⚙️Expert Engine

DRAFT - DRAFT - DRAFT

The Expert Engine is at the heart of the Confidently project. It is a set of instruments that, starting from the Data Input is able to produce the scoring for a textual resource.

The Expert Engine is composed of:

• Simple Reliability Data: simple indicators derived from the resource itself or its URL

• Complex Reliability Data: complex indicators produced by the AI analysis of the resource

• User-Generated Reliability Data: user-generated scoring by supervalidators and validators working on the platform

By resource we mean a news in textual format. By reliability data or trustworthiness data we mean all the numerical indicators of trustworthiness of a certain resource. A resource is considered to be reliable when the weighted score of its indicators signals a very low probability that it is a fake news or misinformation.

Simple Reliability Data

The model we follow shares some similarities with Iffy.news and their attention to "promising indicators". It is important to perform some initial evaluations against the domain name of the resource and the author/newspaper itself.

• Date of domain registration and ages of online presence

• Presence of the domain in accredited journalism lists (e.g. E&P Databook or FIEG for Italy).

• Presence of the domain or newspaper in Wikidata, for example: La Repubblica (accredited Italian newspaper)

• MOZ domain authority score and backlinks

• Historical view of DNS changes via SecurityTrails or DomainTools

• Social present, social media account age, engagement rate

• Presence of the website in a blacklist such as the Iffy Index or Bufale.net for Italy

• Google Fact Check Explorer (working API demo completed already)

• NewsGuard scoring

• Tranco List

Complex Reliability Data

TODO

User-Generated Reliability Data

This component represents the reliability data of a news source generated by users of the platform (validators, supervalidators) and it has mainly two purposes:

• it concurs with a very high weight to the final scoring

• it trains the AI models for Complex Reliability Data