Perspective

Perspective is a machine learning-based tool designed and developed by Jigsaw to help identify and flag toxic and abusive language online. 

Available by API, the system uses natural language processing and machine learning algorithms to analyse text and predict the likelihood that it will be perceived as toxic or abusive by a human reader.

It also takes into account the context in which the text is being used, including the conversation history and the relationships between users.

The API provides a score, known as the "toxicity score", which indicates the likelihood that a piece of text will be considered toxic or abusive. The score is based on a range of factors, including the language used, the tone and the context.

Risks and harms 🛑

Jigsaw's Perspective API is associated with several risks and harms, including inaccuracy and bias, resulting in poor or unfair moderation decisions that negatively impact users and vulnerable groups in the form of bullying, harassment and other physical, psychological and human rights harms. 

The API could be used to silence marginalised voices if implemented without careful consideration of its limitations.

Research, advocacy 🧮

• Friedl P. Dis/similarities in the design and development of legal and algorithmic normative systems: the case of Perspective API

• Raman V. et al. Centering the Margins: Outlier-Based Identification of Harmed Populations in Toxicity Detection

News, commentary, analysis 🗞️

• https://www.fastcompany.com/90929549/google-jigsaw-toxic-speech-ai