Nadia Piet - Design Thinking for AI
•
4 likes•832 views
Questo talk è un invito a designer e innovators di tutto il mondo a partecipare, sfruttando le opportunità e affrontando le sfide dell’intelligenza artificiale per creare human(ity)-centered applications e significative user experiences. Partiremo con un corso intensivo sull’intelligenza artificiale e il machine learning, poi ci interrogheremo sul ruolo dei designer, esplorando alcuni aspetti critici della progettazione, su come applicare le nostre competenze di designer per avvicinare l’IA a valori sociali, economici e per l’utente. Infine presenteremo una panoramica pratica di come utilizzare il design thinking process che conosciamo e condurlo a quello meno familiare dell’intelligenza artificiale. E allora scopriamo, definiamo e progettiamo futuri desiderabili per l’intelligenza artificiale!
Report
Share
Nadia Piet - Design Thinking for AI
- 2. Exploring the role of design (thinking) in AI/ML for WUD Rome by Nadia Piet @nadiapiet
- 3. Hi! I’m Nadia Piet 2006 2019 Freelance service & strategic designer and researcher with a focus on emerging and humanity-centered tech and futures “We shape our tools and then our tools shape us” — Marshall McLuhan
- 4. Where do (service/UX) design and AI/ML intersect? What’s the role of design(ers) in the AI/ML development process?
- 7. artificial intelligence the practice of making computers do things traditionally thought of as requiring human cognition
- 9. machine learning programming ≄ data rules output output data rules
- 10. A new way of communicating with computers Useful for problems where the output is clear, but rules aren’t Predictions are probabalistic (%)
- 11. Prediction Output Cake Model Chef Training Learning Practice GPU Hardware Utensils Algorithm(s) Instructions Recipe Data Input Ingredients Disclaimer: Please note this is a highly simplified representation of the actual process. + System + +
- 14. AI will not tell us problems worth solving or questions worth asking or inefficiencies worth preserving
- 15. human(ity)- centered design / design (thinking) ?
- 16. Where do (service/UX) design and AI/ML intersect? What’s the role of design(ers) in the AI/ML development process?
- 18. system requirements system limitations ML engineering space
- 19. user needs system requirements user experience system limitations design space ML engineering space
- 20. user needs system requirements user experience system limitations design of design for
- 22. Enabling new types of user experiences
- 23. Turning tech capabilities into user and social value user-centered problem solving data-driven opportunity spotting tech-driven opportunity spotting Build on existing applications Leveraging dataResearch to application How might AI/ML help solve [this] in a unique way? How might the data we have access to create value (for our users)? How might we leverage AI/ML (in processes where good outcomes are clear but rules aren’t)? Developing new models
- 24. User research & domain experts for modelling Output (label prediction) User experience Input (data sets) Features (factors) Objective (question to answer) Business value User input
- 25. Trade-offs in choosing an algorithm + training a model Precision % of predictions that are relevant Recall % of objects that are predicted VS How important is .. Accuracy % of predictions are correct Transparency ability to trace back why/how VS
- 26. Benchmarking + evaluating Plot: Current human benchmark Baseline model Minimum confidence level Minimum benchmark to provide value to user 100% accuracy (?) 0% accuracy
- 27. Cost of errors Confusion matrix Positive Negative Positive :) True positive :( False negative Negative :( False positive :) True negative Machine prediction Userreality
- 28. Navigating design values (per use case) Emotional relationship (‘warm tech’) Instrumentalism (‘cold tech’) Automation bias / reliance Lack of trust / manual Personalization Privacy Pro-active (invasive?) Re-active (dormant?) Human touch Computational efficiency
- 29. Prototyping the experience (not the model)
- 31. Explainability
- 35. User autonomy + data consent by Philip van Allen
- 36. Anticipate + design for (graceful) failure
- 37. Data bias & fairness Figure 2–5: ‘COMPAS Software Results’, Julia Angwin et al. (2016)
- 38. Ethics, data privacy & (un)intended consequences
- 39. Translating subjective human experience into computational parameters
- 40. user needs system requirements user experience / trade-offs system limitations design space engineering space Picking + training a model Evaluating your model Cost of errors Explainability User autonomy User feedback + machine teaching Bias + Fairness Spotting opportunities Expectations + graceful failure
- 41. Bridging AI and design
- 42. “Now is our opportunity to shape that future by putting humanists and social scientists alongside people who are developing artificial intelligence” - Marc Tessier-Lavigne President of Stanford University
- 43. The next design (r)evolution industrial economy product design service economy service design experience economy experience design digital / computational economy algorithm / AI design
- 44. “Human-centered design has expanded from the design of objects to the design of algorithms that determine the behavior of automated or intelligent systems” - Harry West (CEO frog)
- 45. 🙋 Thank you all Grazie mille Dankjewel Questions? Curious? Ideas? Let’s connect @nadiapiet hello@nadiapiet.com