In this episode, we look at the document dump provided to Project Veritas. The documents contain references to partisanship, psychological research, fake news, politics, machine learning ‘fairness’ and more. For some reason (although not surprisingly), these have not been thoroughly disseminated anyone in the media — so, let’s delve in.
What is ‘twiddling’? Re-ranking of results from a single corpus (i.e. an existing search). In the doc, there are references to the ease with which twiddlers can be created, and that an example function is to ’strongly demote results’. In other words, Google clearly manually intervenes in its search results. (On December 11, 2018, Sundar Pichai appeared in front of Congress, under oath, stating that Google doesn’t intervene in its search results).
We also look at the ‘Fringe ranking classifier‘, and the leaked documents discuss ideas on how to identify ‘fringe’ news’, and one idea is to directly grade channel quality. Also in the dump, we learn about machine learning fairness; Google’s Orwellian attempt to defy reality by ‘trying to show search results that do not reflect any bias or unfairness‘.
Supporters of taim.io.