Public
- Public
- Network
- Groups
- Featured
- Popular
- People

• Writing styles between humans and generative Al models differ significantly: for instance, the GPT models use more nominalizations and have higher sentence complexity (signaling more complex, 'scientific' language), whereas the students make more use of modal and epistemic constructions (which tend to convey speaker attitude).

Download link

• Writing styles between humans and generative Al models differ significantly: for instance, the GPT models use more nominalizations and have higher sentence complexity (signaling more complex, 'scientific' language), whereas the students make more use of modal and epistemic constructions (which tend to convey speaker attitude).
https://sb-9g02l2bbpw.b-cdn.net/media_attachments/files/111/482/599/239/376/933/original/eb10dfecf3a13eac.jpeg

Notices where this attachment appears

Embed this notice
Clive Thompson (clive@saturation.social)'s status on Monday, 27-Nov-2023 22:09:08 JST Clive Thompson

In this paper in Nature, the researchers find that essays generated by ChatGPT were rated higher for quality than human-written ones
Interesting on its face, but …
… what really drew my eye was this finding: human students are more likely to use language that conveys “speaker attitude“
I wonder how much of that is due to ChatGPT’s relatively low temperature (to minimize weird results) and/or prompting?
https://www.nature.com/articles/s41598-023-45644-9
(Via @jasemrau)
1/2

In conversation about a year ago from saturation.social permalink