interacting-with-ai-2023---module-2---session-1---handout.pdf
- 1. INTERACTION WITH AI AND AUTONOMOUS SYSTEMS – MODULE 2 Session 1, 2023 Asbjørn Følstad, SINTEF
- 2. 2 Chatbots i offentlig sektor Chatbots and interaction design Chatbots in the public sector Chatbots i offentlig sektor Tillit og menneske-AI samarbeid My background: Reseach projects on human-centred AI
- 3. 3 My background: Reseach projects on human-centred AI SINTEF, HCI group Psychology (!?!)
- 4. 4 Interaction with AI – module 2 Five sessions 2.1: Human-AI interaction and design 2.2: Social relations with AI-based chatbots (Marita Skjuve) 2.3: Chatbot interaction and design 2.4: Interacting with generative AI 2.5: Trustworthy interaction with AI
- 5. 5 Literature Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., ... & Teevan, J. (2019). Guidelines for human-AI interaction. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (paper no. 3). ACM. Bender, E. M., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021). On the dangers of stochastic parrots: Can language models be too big?. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (pp. 610-623). ACM Kocielnik, R., Amershi, S., & Bennett, P. N. (2019). Will you accept an imperfect AI?: Exploring designs for adjusting end-user expectations of AI systems. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (paper no. 411). ACM. Yang, Q., Steinfeld, A., Rosé, C., & Zimmerman, J. (2020). Re-examining whether, why, and how Human-AI Interaction is uniquely difficult to design. In Proceedings of the 2020 CHI conference on human factors in computing systems (Paper no. 164). Følstad, A., & Brandtzæg, P. B. (2017). Chatbots and the new world of HCI. interactions, 24(4), 38-42. Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P. B. (2021). My chatbot companion-a study of human- chatbot relationships. International Journal of Human-Computer Studies, 149, 102601. Hall, E. (2018). Conversational design. A Book Apart Skjuve, M., Brandtzaeg, P. B., Følstad, A. (2023) Why people use ChatGPT. Preprint, SSRN 4376834. Bommasani, R. et al. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258. Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv preprint arXiv:2303.10130. Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Preprint, SSRN 4375283.
- 6. 6 Building on and extending the individual assignment in Module 1. Startup today – finish October 26) Individual assignment Group assignment Building on and extending the group assignment in Module 1. Startup today – finish October 26) + Task on Generative AI for concept development (to be detailed later – was originally planned for today, but not enough time :-) Interim reports (Oct. 26)
- 7. Interim report - individual assignment Three topics: • Characteristics of AI-infused systems. • Human-AI interaction design. • Conversational user interfaces to large language models Language: English or Norwegian. Max. pages: 6 Min. articles referenced 4.
- 8. Interim report – group assignment Content – 5-7 pages • A description of the group, who you are - names. • A description of what area of “interaction with AI” you are interested in working with. • A description of the group, who you are - names. • A description of the topic of “interaction with AI” you are interested in. • A description of the user group and context of use you are interested in. • (new) Background section: Position your work relative to existing knowledge and practice • The 1-3 questions that you want to address (can change and evolve) • (updated) Method section – overall approach, design process (optional, but encouraged), data collection methods • (new) Sketches and/or prototypes (optional, but encouraged) • (new) Findings (progress, initial outcomes) • (updated) Minimum six references to literature. Appendix – approx. 1 page • Appendix 1: Interaction design task – briefly describe the process and outcome. Detail reflections and lessons learnt. Brief status on the group task – each group say a few words
- 9. Group 7 – AI in higher education learning 1. How do students use AI for learning? 2. How does the use of AI affect a student's ability to learn? Group 8 – Make a better learning environment for students by AI 1. Why would this be beneficial for the professor? 2. What is the most proficient way to do this? (benefits vs. negative side effects) Group 3 – Generative AI used among students 1. Hvordan bruker studenter generativ AI i studiesammenheng? 2. Hvilke konsekvenser har det? Group 4 – Research Management Systems and AI; critical thinking 1. How does AI simplify reference organization in research management systems? 2. How can AI adapt to diverse user needs in library systems? 3. What are the privacy implications of AI in library systems? Group 5 – Ownership of Generative AI images 1. Who owns the generative AI results produced by DALL-E and Midjourney, and which factors are affecting the ownership of AI generated images? 2. How can generated AI results affect societal problems within creativity, ethics and privacy? Group 9 – Ethical dilemmas; Writing process -authorship -Intellectual property and creativity 1. How Does AI Assistance Impact the Writing Process and Final Output? 2. Who should be credited as the author of AI-generated content? 3. How can we prevent AI-generated content from being used to spread disinformation or propaganda?
