[LOGOS] Trustnet

[Richard Carpiano]

Didn't have a chance to check out the study in detail, Indeed, really
problematic evaluation of it.

That was detailed in the original article through which I found the one I
sent you all and didn't read through because I got pulled into the MIT
press release.

Here is that broader article...
https://www.post-gazette.com/news/health/2024/06/22/online-misinformation-tools/stories/202406230019


it gets into the many downsides but also discusses this:

"Althoff and fellow researchers Xinyi Zhou, Ashish Sharma and Amy Zhang
created a program similar to Trustnet but with an emphasis on using AI to
detect misinformation. Althoff said it’s complementary to Trustnet.

Called MUSE, the tool detected misinformation and provided accurate
corrections to online statements better than ChatGTP 4 and the highest
rated fact checkers on Community Notes. An academic paper on MUSE
<https://arxiv.org/pdf/2403.11169> is currently under review for
publication, said Althoff."

MUSE supposedly outperforms ChatGPT and humans but based on my quick skim,
also a relatively small n evaluation.
Rich
_________________________________________________
Richard M. Carpiano, PhD, MPH
Professor of Public Policy
Faculty Director, Science to Policy (S2P)
University of California, Riverside
School of Public Policy
900 University Avenue
Riverside, California 92521 USA

Office: 4157 Interdisciplinary South
Phone: 951-827-5405
E-mail: richard.carpiano at ucr.edu
Website: https://profiles.ucr.edu/app/home/profile/rcarpian
X (Twitter): https://www.twitter.com/RMCarpiano
<http://www.twitter.com/RMCarpiano>
Bluesky: https://bsky.app/profile/rmcarpiano.bsky.social
LinkedIn: https://www.linkedin.com/in/richard-m-carpiano-36261b30b


On Sun, Jun 23, 2024 at 7:04 AM Kevin M Esterling via LOGOS <
logos at lists.ucr.edu> wrote:

>
>
>
> This is indeed an interesting project, but not for the reasons that the PI
> intended. Instead, there doesn't seem to be any communication theory or
> design principles that would ensure the extension is measuring
> misinformation, which is what they intend. Just wishing users will label
> misinformation does not mean that is what they will do, but instead they
> will certainly tend to label content they don't like, and then those biases
> will propagate to the echo chamber of their "trusted" friends. And then
> their validation study had an N of 32 labeling a total of 64 items. It
> really is amazing to me what MIT folks can get away with....
>
>
>
> Kevin M. Esterling (he/him)
> Professor of Public Policy and Political Science
> Director, Laboratory for Technology, Communication and Democracy
> (TeCD-Lab)
> https://tecd-lab.ucr.edu/
> UC Riverside
> Riverside, CA 92521
> 951-827-5591
> https://profiles.ucr.edu/kevin.esterling
>
>
> What happened when Twitter deplatformed 70,000 misinformation traffickers?
> <https://www.nature.com/articles/s41586-024-07524-8>
>
>
> ------------------------------
> *From:* LOGOS <logos-bounces at lists.ucr.edu> on behalf of Richard Carpiano
> via LOGOS <logos at lists.ucr.edu>
> *Sent:* Sunday, June 23, 2024 4:20 AM
> *To:* Logos Listserv (BCOE) <logos at lists.ucr.edu>
> *Subject:* [LOGOS] Trustnet
>
> Perhaps of interest... Rich
>
> Trustnet: A new web browser extension to help identify misinformation for
> the average person...
>
>
> https://news.mit.edu/2024/new-tool-trustnet-empowers-users-to-fight-online-misinformation-0516
>
> _____________________________________
> Richard M. Carpiano, Ph.D., M.P.H.
> Professor of Public Policy
> Faculty Director, Science to Policy (S2P)
> University of California, Riverside
> School of Public Policy
> 900 University Avenue
> Riverside, California USA 92521
>
> Office: 4157 Interdisciplinary South
> Phone: 951.827.5405
> Fax: 951.827.2303
> E-mail: richard.carpiano at ucr.edu
> Website: https://profiles.ucr.edu/app/home/profile/rcarpian
> X (Twitter): https://www.twitter.com/RMCarpiano
> Bluesky: https://bsky.app/profile/RMCarpiano.bsky.social
> LinkedIn: https://www.linkedIn.com/in/richard-m-carpiano-36261b30b
> _______________________________________________
> LOGOS mailing list
> LOGOS at lists.ucr.edu
> https://lists.ucr.edu/mailman/listinfo/logos
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ucr.edu/pipermail/logos/attachments/20240623/9b9cf71f/attachment-0001.html>