←back to thread

Show HN: Hacker News em dash user leaderboard pre-ChatGPT

(www.gally.net)
358 points tkgally | 9 comments | 30 Aug 25 03:40 UTC | HN request time: 0.808s | source | bottom

The use of the em dash (—) now raises suspicions that a text might have been AI-generated. Inspired by a suggestion from dang [1], I created a leaderboard of HN users according to how many of their posts before November 30, 2022—that is, before the release of ChatGPT—contained em dashes. Dang himself comes in number 2—by a very slim margin.

Credit to Claude Code for showing me how to search the HN database through Google BigQuery and for writing the HTML for the leaderboard.

[1] https://news.ycombinator.com/item?id=45053933

1. IAmGraydon ◴[30 Aug 25 04:25 UTC] No.45071916[source]▶
I guess I’m confused. Why is it interesting to know how many em dashes were used before the dawn of ChatGPT? It’s how many AFTER that seems like it would be far more interesting.
replies(4): >>45071977 #>>45071990 #>>45071991 #>>45072503 #
2. southwindcg ◴[30 Aug 25 04:47 UTC] No.45071977[source]▶
Some people accuse anyone who uses em dashes of using ChatGPT to write their posts. This is "proof" that actual humans use em dashes.
replies(1): >>45072690 #
3. tkgally ◴[30 Aug 25 04:53 UTC] No.45071990[source]▶
As mentioned in the thread that included dang’s suggestion [1], examples of one’s use of em dashes timestamped before ChatGPT could be used as a defense if one is accused, on the basis of em dashes, of having written with AI.

Whether this is interesting or not, well…

[1] https://news.ycombinator.com/item?id=45046883

4. latexr ◴[30 Aug 25 04:53 UTC] No.45071991[source]▶
Because it’s becoming a common belief that any em-dash indicates LLM writing, and us people who regularly use em-dashes are attempting to show that is a poor signal on its own. The goal is to show proof of humans using it.
replies(1): >>45072083 #
5. Tostino ◴[30 Aug 25 05:22 UTC] No.45072083[source]▶
Or at least to have a baseline. If you see a sudden jump, that does tell you something.
replies(1): >>45072436 #
6. bee_rider ◴[30 Aug 25 06:46 UTC] No.45072436{3}[source]▶
Maybe it tells us that, thanks to AI, some folks learned about a perfectly useful piece of punctuation.
7. dragonwriter ◴[30 Aug 25 06:59 UTC] No.45072503[source]▶
Given that GPT-3.5 (like many LLMs) was trained with a large corpus of scraped internet data, including popular discussion fora, the people on the leaderboard are the ones potentially to blame for ChatGPT’s em-dash habit.
8. vntok ◴[30 Aug 25 07:37 UTC] No.45072690[source]▶
Things like books are proof that actual humans use em dashes, that wasn't ever the contention.

What's needed is a writing comparison before/after 2022 for these users. If there's a sudden 200% increase in the use of em-dashes from one month to the next, it's a very strong indicator that the user started LLMing their posts.

replies(1): >>45078852 #
9. southwindcg ◴[30 Aug 25 23:15 UTC] No.45078852{3}[source]▶
Perhaps I should have qualified that humans use them in casual writing, website comments and the like, and not just in formal, published works that probably had an editor.