Agenda for Actionable Guidance

In generating actionable guidance for responsible AI adoption and use, we remain committed to the goal of securing prosperous, equitable, and sustainable futures for all.

Problem Identification

We prioritise issues relating to the design, deployment, and use of AI use-cases that demand the most urgent attention, while factoring for their scale of deployment (i.e., number of users), magnitude of impact (i.e., number of affected groups), and nature of impact (i.e., social, economic, political, cultural, and environmental).

We make this determination through consultations with our affiliate network and external stakeholders, public surveys, and review of government documents (e.g., calls for public consultation, request for proposals by governments), newspaper reports and relevant primary and secondary research.

Guidance Generation

We mobilise members from our affiliate network and external experts, while accounting for their previous work, interest areas, and time commitment to constitute a dedicated working group to debate and discuss an identified issue in greater detail, define it with desirable precision, and formulate and propose suitable multi-stakeholder-led responses or measures to address it. The group receives research, editorial, and other administrative support, from AI Policy Exchange volunteers, as needed. The group remains active for as long as the successful completion of its mandate demands.

The preliminary version of the guidance is made available for stakeholder appraisal via social media handles of AI Policy Exchange. Feedback is solicited via comprehensive feedback forms and through policy parleys. Based on this feedback, the concerned working group attempts to make amends to its proposed guidance before signing it off for adaptation and use by relevant public and private sector actors. The proposed guidance may include suggestions for public and private bodies to implement or facilitate necessary social, legal, economic, political, or cultural reforms reforms.

Guidance Communication

We realise that mobilising stakeholder cooperation and consensus on our proposed corrective responses or measures to address an issue is critical for our proposals to actualise and achieve their intended outcomes and impact. Hence, we do not limit the communication of our guidance to the specific stakeholder for whom it is originally designed for adoption and implementation: rather, we proactively ensure that they are suitably extended to all stakeholders, including regulators and public policy decision-makers, public and private enterprises, technology workers, users of AI-based products and services, AI policy think tanks and advocacy groups, the press, and the general public.

Our formats for guidance communication, therefore, vary depending on the bandwidth (technical expertise and attention time) of the target audience. They could include long technical reports, white papers, policy memos and briefs, short audio-visual explainers, commentaries, lightning talks, infographics, and guidebooks.

Lastly, we caveat our proposed guidance with proper disclosures about underlying assumptions and limitations. This, we believe, is critical for enabling stakeholders to accurately assess the merit and utility of the guidance given the policy or business context in which it is sought to be applied, and adapt it accordingly to suit contextual realities; after all that there is no such thing as perfect guidance.

All our policy communication materials are released on the AI Policy Exchange portal for open access under a CC BY 4.0 International License, in downloadable formats. Moral rights over our policy communication materials jointly reside with all contributors.

Exceptional Guidance

We realise that policy issues sometimes might arise without an advance warning. This can happen, for instance, when a certain AI use-case is deployed by a government department or agency without adequate risk assessment or when a ruling government forces a regulatory overhaul without adequate stakeholder consultation or forethought about regulatory impact. In these exceptional circumstances, given that there is not enough time to conduct thorough policy analyses, we leverage our affiliate network to convene closed-door roundtables to dissect the issue at hand and explore possible courses of action for prevention of any imminent risk of harm. 

We then publish our policy recommendations based on the insights that emerge from these roundtables, further substantiated by our existing body of work (published or unpublished) as it relates to the issue, in the form of open letters, advisories or quick issue briefs for review by relevant public policy decision-makers and other concerned stakeholders.