In Defence of the Em Dash

The em dash (—) has recently been singled-out as a sure tell of AI-generated writing. Popular LLMs — and in particular, ChatGPT — are certainly fond of the em dash. But the level of scorn for those who use — and, for the mark itself, has reached a level where I feel compelled to write in defence of a punctuation mark.

Why Do I Really Care?

The association between AI-generated writing and the em dash has become so strong that I now feel reluctant to use one of my favourite — and most versatile — bits of punctuation, lest my writing be dismissed as something I did not generate. This post has three purposes:

1. to explain how I use dashes;
2. to pre-emptively defend myself against accusations of AI-generated text for using em-dashes; and
3. to share information about how to type em- and en-dashes, and other special characters using compose key.

I have used em-dashes extensively in my other posts on this website, some of which pre-date the first release of ChatGPT. The texts for the posts on my site have not been so much as edited by AI. If you ever receive an email from me with em-dashes, do not take it as evidence of AI-assisted writing.

Hyphens, En Dashes, and Em Dashes

I take a relaxed approach to punctuation. I loosely choose to stick to a set of (fairly common) rules, without feeling the urge to prescribe these choices to others. Roughly I use:

• Hyphens (-) for
  • various types of compound word formation (fast-approaching, top-of-the-range, sometimes co-operative)
• En dashes (–) for
  • specifying numerical ranges (1920–1929, pp. 107–112)
  • connecting items (Leeds–Liverpool canal)
• Em dashes (—) for
  • parenthetical remarks (personally — though somewhat controversially — I surround the dashes with spaces when using them parenthetically)
  • long pauses — such as this

I find these choices typographically pleasing.

As to why I enjoy using the em dash so much — that essentially comes down to its versatility. In many cases, it really does just feel like the most appropriate punctuation for when a mental (or verbal) pause is needed. The line itself evokes a sense of deliberate, considered silence within the horizontal flow of the text.

Writing Em Dashes

A common justification for suspicion about the artificial provenance of text with em dashes in is that nobody’s keyboard has an em dash key. Ignoring the fact that many text editors will automatically convert two or three dashes into an em dash, it is indeed possible to type em (or en) dashes. If you are unable to type an em dash, you should consider upgrading your keyboard. Fortunately, this does not require spending any money as this can be remedied with a bit of software.

The most basic way to include symbols not present on your keyboard is to search from them online, and copy and paste the result into your document. Fine if you have internet connection and only need to do this rarely. Otherwise, it’s suboptimal.

Remember Alt codes? If you’d like to type “Tromsø”, you simply need to rake the depths of your memory to recall that the corresponding code is Alt-0248. Of course unless you’re a zoomer who refuses to use capital letters, you’ll simply need to recall that Ø is Alt-0216 when writing Øresund. Easy, right? Memorising these is feasible if you only need to use a few of them (for someone who occasionally needs to write Norwegian place names, æ, Æ, ø, Ø, å, and Å will do if a Norwegian keyboard is not available). For anyone else — anyone who occasionally wants to type “20°C”, “πr²”, “Thing™”, “3½”, “a ∈ A”, or “{1, 2, …, N}” — there is simply too much to memorise.

Many operating systems have some sort of generic “character picker” functionality that allows you to search for special characters including emojis. These usually come with a search option, but this can add a bit of friction to typing special characters.

A better system, would be to use intuitive mnemonics rather than codes. For example, if we could represent æ by ae, get á by combining a', or â by typing a^, ã by a~, or even ā bya-. Fortunately, such a system exists, and is available on many computers. Of course, there are not enough two-character combinations to cover all the possible unicode symbols with an intuitive digraph (two characters), leave some characters unrepresented or with unintuitive digraphs.

On Linux one can use ComposeKey to compose key sequences. For example, upon pressing AltGr, instead of entering a character immediately, you will enter composition mode. This gives you the opportunity to type the symbol you want. So pressing AltGr followed by -> yields →. To get an em dash, simply press AltGr and then type ---. An en dash is just --.. (Bonus for vim users: take a look at digraphs)

On MacOS, ComposeKey is a little more tricky. Macs, having an extra function key compared to many keyboards, have more shortcuts for special characters. These include writing the en dash with ⌘-, or the em dash with ⌘⇧-. Going beyond these basic special characters, one can use the emoji picker, as previously mentioned. MacOS also has a similar ‘hold-for-diacritics’ to mobile phone keyboards. If those options are not sufficient, there is a way to get ComposeKey-like behaviour using Karabiner Elements. See this blog post for details.

On Windows, there is WinCompose.

On Mobile it is usually possible to type em and en dashes by holding down the dash key until the additional variants pop up.