Bridging AI and HCI: Incorporating Human Values into the Development of AI Technologies

on YouTube ↗48m

" summary

A Wikipedia recruitment system is built and field-deployed to match newcomers to wiki projects using four models (two interest-based — rule and category similarity — and two relationship-based — message/Bump ties and co-edit similarity). Design follows a process-oriented design that elicits stakeholder values, separates newbies vs experienced candidates, and surfaces recommendations to project organizers. Over six months the system delivers four batches of 385 recommendations to 18 projects, receives ~16,000 responses, shows the rule-based algorithm wins on ratings and invites, and does not disadvantage newcomers.

AI-written summary · about

" chapters (23)

1. 0:00Intro
2. 0:41Today's Theme
3. 4:51Wikipedia: The Largest Collaborative Project in Human History
4. 8:01Recruitment Algorithm
5. 8:34Challenges: Community Acceptance to Al technologies
6. 9:06Challenges: Tension between Wikipedia Community and Research Community
7. 13:51Challenges: Multiple Stakeholders
8. 14:18Challenges: Different Needs and Values
9. 16:29Traditional Algorithm Design
10. 17:36Our Approach: Value Sensitive Algorithm Design (Zhu, Yu, Halfaker & Terveen, CSCW 2018)
11. 18:47Step 1. Understand Stakeholder Values
12. 20:12Stakeholder Values and Trade-offs
13. 25:16Algorithmic Approach - "How to determine the fit": Interest-based or Relationship-based?
14. 28:17"How to communicate the algorithms' outputs": Directly invite newcomers Or Communicate with current members first?
15. 29:14Interface
16. 30:06Deploy the Recruitment Tool
17. 31:29Accuracy Evaluation
18. 34:41Acceptance Evaluation
19. 37:54Summary of findings
20. 40:59Work Evaluation in Wikipedia
21. 41:42Challenges of ML-based Evaluation System
22. 43:08On-going: Redesign Al Evaluation System for Wikipedia
23. 44:10Application on Algorithmic Content Curation

" related talks

• 31mA more human-centered approach to AIJess Holbrook · Big Data & AI Toronto 2020
• 19mAI in the Field: Humans at the Heart of ValueMarjory Canonne · Devoxx France 2026
• 37mFinding the Problem to Solve: Human-centered AIRaleigh Tomlinson · Data Con LA 2020
• 38mEthical Artificial Intelligence to Advance Welcoming Communities and Identify Potential Communica...Sean P. Goggins · Emerging OS Forum @ Open Source Summit North America 2022
• 01h 00mAI for Social Responsibility: Embedding (...) into AIVeronica Dahl · Code Mesh LDN 2019
• 16mMachine Intelligence With Humans in MindTannaz N. Roshandel · GeeCON 2023
• 26mHow to develop useful and unbiased AIZamina Ahmad · code.talks 2023
• 20mBridging the gap between AI and UILiad Magen · WeAreDevelopers AI Congress Vienna 2018