Follow Alex, Aviva and Beatrice as they venture into the daunting world of A.I. in order to shed light on the intersection of artificial intelligence, journalism and bias.
In an effort to explore algorithmic auditing and AI in more depth, our team created a three-part podcast series that delves into the research-creation process of algorithmic auditing. Through this podcast, Alex, Aviva and Beatrice sought to explore the potential biases that may be present when using AI to generate content.
As A.I. and machine learning technologies become more sophisticated, the need for proper auditing becomes increasingly necessary. Algorithmic auditing refers to a range of approaches to reviewing algorithmic processing systems. It can take different forms, from checking governance documentation, to testing an algorithm’s outputs, to inspecting its inner workings
The podcast’s three episodes provide in-depth discussions and analysis of the project. The episodes detail the initial project plan, the creative ideation process, the production process, and the final results and takeaways from the algorithmic auditing performed.
- Episode One discusses and breaks down the group project, which is at the intersection of journalism and communication studies, speaking to issues of artificial intelligence and bias. It details the inspiration for the project and delves into the fundamental theories of the project, which are connected with the concepts and discussions had and seen throughout the course of the class. The episode also covers the team’s initial expectations and the original course of action we planned on taking. Inspired by questions surrounding bias and news, the project seeks to perform algorithmic auditing of headlines fed to an A.I. In this case, OpenAI’s GPT-3’s playground was used as the site of study in hopes to shed light on bias present in the news industry and/or within the artificial intelligence itself. The episode ends with a conversation surrounding the initial expectations team members had going forward with the project, setting the scene for the second episode, which dives into the process.
- Episode Two provides an in-depth look at the research-creation process and features live reactions to the algorithmic auditing session. Issues and roadblocks that were encountered throughout the process are discussed and examined. The team delves into the choice to stick with only OpenAI’s GPT-3, as well as the possible prompts that were optimal to use with the A.I.
- Episode 3 of the podcast provides an overview of the first two episodes and engages in a conversation about the future implications of using A.I. software such as GPT-3 in the
field of journalism. The episode also speaks to the issue of transparency in relation to journalism and AI, providing an interesting perspective on the matter.
Overall, this podcast series sheds light on the current state of algorithmic auditing and provides an in-depth analysis of the research-creation process. By discussing the implications of using artificial intelligence in the field of journalism, the podcast provides a unique and thought-provoking approach to the topic. It is clear that algorithmic auditing is an important tool in the fight against unethical bias and requires more research in order to be properly understood.