QUALITY OF RESOURCES DETERMINED BY PRISMA
Preferred Reporting Items for Systemic Reviews & Meta-Analysis
Systematic quality selection of standards, laws, experts, and guides was conducted similar to Jobin, Ienca & Vayena’s (2019) method. However, we focused our efforts on both “grey literature” and academic publishing and distribution channels. Grey literature are materials and research produced by organizations outside of the traditional commercial or academic publishing channels. Common grey literature included reports, working papers, governments documents, white papers and evaluations. So both academic and grey were considered depending on quality.
Quality of AIS ethics resources in this portal (excluding AIS Experts, that preliminary selection process is described here) was adapted from Floridi & Cowls (2019) process for inclusion of principles into their unified framework where each publication on ethics principles had to meet three basic criteria: “they are recent, published within the last three years; directly relevant to AI and its impact on society as a whole; and highly reputable, published by authoritative, multi-stakeholder organizations with at least national scope.
A scoping review of the gray and academic literature for standards, laws, principles and guidelines for ethical AIS resources was conducted. A scoping review is a method aimed at synthesizing and mapping the existing literature. Considered particularly suitable for complex and heterogeneous areas of research, this method is especially well adapted for evaluating the thousands of continuously proliferating texts and other media that are focused on AIS ethics.
There is still an absence of rigorous and globally accepted unified databases for AI-specific ethics resources. As such, in addition to the scoping review; we developed a protocol for screening and eligibility, adapted from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework.
Following best practices for gray literature retrieval, a multi-stage screening strategy is shared. Both inductive screening and eligibility were considered via search engines. Then deductive identification and data organization of the relevant entities with associated websites and online collections was conducted. To achieve comprehensiveness and systematicity, relevant document evaluation and database preparation was completed by relying on three sequential search strategies: (1) manual search of linkhub websites (2) manual keyword Google search of excerpts and full texts and (3) citation chaining to systemically screen full texts. Content analysis of sources including both grey and academic literature was independently conducted by two researchers in two cycles of manual coding.
REFERENCES
