Institutionalizing ethics in AI through broader impact requirements

Abstract

Turning principles into practice is one of the most pressing challenges of artificial intelligence (AI) governance. In this article, we reflect on a novel governance initiative by one of the world's largest AI conferences. In 2020, the Conference on Neural Information Processing Systems (NeurIPS) introduced a requirement for submitting authors to include a statement on the broader societal impacts of their research. Drawing insights from similar governance initiatives, including institutional review boards (IRBs) and impact requirements for funding applications, we investigate the risks, challenges and potential benefits of such an initiative. Among the challenges, we list a lack of recognised best practice and procedural transparency, researcher opportunity costs, institutional and social pressures, cognitive biases, and the inherently difficult nature of the task. The potential benefits, on the other hand, include improved anticipation and identification of impacts, better communication with policy and governance experts, and a general strengthening of the norms around responsible research. To maximise the chance of success, we recommend measures to increase transparency, improve guidance, create incentives to engage earnestly with the process, and facilitate public deliberation on the requirement's merits and future. Perhaps the most important contribution from this analysis are the insights we can gain regarding effective community-based governance and the role and responsibility of the AI research community more broadly.

Operationalising AI governance through ethics-based auditing: an industry case study

“…Fortunately, companies need not start from scratch: numerous translational mechanisms for AI governance have been proposed and studied [ 55 – 57 ]. These include impact assessments lists [ 58 – 60 ], model cards [ 61 ], datasheets [ 61 – 63 ], as well as human-in-the-loop protocols [ 64 ], standards and reporting guidelines for using AI systems [ 65 – 67 ], and the inclusion of broader impact requirements in software development processes [ 68 ]. 5…”

Towards a standard for identifying and managing bias in artificial intelligence

“…Identifying and addressing potential biases is an important step in the assessment process. There is currently momentum for AI researchers to include statements about potential societal impacts [225] when submitting their work to journals or conferences. Similar to privacy impact assessments, which are relied upon by data protection and privacy frameworks to gauge and respond to data privacy risks, such impact assessments provide a high-level structure that enables organizations to frame the risks of each algorithm or deployment while also accounting for the specifics of each use case.…”

Towards Responsible Natural Language Annotation for the Varieties of Arabic

“…Other practices to aid the practitioner in envisioning the possible impacts of their work include: staged system roll-out or prototyping to get ahead of any unforeseen issues before full launch, and performing an impact investigation. Impact investigations are worthwhile, though they are neither simple nor straightforward and there are no clear norms (Prunkl et al, 2021;Partnership on AI, 2021).…”