BiasLenz

Team Name

Questionable Research Labs

Team Members

Project Description

BiasLenz: Unveiling Media Bias through Imagery

In the midst of looming elections, the way different media outlets depict news stories has never been more apparent. Picture this: as you scroll through your news feed, your attention is primarily drawn to two elements - the headline and the featured image. This initial impression forms the lens through which you view the entire article, with the image often carrying more weight. For example, a disheveled portrayal of Judith Collins juxtaposed with Jacinda Ardern's confident pose and winning smile can profoundly shape your interpretation of the accompanying information. However, this visual framing, often aligned with political allegiances, can significantly bias your perception of the story's actual content.

Empowering Critical Thinking through BiasLenz

In this landscape, cultivating critical thinking becomes paramount as a defence against disinformation. This is where BiasLenz steps in. BiasLenz is a sophisticated tool meticulously designed to spot and analyze biases in news photography. It diligently compares image datasets associated with different politicians and various media sources. Through this analysis, BiasLenz empowers us to critically assess how media outlets shape our perceptions and influence our opinions. Our publicly accessible dashboard takes a deep dive into the realm of political imagery employed by mainstream media outlets, offering a straightforward method to gauge their overall bias. BiasLenz aims to illuminate the subtle yet influential ways media bias can shape our understanding of news and encourages a more discerning and informed readership.

Data Story

To find the biases of News platforms, we first need to build up a database of News Photos. To do this, we used a combination of Google's custom search API, and scraping RSS feeds.
We find faces in the news photos, crop them, and then match the faces to the Parlament Services official MP list. Once we had a labelled list of MP photos from different orgs, we used a Neural Net to scan the headshots for positive and negative images.