The internet is an ecosystem with a rich diversity of content, both enlightening and damaging. The problems of misinformation, fake news, deepfake images, and videos have all worsened due to the rise of AI, making internet navigation a potentially harrowing experience. The malign influence of AI on internet information pollution isn’t easily attributable to a specific tech giant, though Google is making efforts to counteract the trend via its Jigsaw unit.
Jigsaw has introduced an advanced edition of its artificial intelligence application, Perspective. Perspective will assess the level of toxicity in a person’s social media posts and gauge the potential impact of a toxic post on others. This capacity enables Perspective to be simultaneously a social media analyst, researcher, literature expert and counselor.
The primary mission of Google’s Perspective is to combat misinformation, censorship, internet toxicity, and violent extremism. Originally introduced in 2017, ahead of the current boom in the field of AI, Perspective was in need of a significant update, which it has now received. The tool operates as an API, like a third-party plugin for a website. Users need to log into a Google account, establish a Google Cloud account, and then download the API.
Perspective primarily evaluates text comments, whether they’re posted on news sites, forums, blogs, or chat messages. The application will determine the character of the comment and assign a label accordingly. Abusive or discourteous comments are flagged as either “Toxic” or “Severe Toxicity”. Comments that single out someone based on their identity receive an “Identity Attack” label. Comments featuring vulgar language are awarded a “Profanity” label.
For well-behaved online citizens who don’t leave offensive comments, Perspective still provides labels such as “Affinity”, “Compassion”, “Curiosity”, or “Respect”. In addition to the label, posts deemed to be toxic are assigned a score on a scale from zero to one. Zero signifies a low likelihood of toxicity, while one suggests a high likelihood. A score of 0.8, for example, would be deemed significantly toxic, while 0.1 would be considered mildly toxic.
Despite the apparent efficacy of Perspective, Jigsaw cautions users about potential limitations and errors. The tool has trouble analyzing overly lengthy comments and cannot support certain languages. Regardless, Perspective can serve as a practical resource for academic researchers studying the connection between social media content and algorithms.
