Human participants in AI research: Ethics and transparency in practice

TL;DR

This manuscript outlines a set of guidelines for ethical and transparent practice with human participants in AI and ML research to equip technical researchers with practical knowledge for their work, and to position them for further dialogue with social scientists, behavioral researchers, and ethicists.

Abstract

In recent years, research involving human participants has been critical to advances in artificial intelligence (AI) and machine learning (ML), particularly in the areas of conversational, human-compatible, and cooperative AI. For example, roughly 9% of publications at recent AAAI and NeurIPS conferences indicate the collection of original human data. Yet AI and ML researchers lack guidelines for ethical research practices with human participants. Fewer than one out of every four of these AAAI and NeurIPS papers confirm independent ethical review, the collection of informed consent, or participant compensation. This paper aims to bridge this gap by examining the normative similarities and differences between AI research and related fields that involve human participants. Though psychology, human-computer interaction, and other adjacent fields offer historic lessons and helpful insights, AI research presents several distinct considerations$\unicode{x2014}$namely, participatory design, crowdsourced dataset development, and an expansive role of corporations$\unicode{x2014}$that necessitate a contextual ethics framework. To address these concerns, this manuscript outlines a set of guidelines for ethical and transparent practice with human participants in AI and ML research. Overall, this paper seeks to equip technical researchers with practical knowledge for their work, and to position them for further dialogue with social scientists, behavioral researchers, and ethicists.

Figures (4)

• Figure 1: The automated annotation process applied to each paper. Boxes with sharp corners represent steps that produce annotations by applying custom prompts to the instruction-tuned large language model (LLM). Boxes with rounded corners depict steps that do not include the LLM. See Appendix \ref{['sec:human_data/prompts']} for the exact text used in prompts.
• Figure 2: The letter from the Center for Disease Control ending the US Public Health Service study at Tuskegee. Source: US National Archives.
• Figure 3: The "shock machine" used in Milgram's experiments. Source: saylor2012principles.
• Figure 4: The cover of the Belmont Report. Source: The Internet Archive.