Research:The Practice of Being an AI Participant Through Data Contribution (interview study)
This page documents a research project in progress.
Information may be incomplete and change as the project progresses.
Please contact the project lead before formally citing or reusing results from this page.
As human-centered AI has become an emerging field, researchers are developing innovative methods for incorporating human perspectives into the AI pipeline. As such, participatory AI (PAI), an intersection between participatory design and artificial intelligence development methods, has emerged as a unique mechanism for incorporating diverse stakeholders into the development of technologies that directly impact their lives. However, PAI is not without its challenges. Due to the technical nature of AI development, previous work has noted the extra work that practitioners and participants must engage in for successful outcomes. Moreover, this extra work can disproportionately fall on communities whose voices are already in the margins of human-centered practices.
We are fundamentally interested in understanding the vast landscape of being an “AI participant.” We hypothesize that participatory AI (PAI) is a burdensome method for these participants, who are often already negotiating systemic harms from emerging technologies. To that end, our findings will inform complementary methods to participatory AI that incorporate diverse perspectives without placing an undue burden on participants. Moreover, researchers should incorporate support structures that relieve participants from anticipated burdens when engaging with PAI methods.
To study this, we would like to interview AI participants across three large categories: Researchers and Developers, Data Contributors, and Advocates. The term “AI participant” has been used to describe a large range of roles. However, participatory methods have been criticized for overly focusing on the end-user. Many data contributors, such as writers and Wikipedia editors, have recently advocated for larger consideration within AI development. Meanwhile, researchers have their own positionalities that affect the design process, and advocates often bear the brunt of activism in this emerging problem space. In short, we hope to provide a nuanced and balanced perspective of participating in AI that can inform future PAI methods.
Methods
editWe will interact with participants by carrying out online, semi-structured interviews on Zoom. Each interview will be about an hour long. These interviews will be based on past participant experiences and informed by the intake survey. Our consent form and intake survey can be found here. There are no other interventions we will use in this study.
Timeline
editWe are currently in the interview stage of this research (Jan - April 2024), where we are conducting and analyzing interviews. Starting May 2024, we plan to begin writing a research paper based on a thematic analysis of the interviews. We will post preliminary results as blog posts and ensure that accepted research manuscripts are freely available to the Wikipedia community.
Policy, Ethics and Human Subjects Research
editThis work was approved by the University of Minnesota IRB on October 31, 2023 (IRB#: STUDY00020382).
Results
editNo results yet. Stay tuned, though!
Resources
editWe plan to release summaries of our results as blog posts on the GroupLens Research blog at grouplens.org