<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i" -->

---
title: How to talk about "AI" without adding to the...
description: Emily M. Bender and Nanna Inie propose concrete rephrasings to strip anthropomorphizing language out of how people talk about AI systems, building on their...
canonical: https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: How to talk about "AI" without adding to the anthropomorphization | daily.dev
og:description: Emily M. Bender and Nanna Inie propose concrete rephrasings to strip anthropomorphizing language out of how people talk about AI systems, building on their...
og:url: https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i
og:image: https://api.daily.dev/og/posts/vggaWw36i.png
og:image:alt: How to talk about "AI" without adding to the anthropomorphization
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# How to talk about "AI" without adding to the anthropomorphization

**[Lobsters](https://daily.dev/sources/lobsters)** · 8 min read · 2 upvotes · 1 comments

## Summary

Emily M. Bender and Nanna Inie propose concrete rephrasings to strip anthropomorphizing language out of how people talk about AI systems, building on their Tech Policy Press op-ed and academic research (Inie et al 2026). They organize suggestions into categories - cognizer/products of cognition, emotion, communication, agency, human role analogy, names/pronouns, and biological metaphors - offering alternatives like 'probabilistic automation' for AI, 'undesirable output' for hallucination, and 'probabilistic, unverified software manipulator' for AI agent. The piece argues that shifting vocabulary to emphasize functionality and human agency, rather than machine cognition or emotion, clarifies what these systems actually do.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://buttondown.com/maiht3k/archive/how-to-talk-about-ai-without-adding-to-the>

## Questions this post answers

### What is a good non-anthropomorphizing alternative term for 'AI hallucination'?

A recommended replacement is 'undesirable output.' Hallucination originally means perceiving things that are not there, but software systems do not perceive anything; all large language model output is probabilistically produced synthetic text, and there is no fundamental difference between desirable and undesirable outputs on the system side, only for the people interpreting them.

_Developers debating how to describe model failures can find similar terminology discussions curated on daily.dev._

### What alternative term is suggested instead of 'AI agent' or 'agentic AI'?

The suggested alternative is 'probabilistic, unverified software manipulator,' abbreviated PUSM, chosen partly because the acronym sounds deliberately unappealing. The term describes software connecting large language models or other components to systems that affect the world, such as scheduling appointments or booking flights, work previously done by people or purpose-built software.

_Anyone weighing how to name emerging agentic tooling can track these framing debates on daily.dev._

## Community take

How the wider developer community reacted, aggregated from 4 discussions and 15 comments across hackernews, lobsters (as of 2026-09-23).

**TL;DR:** Commenters largely agree that anthropomorphizing language obscures how AI systems actually work, though several find the proposed replacement terms clunky, and there's real debate over whether the linguistic reform matters much in practice.

**Sentiment:** 40% positive · 45% mixed · 15% skeptical

**The case for**

- Agreeing that phrases like 'the AI thinks' are analogies that normalize misleading claims about machine cognition.
- Some argue precise language matters because it prevents non-technical audiences from believing AI acts autonomously, which deflects blame from the people deploying it.
- One commenter shares their own practice of substituting 'generate' for 'write' and 'indicates' for 'explains' to stay consistent.

**The pushback**

- Several proposed replacements (like 'undesirable output' for hallucination) are seen as far less catchy or usable than the terms they'd replace.
- Some doubt the effort is worthwhile, noting humans anthropomorphize everything from roombas to lab equipment as a matter of course.
- One commenter prefers 'confabulation' over 'hallucination' and argues 'LLM' is more precise than 'AI' as a broader category term.

**By community**

- hackernews (mixed): Sparse but favorable commentary, with one detailed comment supporting the argument against normalized anthropomorphizing language.
- lobsters (mixed): Broad agreement with the article's goal but genuine debate over whether the specific replacement terms are usable and whether the stakes are as high as claimed.

**Hottest debate:** Whether avoiding anthropomorphizing language actually changes public perception/accountability or is just an unworkable stylistic preference.

