[ * * Announcement * * ] Fair and Unbiased AI data: Lack of Mandarin Speakers Representation

edited July 2021 in General

My name is Snehesh Shrestha, and I am 5th year PhD student working in the area of Artificial Intelligence for Human-Robot Interaction. As you might already know that data used in AI is biased, therefore inheriting human bias that tends to propagate into technology of the future. One of the issue we find from meta-analysis of academic papers is that there is a lack of diversity in the participants in academic research studies. In my work, we argue that it is not sufficient to design algorithms with one-model-fits-all policy without proper representation in data. So we started a long journey of collecting human-robot interaction data of native English speakers and non-native English speakers. To demonstrate this gap, we find that even though there is a large number of native Mandarin speakers in the US, there is a disproportionately low representation of Mandarin speaking participants in the data.

In this National Science Foundation (NSF) sponsored research, we have collected 77 participants interacting with a kitchen robot. Due to the COVID-19 pandemic, we had difficulty recruiting participants to visit the lab. We are looking for help from you as a native Mandarin speakers to represent the community in AI data. We just need 5 more participants to complete the study.

We need your help in representing native Mandarin speakers in AI data. This is a paid study at $10 for the hour (the study usually lasts about an hour or less) where participants interact with a robot agent. Additionally, you will also automatically be added to a $50 raffle lucky draw. To qualify for the study, participants must be 18 years or older, must be native Mandarin speakers, and must have been in the US less than 2 years (a few months more might still be eligible), and be willing to visit the UMD College Park campus lab for approximately 1 hour.

For more information and as well as to sign up, you can visit: http://snehesh.com/hri/ or they can email me directly at [email protected] if they have any questions.

Thank you in advance for your help!


Sign In or Register to comment.