- 10. Group 1 – AI in the workforce 1. How is the introduction of AI into the mainstream workforce impacting IT professionals and to what extent? 2. What are the workers’ expectations towards the introduction of AI in the workforce? Group 2 – Recruiting by AI 1. Hvordan forholder arbeidstakere/arbeidsgivere seg til en rekrutteringsprosess med / uten AI? 2. Har arbeidstakere/arbeidsgivere en relativ tillit til AI’s rekrutterings-evne? 3. Hva kan være konsekvensen av for høy eller for lav tillit til AI i rekrutteringssammenheng? Group 6 – Cheating with AI - gaming 1. How can AI be used to cheat 2. How can AI be used to cheat in video games? 3. How can AI prevent cheating? 4. How does the future of AI-based anti-cheating look like? Group 11 – Geneartive AI - Chatbots in social media 1. How can the use of chatbots be a valuable resource for people with special needs? 2. What problems could arise when exposing children to “commercial” chatbots in social media? 3. How can we facilitate good practice, ethics and privacy for interaction with chatbots in Norway? Group 10 – Autonomous Vehicles – safety and regulatory – inclusion – ethical dilemmas 1. What are the safety considerations and regulatory challenges associated with the integration of autonomous vehicles into the Norwegian transportation system? 2. Do autonomous mobility solutions enhance inclusivity and equitable access? 3. What ethical dilemmas emerge from the implementation of autonomous vehicles, and how can they be ethically navigated to ensure public acceptance?
- 11. 11 Human-AI interaction and design
- 12. 12 What types of systems are we designing when designing for interaction with AI? AI-infused systems
- 13. 13 "Systems that have features harnessing AI capabilities that are directly exposed to the end user." AI-infused systems Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., ... & Teevan, J. (2019). Guidelines for human-AI interaction. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (paper no. 3). ACM.
- 17. Individual assignment – task 1: Characteristics of AI-infused systems • AI-infused systems are ' systems that have features harnessing AI capabilities that are directly exposed to the end user' (Amershi et al., 2019). • Drawing on the lectures of Module 2 and three of the mandatory articles (Amershi et al. (2019), Kocielnik et al. (2019), Yang et al., (2020)). Identify and describe key characteristics of AI- infused systems. • Identify one AI-infused system which you know well, that exemplifies some of the above key characteristics. Discuss the implications of these characteristics for the example system, in particular how users are affected by these characteristics. What are the characteristics of AI-infused systems? Identify one system and use this to exemplify
- 18. VS.