**Open questions**

- Is there a good, natural-sounding replacement term for 'reasoning' that avoids implying cognition?
- Does using clinical alternative terminology actually shift how non-technical people perceive AI capabilities?

**Highlights**

> Since these technologies are used for stuff like committing genocides, mass displacement of workers, surveillance by authoritarian governments and other evil purposes, the stake of using such language is deflecting the attention from the perpetrators (tech oligarchs and governments) who want you to believe the machines are autonomous and unstoppable. It's definitely worth it.
> — [chobeat on lobsters · 3 points, 1 comments](https://lobste.rs/s/oqipmz/how_talk_about_ai_without_adding#c_n44jsh)

> Oh, interesting find! I've thought about this a lot, but I still learned a few things from this article. I have heard good arguments in favour of anthropomorphisation, but I still want to be a little careful in my own writing. TFA misses a very common case: the verb "to write". In my head, an LLM cannot "write" anything, but it can "generate" text. I try to be consistent in using "write" when referring to humans, and "generate" when referring to language models. One word I've struggled with a bit in my writing, which is also not mentioned in TFA, is what people commonly refer to as "reasoning". I settled on calling the token stream that appears to work through a problem as "the reasoning tokens", which means language models do not "reason", but they can "emit reasoning tokens". This is a bit of a cop-out though, because it still uses that word – only as a noun with an alternative definition, rather than a verb. I also notice I've avoided saying that a language model "explains" something. Instead I've written that "a rationale is provided in the output". This indicates how awkward it can get to avoid anthropomorphisation, though. Then there are edge cases like saying a language model "has no sense of X" which is true, but also anthropomorphises in that it affirms an expectation that they could have a sense of something. I'm also nowhere near good at this. In my writing I find plenty of examples where the model "gets confused" or "doubles down" or "decides".
> — [kqr on lobsters · 1 points, 2 comments](https://lobste.rs/s/oqipmz/how_talk_about_ai_without_adding#c_wwou4u)

> Strong agree with the ideas. Have commented here frequently about a "boiling the frog slowly" trend of using Anthropomorphisms to soft pass the idea of AGI. Neologisms are awful, but so are analogies. if I write "I think" its not an analogy, it's a literal truth. I just thunked it. If I write "the AI thinks" then it begs questions. It isn't literal truth, its analogous, its a linguistic normalisation of describing what it did. Did it think? Who knows! For the LISP amongst us, "thunk" goes all the way back to 1961 and ALGOL 60. And it was a neologistic analogy even then.
> — [ggm on hackernews](https://news.ycombinator.com/item?id=49794432)

> I don't know if avoiding using anthropomorphizing language is worth it. As a species, we're *great* at anthropomorphizing everything from roombas to computers to hulking lab equipment. Besides, some terms here are fairly uncontroversial or even ancient in computing terms - "speech recognition" was coined in the 60s.
> — [DustyFuzzy on lobsters · 1 points, 1 comments](https://lobste.rs/s/oqipmz/how_talk_about_ai_without_adding#c_zz5wph)

**Source threads**

- [hackernews](https://news.ycombinator.com/item?id=48741809) · 3 points · 0 comments
- [hackernews](https://news.ycombinator.com/item?id=49794261) · 2 points · 2 comments
- [lobsters](https://lobste.rs/s/oqipmz/how_talk_about_ai_without_adding) · 24 points · 11 comments
- [hackernews](https://news.ycombinator.com/item?id=48822019) · 2 points · 2 comments

## Community discussion

Top comments from developers on daily.dev.

**@kartiknvj** · 0 upvotes

> Bender and Inie's rephrasing list is useful, and the hardest entries for me to adopt are the ones that remove agency from failure descriptions. When I write an incident postmortem, saying a model 'refused' instead of 'declined' sounds trivial until the wording changes who gets blamed. Have you tried applying this to eval reports, where 'the model lied' versus 'the output diverged from ground truth' steers the fix?

