The US Politics Labeler, @uspol.blakeslabs.com, is intended to help with mental health issues related to United States politics. It's to help people who are fed up with US politics, for whatever reason: they don't live here, or they're tired of seeing the government trying to kill them, so on and so forth.
We try to give granular labeling, but we're not going to make a label for every single thing. If we notice a trend emerging in the miscellaneous category, we'll make a new label for it and (later, when the tools exist for it) re-label the appropriate posts.
We may not label:
- Posts with the Donald Trump (or Kamala Harris) label which already have other, more appropriate labels (such as if his name is only in passing during an encouragement to vote; this would be under the Elections label, and not the Trump label).
- Racial or LGBTQIA+ topics, if it is not from a prominent politician or in relation to legislation.
- Posts about Elon Musk, if they are not directly related to his position in DOGE or his alliance with the Trump Administration.
Posts without alt text cannot be reliably labeled automatically, so all political image-only posts must be hand-processed. Therefore, we recommend that you hide posts that have images but no or incomplete alt text. One way to do this is by subscribing to @baatl.mastod.one and setting "No alt text" to "Hide." Another way to do this is being proposed in Bluesky issue APP-4550.
We do take appeals. Please only use this if your post was not talking about any politically-charged topic in the US -- it doesn't matter if you really need people to see it; if they don't want to see it, congratulations, you're trying to violate their consent and liberty. If you do this repeatedly, your reports will be muted. Appeals are processed manually, so it may take some time for an appeal to go through. Don't spam it, just wait it out.
We also accept reports. Please only report posts that are talking about a US political topic or politician. When you do this, simply mark it as "Other". We don't need any reason. Reports are processed automatically, and if no criteria match, it is automatically escalated for manual review. Furthermore, reports may not be processed right away. The script checks for them on a timer, getting only a few at a time, and if it is escalated for manual review, that may take longer.
Please don't spam reports or appeals. It makes it harder to identify what triggered an initial report.
Account labeling is an extreme intervention which will typically cause many false positives. However, there are a few edge cases where labeling an account may be justified:
- An account which posts exclusively political posts, especially if it would be difficult to automatically label each post individually. Such accounts will typically have the "Miscellaneous" label.
- An account which posts primarily political content, and is politically branded (i.e. its name or avatar reflect common slogans, logos, or party or politician names). We do not label personal accounts, or accounts which appear to be personal, unless they belong to a politician. An example of that is any account which has "fuck Trump" or "MAGA" in the name, but seems to be a person's normal account. If the account represents a movement or institution, it MAY be labeled under this rule.
- An account belonging to a politician or political government entity. These accounts may be labeled by party and/or as "Miscellaneous."
- An account which intentionally evades the labeler. Warnings are issued first at the discretion of the operator (of whether or not the evasion is intentional; there are many cases of people trained to avoid algorithmic censorship), and if evasion continues, particularly if some retributive action is taken (blocking, harassing, etc), the account will be labeled. Evasion of other labelers is not relevant.
We do not label personal accounts, except under exceptional circumstances, such as those listed above.
We do not label accounts based on the contents of their profile (nor do we label the profile). We're concerned with the contents of posts.
Not all posts are received by the labeling script. This is not our fault, and we cannot fix it at this time, although we are attempting to extend our reach. What the labeling script does find and receive will be labeled, and that is supplemented by your reports.
Media with no alt text cannot be appropriately labeled due to technical and financial constraints (this would cost THOUSANDS of dollars per month). We recommend hiding all such media. We do not have a mechanism we can recommend to do this at this time, though.
This is a living document, and it will be subject to many revisions over time. Donations are appreciated and the source code has been published, though the configuration (keywords) will not be.
