EquiRank - A Research In Avoiding Algorithmic Bias on Instagram
Equirank is a concept leaderboard system embodied in Instagram and potential available for other social medias & platforms. It allows users to collect virtual points in order to engage communication between friends and redeem real-word prizes by reporting the potential algorithmic bias on the platform.
Feat.
Carnegie Mellon University 05-410: User Centered Research and Evaluation
Team Member: Siyi Liu, Daniela Munoz, Zulekha Karachiwall, Me
CLICK ONCE TO SHARE EQUITY.
Goals
Algorithmic bias can lead to unfair outcomes for certain groups of people and perpetuate discrimination and inequality. Biased algorithms can reinforce existing stereotypes and discrimination, which can have significant real-world consequences. Therefore, it's crucial for people to be aware of it on the internet.
Our research field is algorihmic bias on Instagram. Some existing functions or preferences on it is aggressive toward specific groups or races. Hence, we need to test and figure out a proper way to deal with this problem.
Automated risk assessments used by U.S. judges to determine bail and sentencing limits can generate incorrect conclusions, resulting in large cumulative effects on certain groups, like longer prison sentences or higher bails imposed on people of color.
“Algorithms are not inherently fair, they reflect the bias of the data they are trained on."
“Even when flaws in the training data are corrected, the results may still be problematic because context matters during the bias detection phase.”
User Surveys & Problems
In the primary phase, our team provided a survey for mid-heavy Instagram users to test if they are aware of the algorithmic bias, and their willingess to deal with it.
In the survey, we also asked their attitudes about the online educational contents, and what potentiallty interactive modes do they prefer on Instagram.
In summary, the users' willingness to participate in controlling the algorithmic bias is generally high. Most of them are eagering to be in a relatively fair online environment. On the other hand, their willingness to read educational contents are not high as expected. But most of them are glad to see short, visually appealing contents.
Knowing What To Do
During the primary pilot testing, we asked 4 participants who are medium to heavy Instagram users regarding their preference on algorithmic bias and online learning. From their reflection, we summarized two key elements that attracts online learning: Visual Appeals and Interaction Level
Affinity Diagramming
Takeaways
From this step, we designed and conducted several ideas and brainstormed the scenarios. We wanted to try most possibilities of the product and its potential usabilities for further research use.
We valued different hypothesis from those scenarios and created storyboards as the presentation of ideas. These ideas are then reviewed and evaluated by the speed dating session in the next phase.
Idea Expansion
During the storyboard & speeddating session, we created storylines and explore potential functions by separating the proposal into low, medium and high level of risks. This aims to generate new ideas towards the final design proposal about the functionality.
Crazy 8s
Storyboards
Takeaways
We found out that users prefer educational content that is concise, direct, and presented through interactive mediums like games or polls. They appreciate quick and clear feedback and are attracted to visually appealing content. Report forums that explain what and why to report are not engaging for users. We need to be cautious about using monetary incentives since users tend to prefer physical rewards rather than virtual points, except when there's a competitive aspect. However, using monetary incentives may increase the risk of misreporting. Lastly, different age groups may prefer different engagement techniques, such as articles and forums for older individuals and games for younger ones
Hence, we developed a Lo-Fi prototye and made a physical model of it. We selected the idea of reporting leaderboard with the mechanism of online ranking. Users can add their points to the scoreboard by reporting contents that contains algorithmic bias successfully (unsucceed reports do not count). We picked several contents and crafted them into the physical model to make sure they convey the idea of education of algorithmic bias well.
Final Takeaways
To enhance the impact of the leaderboard, consider incorporating more comprehensive profiles that include pictures and recognizable names. According to feedback received, participants were less motivated by the existing design as the profiles were unfamiliar. Additionally, create a section that clearly outlines the rulesets for moving up and down the leaderboard, which participants can refer to at any time.
To make the content more appealing, use colors instead of black and white visuals and include interactive features. Participants expressed a preference for interactive content over just visuals. Furthermore, avoid extreme themes and stick to more neutral topics that appeal to a wider audience.
Summary
Throughout this project, my main focus was on learning and gaining practical experience in user experience research. I was able to explore a wide range of methods and approaches for testing users, and through this process, I have gained a wealth of new knowledge and insights.
One of the most significant takeaways from this project was the importance of developing a deep understanding of the users. By utilizing various research methods such as surveys, user interviews, and usability testing, I was able to gather valuable information on user behavior, preferences, and pain points. This knowledge allowed me to develop a more user-centric design approach, resulting in a more effective and satisfying user experience.
Another key learning experience was the importance of utilizing multiple approaches to testing users. By utilizing a combination of qualitative and quantitative research methods, I was able to gather both subjective and objective data on user behavior, providing a more comprehensive view of the user experience. Additionally, I was able to better understand the limitations and strengths of each method, enabling me to select the most appropriate approach for each stage of the design process.
Overall, this project has been an invaluable learning experience for me, providing me with practical skills and knowledge that I can apply to future user experience research projects. I am confident that the insights and techniques that I have gained will enable me to create more effective and user-friendly designs in the future.
Final Poster