We need to do work to rebuild this fence with our users What we can actually answer What users think we can answer
  53. 53. Have fun! You know people are going to try to test your software, so have fun with it, and let the AI dazzle them in simple and unexpected ways. © 2019 Emily SappingtonDesigning for AI
  54. 54. Closing: Minimum Intelligence © 2019 Emily SappingtonDesigning for AI
  55. 55. Principles for Minimum Viable Intelligence • Do a few things, very well • Don’t expect to be intelligent before you’re smart • Under promise, over deliver • Surprise users with intelligence that suits your scenario © 2019 Emily SappingtonDesigning for AI
  56. 56. Designing for AI Send your CV! emily.sappington @babylonhealth.com Emily Sappington https://linkedin.com/in/ emilysappington @SappingtonEmily
  57. 57. Designing for AI Emily Sappington Try Babylon for free! Promo code: WorkWithUs Send your CV! emily.sappington @babylonhealth.com https://linkedin.com/in/ emilysappington @SappingtonEmily

×