## Similar posts on daily.dev

- [How To Talk To A Machine Without Anthropomorphising It](https://daily.dev/posts/how-to-talk-to-a-machine-without-anthropomorphising-it-1spfingcn) · Hackaday · 0 upvotes · 0 comments
- [We Need to Talk About How AI Actually Makes You Feel](https://daily.dev/posts/we-need-to-talk-about-how-ai-actually-makes-you-feel-c76dzwfgx) · SQL Authority · 0 upvotes · 0 comments
- [AI Myths That People Still Believe](https://daily.dev/posts/ai-myths-that-people-still-believe-3zuuqphee) · Medium · 1 upvotes · 1 comments
- [Three Inverse Laws of Robotics](https://daily.dev/posts/three-inverse-laws-of-robotics-v3ajoy7pt) · Lobsters · 3 upvotes · 0 comments
- [The Future of AI](https://daily.dev/posts/the-future-of-ai-i7ojr8lxm) · Hacker News · 2 upvotes · 0 comments

---

Tags: [#llm](https://daily.dev/tags/llm), [#bots](https://daily.dev/tags/bots), [#ethical-ai](https://daily.dev/tags/ethical-ai)

[View this post on daily.dev](https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"How to talk about \"AI\" without adding to the anthropomorphization","url":"https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i"},"datePublished":"2026-09-23T13:40:01.844Z","dateModified":"2026-09-23T21:25:59.545Z","description":"Emily M. Bender and Nanna Inie propose concrete rephrasings to strip anthropomorphizing language out of how people talk about AI systems, building on their...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bf2b3d9c346d53a93371390a4abf00da?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/bf2b3d9c346d53a93371390a4abf00da?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Lobsters","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Lobsters","logo":"https://media.daily.dev/image/upload/s--tl8v_Fku--/f_auto,t_logo/v1698841318/logos/lobste.jpg","url":"https://daily.dev/sources/lobsters"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"llm,bots,ethical-ai","timeRequired":"PT8M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Lobsters","item":"https://daily.dev/sources/lobsters"},{"@type":"ListItem","position":3,"name":"How to talk about \"AI\" without adding to the anthropomorphization"}]}
{"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i","comment":[{"@type":"Comment","text":"Bender and Inie’s rephrasing list is useful, and the hardest entries for me to adopt are the ones that remove agency from failure descriptions. When I write an incident postmortem, saying a model ‘refused’ instead of ‘declined’ sounds trivial until the wording changes who gets blamed. Have you tried applying this to eval reports, where ‘the model lied’ versus ‘the output diverged from ground truth’ steers the fix?","datePublished":"2026-09-24T20:00:41.560Z","url":"https://daily.dev/posts/vggaWw36i#c-GOXOnYszt","author":{"@type":"Person","name":"kartik-nvjk","url":"https://daily.dev/kartiknvj","image":"https://media.daily.dev/image/upload/s--3gGgsVCw--/f_auto/v1781456774/avatars/avatar_TvTVeiMdkRCqWUDullFmy?_a=BAMAMiWQ0"}}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/how-to-talk-about-ai-without-adding-to-the-anthropomorphization-vggaww36i#faq","mainEntity":[{"@type":"Question","name":"What is a good non-anthropomorphizing alternative term for 'AI hallucination'?","acceptedAnswer":{"@type":"Answer","text":"A recommended replacement is 'undesirable output.' Hallucination originally means perceiving things that are not there, but software systems do not perceive anything; all large language model output is probabilistically produced synthetic text, and there is no fundamental difference between desirable and undesirable outputs on the system side, only for the people interpreting them. Developers debating how to describe model failures can find similar terminology discussions curated on daily.dev."}},{"@type":"Question","name":"What alternative term is suggested instead of 'AI agent' or 'agentic AI'?","acceptedAnswer":{"@type":"Answer","text":"The suggested alternative is 'probabilistic, unverified software manipulator,' abbreviated PUSM, chosen partly because the acronym sounds deliberately unappealing. The term describes software connecting large language models or other components to systems that affect the world, such as scheduling appointments or booking flights, work previously done by people or purpose-built software. Anyone weighing how to name emerging agentic tooling can track these framing debates on daily.dev."}}]}
```

