RAI Guidelines: Method for Generating Responsible AI Guidelines Grounded in Regulations and Usable by (Non-)Technical Roles

TL;DR

This work presents a four-step methodology to produce responsible AI guidelines that are anchored to current regulations and usable by diverse roles. The authors (i) manually code foundational papers to extract active guideline techniques, (ii) assemble an initial 16-technique catalog, (iii) iteratively refine it through interviews and ISO/EU Act mapping to reach 22 guidelines, and (iv) finalize and exemplify them in an interactive tool evaluated with AI practitioners. In a two-phase study, the guidelines demonstrated regulatory alignment and broader cross-role applicability, achieving higher usability than a baseline checklist and facilitating early-stage ethical reflection. The research advances the concept of Responsible AI by Design by situating guidelines within regulatory frameworks and organizational workflows, offering concrete design requirements and a pathway for ongoing updates and accountability across teams.

Abstract

Many guidelines for responsible AI have been suggested to help AI practitioners in the development of ethical and responsible AI systems. However, these guidelines are often neither grounded in regulation nor usable by different roles, from developers to decision makers. To bridge this gap, we developed a four-step method to generate a list of responsible AI guidelines; these steps are: (1) manual coding of 17 papers on responsible AI; (2) compiling an initial catalog of responsible AI guidelines; (3) refining the catalog through interviews and expert panels; and (4) finalizing the catalog. To evaluate the resulting 22 guidelines, we incorporated them into an interactive tool and assessed them in a user study with 14 AI researchers, engineers, designers, and managers from a large technology company. Through interviews with these practitioners, we found that the guidelines were grounded in current regulations and usable across roles, encouraging self-reflection on ethical considerations at early stages of development. This significantly contributes to the concept of `Responsible AI by Design' -- a design-first approach that embeds responsible AI values throughout the development lifecycle and across various business roles.

Figures (4)

• Figure 1: Four-step method for generating responsible AI guidelines. These guidelines were derived from research papers, and are in line with ISO standards and the EU AI Act eu_ai_act_2022.
• Figure 2: Interactive Responsible AI Tool with 22 guidelines. The first part (A) allows for entering information about the developed AI system and (B) selecting the applicable user role. The second part (C) enables interaction with the guidelines. The third part (D) presents a summary of user responses for post-hoc reflections. Guidelines for other project phase can be viewed through the phase selectors (E/A).
• Figure 3: Guideline sorting procedure. Users can place a guideline in any of the three stacks (i.e., successfully implemented, should be considered, inapplicable) by: (1) considering two guiding questions, and (2) using the Yes/No buttons located next to the card and on the back side of it.
• Figure 4: SUS (usability) results. Guidelines are more usable than checklist items.