Abstrakt: |
In the advent of the fourth industrial revolution, leaders in the commercial artificial intelligence (AI) market have shaped parameters of trustworthy AI design. Resulting directives, however, typically approach this issue from a technical perspective, while largely ignoring its human factors counterpart. As such, we conducted an information synthesis to capture the current landscape of user-centered design for developing trustworthy human-AI interactions. As part of our review, we analyzed resources from 50+ publications and summarized their respective protocols into three major categories: (1) design standards and guidelines based on ethical principles, 2) best practices for designing the nature of human-AI relationships across the user experience, and 3) best practices for designing redress mechanisms when trust is at risk, low, or broken. Part and parcel to this review, we provide concrete prescriptions for designing trustworthy human-AI interactions, identify limitations of existing protocols, and suggest areas requiring further exploration. |