15 Comments
Mar 21Liked by Caitlin Dewey

I like the perspective, yet I would like to add that 1990s Internet wasn't nearly as monetized as it is now. Therefore the incentive to keep people on websites (and algorithms rewarding disagreement - see Jaron Lanier's arguments on social media) wasn't as high. Yes, that means more visibility as already pointed out, but also that parts of the Internet have been actively designed to be worse!

Expand full comment

I was not on any political newsgroups in the '90s but I was on Usenet. Google Groups used to support Usenet (sunset February 2024) and about five or six years ago, I went searching for myself and hoo boy, the CRINGE of my posts. I was so young.

Expand full comment
author

This is the #1 reason why I am glad Livejournal is a shell of its former self, lol. But wait am I reading this correctly -- did Usenet just became inaccessible?

Expand full comment

No, you can still access Usenet. It will probably never die. lol (Though it seems to be filled with pirated video, audio, and warez. Not that I know anything about this.)

It's Google Groups that dropped support as of February 2024.

Expand full comment
Mar 21Liked by Caitlin Dewey

Great piece of reporting and insights. We do a lot of social listening/monitoring at work and use an AI tool. What’s interesting is not the positive or negative sentiment we uncover but the neutral. The listening platforms are terrible at picking up sarcasm, for example, so it often ends up as “neutral.” I think the death of a thousand cuts comes from that squishy middle, the tiring back-and-forth of it all.

One thing the reporting and study bears out is that (looks around room) most of us probably just move on with our lives. And AI isn’t (yet!) telepathic; it’s never going to tell us what we think or feel afterward, unless you’re part of a motivated base, then you’re probably running hot a lot.

Expand full comment
author

Huh that's interesting re: sarcasm. Also a tricky distinction for second language learners, if I'm not mistaken!!

Expand full comment

Exactly. I would be interested to know whether the tools the researchers used captured traditional sentiment, though likely not since they aren't marketers :) But even more broadly, how is machine learning or LLMs capturing the nuance of the equivalent of eye rolling and shade.

Expand full comment

Fantastic! I wonder what this study would say about Substack? Are we deluding ourselves that it’s this happy place?

Expand full comment
author

IS this the happy place? Nazis etc? 😬😬

Expand full comment

A second, more considered answer: for me anyway, the Nazi furor stayed off on the fringes of my use of Substack. I don’t spend a lot of time in Notes, so it just never occupied center stage.

Expand full comment

How quickly I forget!

Expand full comment
Mar 21Liked by Caitlin Dewey

This is so counterintuitive and wonderfully hopeful. Thank you. Yes, we can raise the level of discourse! Yes, communities can develop and enforce standards of comportment while still promoting and supporting the free exchange of thought and ideas. Yay!

Expand full comment

I have to wonder, however, if some context to understanding the language used three decades ago (by Italian researchers, no less!) isn't missing. Calling someone "retarded" didn't carry the heft of damage in 1997 that it does in 2024.

Expand full comment
author

The evaluation of the language is automated, so I don't think the geography of the researchers makes a difference ... I actually just tried to test a phrase with the r-word on the Perspective API site, but it's not loading for whatever reason. https://www.perspectiveapi.com/

Your point about language change over time is interesting. I'll have to think on that. I'm inclined to think that's an issue of degree, though, not of kind. R-word might be considered "more" rude now, but it was still rude in 1992. At least I assume it was. I was three years old. 😂

Expand full comment
deletedMar 21
Comment deleted
Expand full comment
author

K., it actually sounds like you agree with the researchers! Were you in any Usenet politics groups? I'm curious because that's the one type of community, in this study, where conversations got LESS toxic the longer they went on.

Expand full comment