I will be presenting it at CHI, a premier academic conference for work in human-computer interaction, in May.

These people deserve to be treated better and appreciated for their work. But the vast majority of people don’t know. This information will not be new to those people who have ever been moderators. The research I want to introduce today is about how people become moderators and the range of difficulties they experience. Some of it is about building statistical models and unique data-driven interfaces.

Some of it is about relational work between moderators, streamers, and viewers. Some of it is about the technical aspects and how bots are helping out human moderators. In the past three years, I have been looking at different aspects of content moderation. How can we learn from the practices of human moderators to help design better tools from them and/or offset their load? I quickly learned that there are things machines are not going to be able to do and that the work of moderators is much more complex than pushing a button to delete a nasty comment. I started out my research project on content moderation with the naive goal of trying to improve artificial intelligence. If you think what you see online is bad, remember that it is the cleaned-up version. Nasty comments online are nothing new, but on livestreaming platforms like Twitch, they are more challenging to deal with because they happen in real time. There is someone, often behind the scenes, constantly cleaning up the trash, making sure that the majority of us have a pleasant experience. Maybe this is a bad analogy, but it’s similar to what we see online.

Remember when you were a kid, you would come home from school and your room would be “magically” clean? Now that I am an adult I know that it is not magic, it is a lot of effort that was done by a person and despite the development of technology, there are so many things that a Roomba just cannot do.

Twitch how to mod