Designers should evaluate algorithms for accuracy, acceptability to users, and impacts on target performance to ensure comprehensive and effective tool development.
About this paper
The author critiques the prevalent machine learning-based approaches to algorithm design for their lack of user engagement and reliance on historical data, proposing instead a method called Value Sensitive Algorithm Design that incorporates stakeholders' feedback early in the process.
This novel method aims to balance multiple stakeholders' needs and includes an example project on designing intelligent socialization algorithms for WikiProjects in Wikipedia.
Here are some methods used in this study:
Which part of the paper did the design guideline come from?
“Results. We now present our answers to our questions. Question 1: What happened to the newcomers if they received invitations from WikiProject organizers? Specifically, do the newcomers participate and contribute? And how is receiving organizers' invitations different from receiving template invitations from the researcher team, or receiving nothing? In Models 1 and 3 (in Table 4 and Figure 3), we compared the within-project contributions of the six newcomer groups (i.e., NQ, Q/XX, Q/OI, (...)” (Section 4.6: Step 5: Evaluate Algorithms' Acceptance, Accuracy, and Impacts)
Zhu, H., Yu, B., Halfaker, A., & Terveen, L. (2018). Value-Sensitive Algorithm Design. Proceedings of the ACM on Human-Computer Interaction, 2(CSCW), 1–23.