AI-generated disinformation threatens to make debunking climate misinformation even more challenging, swamping already scanty capacity. We are exploring ways to counter this burgeoning flood of harm by employing the same tools now employed by malefactors as a force for good. We've already published research on detecting and categorising climate misinformation (and spoiler alert, we have new research coming out very soon on detecting logical fallacies). The next step is combining this research with recent developments in generative AI and large language models to automatically generate debunkings as we detect new misinformation. In other words, generative debunking.
We've reached the stage of the project where we are evaluating the output generated by our AI models. For that, we need people to help us evaluate our AI debunkings.
Generative LLMs are prone to failures best spotted by real people. Here, we need help with evaluating output from our newly developed chain of tools producing automated replies to climate bunk. This activity is especially suited to people having a fair degree of familiarity with climate disinformation and misinformation.
Activity details:
We estimate this assistance to require about 5 hours of contributed time. Helpers will be personally familiarized with the process by Dr. John Cook, founder of Skeptical Science and a leading climate disinformation researcher.
This activity is also a great way to get acquainted with Skeptical Science as a volunteer, leading to other opportunities to help fight Earth's climate crisis by fostering a better-informed public.
Interested?
We hope so. If you are, please let us know some contact details and we'll be in touch. Your contact information will not be shared with or used by anybody outside of Skeptical Science, Inc. and will only be used in connection with this activity.
[googleform form="1FAIpQLSc7bGSCGLGPkqKKP0_iq_il4Bcu6TSc3iVGThHYcViMp8HxiQ"]
Posted by John Cook on Monday, 4 November, 2024
 |
The Skeptical Science website by Skeptical Science is licensed under a Creative Commons Attribution 3.0 Unported License. |