Publication Bias in the Output of Generative AI

Abstract

Throughout this study it became apparent that Generative artificial intelligence lacks the ability to identify publication bias and is not always able to grasp the central message, despite several attempts and methods seeking to correct such publication bias. Due to the training of AI which consists of consuming information from the internet and providing information by predictive algorithm methods the presence of bias was expected. The goal then was to investigate if generative AI, using ChatGPT as our tool, could correct its outputs based on inputs provided. The subject matter focused on research conducted in our laboratory, gadolinium based contrast agents (GBCAs), which there is significant misconception and disagreement in regard to safety concerns.

Ähnliche Arbeiten