- 19. 19 Kocielnik et al. (2019) Probabilistic – almost always operate at less than perfect accuracy Impacted by user actions – such as user-generated content Transparency issues – how to mitigate? e.g. by showing decision rules
- 20. 20 Amershi et al. (2019) Definition of AI-infused systems: Systems that have features harnessing AI capabilities that are directly exposed to the end user Uncertainty -> errors common, both false positives and negatives Inconsistency -> sensitive to context and small changes in input Behind the scenes personalization (e.g. automated filtering) -> potentially undesired information hiding
- 21. 21 VS. Learning | Improving | Black box | Fuelled by large data sets Dynamic Mistakes inevitable Data gathering through interaction Opaque
- 22. Individual assignment – task 2: Human-AI interaction design • Amershi et al. (2019) and Kocielnik et al. (2019) discuss interaction design for AI- infused systems. Summarize main take- aways from the two papers. • Select two of the design guidelines in Amershi et al. (2019). Discuss how the AI-infused system you used as example in the previous task adheres to, or deviates from these two design guidelines. Briefly discuss whether/how these two design guidelines could inspire improvements in the example system. • Bender et al. (2021) conduct a critical discussion of a specific type of AI- infused systems – those based on large language models. Summarize their argument concerning problematic aspects of textual content and solutions based on large langue models. https://www.microsoft.com/en-us/haxtoolkit/ai-guidelines/ Later lecture
- 24. 24 Learning system - design for change • M1: make clear what the system can do • M2: make clear how well the system can do what it can do • Explain dynamic character (?) Learning | Improving | Black box | Fuelled by large data sets
- 25. 25 Learning system - design for change • M1: make clear what the system can do • M2: make clear how well the system can do what it can do • Explain dynamic character (?) Learning | Improving | Black box | Fuelled by large data sets
- 26. 26 Learning system - design for change • M1: make clear what the system can do • M2: make clear how well the system can do what it can do • Explain dynamic character (?) Learning | Improving | Black box | Fuelled by large data sets
- 27. 27 Learning system - design for change • M1: make clear what the system can do • M2: make clear how well the system can do what it can do • Explain dynamic character (?) Learning | Improving | Black box | Fuelled by large data sets
- 28. 28 Mistakes inevitable - design for uncertainty • M9: Support efficient correction • M10: Scope services when in doubt Learning | Improving | Black box | Fuelled by large data sets
- 29. 29 Mistakes inevitable - design for uncertainty • M9: Support efficient correction • M10: Scope services when in doubt Learning | Improving | Black box | Fuelled by large data sets
- 30. 30 Mistakes inevitable - design for uncertainty • M9: Support efficient correction • M10: Scope services when in doubt Learning | Improving | Black box | Fuelled by large data sets
- 31. 31 Mistakes inevitable - design for uncertainty • M9: Support efficient correction • M10: Scope services when in doubt Learning | Improving | Black box | Fuelled by large data sets
- 32. 32 Mistakes inevitable - design for uncertainty • M9: Support efficient correction • M10: Scope services when in doubt Learning | Improving | Black box | Fuelled by large data sets
- 33. 33 Difficult to understand and validate output – design for explainability • M11: Make clear why the system did what it did Learning | Improving | Black box | Fuelled by large data sets
- 34. 34 Difficult to understand and validate output – design for explainability • M11: Make clear why the system did what it did Learning | Improving | Black box | Fuelled by large data sets
- 35. 35 Difficult to understand and validate output – design for explainability • M11: Make clear why the system did what it did Learning | Improving | Black box | Fuelled by large data sets
- 36. 36 Data wanted – design for data capture • Accommodate gathering of data from users • … but with concern for the risk of being gamed • Make users benefit from data • Design for privacy Learning | Improving | Black box | Fuelled by large data sets
- 37. 37 Data wanted – design for data capture • Accommodate gathering of data from users • … but with concern for the risk of being gamed • Make users benefit from data • Design for privacy Learning | Improving | Black box | Fuelled by large data sets
- 38. 38 Data wanted – design for data capture • Accommodate gathering of data from users • … but with concern for the risk of being gamed • Make users benefit from data • Design for privacy https://www.technologyreview.com/s/610634/microsofts-neo- nazi-sexbot-was-a-great-lesson-for-makers-of-ai-assistants/ Learning | Improving | Black box | Fuelled by large data sets
- 39. 39 Data wanted – design for data capture • Accommodate gathering of data from users • … but with concern for the risk of being gamed • Make users benefit from data • Design for privacy Learning | Improving | Black box | Fuelled by large data sets
- 40. 40 Data wanted – design for data capture • Accommodate gathering of data from users • … but with concern for the risk of being gamed • Make users benefit from data • Design for privacy Learning | Improving | Black box | Fuelled by large data sets
- 41. 41 Human-AI interaction and design – practical application of guidelines
- 42. Getting familiar with the Amershi et al. (2019) design guidelines for Human-AI interaction https://www.microsoft.com/en-us/haxtoolkit/ai-guidelines/ Identify one generative AI service you know and like Discuss service from perspective of 2 or more guidelines Go to design guidelines – online version. Wrap-up and present
- 43. 43 End 1