Colby Brown @colby@transportation.social

The fact that the developers of Lemmy also run an instance whose logo is a tank, a symbol of the tool of oppression used to literally crush democratic dissent in China, seems very significant. And it's not just symbolism--they have refused to deal with serious content moderation issues on their site from its authoritarian leftist users specifically. I'm hoping that #Kbin becomes more stable and I'll check it out eventually, but for now I'll get my link aggregation fix from Hacker News.

(4/4)

Granted, machine learning bots aren't the only answer to content moderation and intelligent, compassionate human response to reports are important. We've had hiccups with our bot but because I'm in control of it and not some faceless admin at a corporation, I was able to tweak settings as needed to get it to distinguish users venting from actual toxicity.

(3/4)

And that kind of gets to the heart of why software isn't politically neutral--at least not social media platforms. Decisions that companies and developers make about APIs have huge impacts on how easily moderators can monitor and deal with hate speech or harassment. That in turn has a huge impact on how safe women and minorities of all kinds feel on social media, I believe.

(2/4)

As a former Redditor, I'm watching the API change protests crying into my 🍿 (figuratively speaking). As a fediverse instance operator, I'm looking at Lemmy and even though supposedly "software is politically neutral" it's a hard nope if for no other reason than there's no Python wrapper for the client API yet. The ease with which I was able to throw together a content moderation bot using the Detoxify library was important to my decision to host a mastodon server.

(1/4)

We can definitely save downtowns by killing telecommuting /s 😂

https://www.tiktok.com/t/ZT81Xsfww/

‘Only 11 other people have completed it’: cycling the 7,600km European Divide

https://www.theguardian.com/travel/2023/jun/08/cycling-the-great-european-divide-long-distance-ride?CMP=Share_iOSApp_Other

https://www.tiktok.com/t/ZT81dFCXA/

Equity is a big concern, too. If the most effective GHG reduction policies are highly regressive, i.e. disproportionally hurting lower-income households and benefiting high-income ones, it's hard for those of us with a conscience to loudly take a firm stand in favor of them. But particularly when one of these policies--road pricing--generates revenue, it seems feasible to craft an implementation that has a redistributive element. (This idea isn't new.)

Bit of soul-searching about what to do about this. Most travel demand modelers are generally highly skeptical of their own model results, and that may be why we're a lot less vocal about them. I see parallels there with the ML research community and its apprehension about future AI. Except the open letter we need to write is probably in defense of our analyses, something that's really hard for a field as self-critical as travel forecasting.

Certain political types only want to hear about certain types of solutions, though. This mirrors what I see on social media. So many pro-bike advocates, urbanists, and train aficionados... but very seldom do I see EVs or teleworking mentioned in a positive light on Mastodon or the birdsite.

It's not that we analysts don't also see benefits from transit projects, bikes, and compact walkable cities--but they don't rank among the most effective solutions, in terms of ability to reduce VMT and GHG emissions.

Fascinating discussion today at #TRBITAP conference about the unpopularity of the transportation GHG reduction measures that our best analysis tools tell us are most effective: pricing, telework and electrification of the vehicle fleet. Staff from multiple agencies across the country reporting similar findings.

If you're in Indianapolis for #TRBITAP (aka the TRB conference formerly known as Innovations in Travel Modeling) come stop by my lightning talk and poster at 3:30-5 (session C3) today to say hi and learn about applying GPT language models to travel analysis!