← Back to context

Comment by zamadatix

2 hours ago

No space em dash = a real person who has been bulldozed by LLMs using it with spaces.

> No space em dash = a real person who has been bulldozed by LLMs using it with spaces.

Setting an em-dash used for parentheticals closed (with no space)—or sometimes with thin spaces—is the common American literary/academic style (Chicago Manual, APA, and MLA all prefer closed); setting it open—with full word spaces—is the common American practice in journalism (reflected in the AP style guide). Not using em-dash at all for that use, but instead using an en-dash set open is the common British practice.

  • And that's why it should be interpreted as a one way signal. I.e. lack of spaces is a great indication a human wrote it but spaces sround an em dash alone is not a particularly strong indication it was AI.

    Even for Americans, there are several style guidelines/modern preferences (particularly around web content) which don't guarantee the lack of spaces around an em dash. Hence even American LLMs using spaces. Ecen my natural em dash usage always included spaces as an American.

I've had to configure a few editors to stop turning my natural '<space>--<space>' into endashes...

  • I hate any editor changing anything without permission (especially ones that insert the closing paren/bracket).

    I used to love Linux's compose keys, though, where I could just press `<win>--.` to get an en dash and `<win>---` to get an em. Most of the compose combinations were guessable, like `<win>a'` for á, `<win>co` for ©.

    MacOS has a great keyboard locale switcher, but the lack of real compose keys limits things. Most characters you can press and hold and get some accented versions, but it's very slow if you're typing in French or something in the EN layout. It also has a built-in character picker, which is really nice but even more slow.

I used to use em dashes with spaces. I started using them without when I was more into reading style guides and it was—if I recall correctly—the Chicago Manual of Style which doesn’t use spaces. This was way before LLMs came onto the scene as a consumer technology.

  • The Chicago Manual indeed advises against spaces, but it’s written for books. AP Stylebook — that recommends spaces — is for papers. Comments are an even shorter form, so they’re closer to AP than to Chicago.

    When I go to print, I use hairspace, but it’s not worth the trouble in comments.