Language: English (mozilla)
This session will be a hands-on exploration and solution-generating session focused on political and ideological bias in image search. We will briefly introduce the topic with some examples of two problems: 1) partisan over-representation in searches related to topics like government, elections, and voting and 2) explicitly partisan images in searches for unrelated, non-political topics. Small groups of participants will then examine and discuss additional case studies in breakout rooms and share their thoughts on possible root causes and potential solutions. After sharing back these case studies, participants will help quantify the scope of the problem by conducting some standard searches for their local geographies and completing a table to summarize relevant findings. We will close by revisiting proposals for potential solutions and next steps.
Outcomes: to increase awareness of political / ideological bias as a relevant dimension of representation in image search; to begin to quantify the scope of the problem, especially by expanding examples beyond a primarily US context; and to tap into the insights of the crowd about potential solutions and new partners to engage with this work. The goal of the work overall is to improve the usefulness of image search tools (with a focus on openly-licensed images)
Why did you choose that space? How does your session align with the space description?:I chose the misinformation / disinformation space for this proposal because I believe amplification of the prevalence of partisan images is a form of disinformation. For example, images sometimes appear to be amplified through the use of "spammy" or inauthentic metadata. There can be a fine line between effective search engine optimization and illegitimate forms of amplification. I'm interested in helping search tools address instances that cross the line into inappropriate amplification. In particular, the examples we will examine contaminate public discourse. Addressing those cases will increase the value of image search tools. The session also has some connection to Decolonized AI Futures, because artificial amplification can be seen as an active process of colonization. However, that is not the primary lens of the session.
How will you deal with varying numbers of participants in your session? What if 30 participants attend? What if there are 3?:The main part of the session where this could be a challenge is the first group activity with the case studies. I have about 5 case studies. Case studies could be examined by groups from 1 to 5, and if necessary more than one group could discuss the same case study. The second activity, quantifying the scope, is very flexible in terms of number of participants.
What happens after MozFest? We're hoping that many efforts and discussions will continue after MozFest. Share any ideas you already have for how to continue the work from your session.:Next steps may include actions that can be taken by individuals (content creation, metadata editing, reporting for moderation) and by organizations creating and maintaining image search tools ( metadata quality control, algorithm design, UX design, moderation enforcement).
What language would you like to host your session in?:English