Fact-checking API

[AdamSobieski]

Introduction

Using combinations of local and remote artificial-intelligence services, end-users could more easily fact-check content that they are either reading or writing in a Web browser.

Previous Works

• Citation Needed
• Add a Fact

[christianliebel]

I'm not sure if we need a web platform API for this. This sounds more like a capability that is best provided by the user agent or through an extension.

Given that AI models can hallucinate, facts could be misclassified. A key consideration is which model or service will be used to define what constitutes a "fact." This definition can vary based on the model's origin, particularly in areas like geopolitics.

I think users should explicitly choose to enable such a capability and decide which service or model to use. Implementing a web platform API that categorizes information as simply "true" or "false" sounds dangerous.

[AdamSobieski]

On these topics of AI and fact-checking, I would add that, in addition to browser-integrated, user-selectable, user-configurable, AI assistants (e.g., text selection + context-menus), there are emerging social-media-platform-integrated solutions such as end-users being able to use the @ symbol to invoke one or more named AI assistants and subsequently ask them whether a preceding social-media post of interest is factually correct or verifiable.

Previously, I had considered UXs involving end-users making text-based (and image-based) selections in Web browsers and accessing fact-checking and verification services by using context-menus, across websites. I am impressed by the unfolding development of end-users being able to invoke social-media 'bots, by name, to request these services.

To your points, beyond categorizing information as simply being either "true" or "false", AI assistants can respond more fully, making use of forum contexts, providing conversational fact-checking and verification services. They could do so either in direct messages with end-users or in online forum threads.

According to this preprint, analyzing 1,671,841 observations of AI-related fact-checking requests made of two AI 'bots on the X platform, there appears to be a demand for AI fact-checking on social media.