| A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | AA | AB | AC | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | About: This is a list of ways to describe large language model outputs that do not match reality. I compiled it mostly based on responses to a December 13, 2023 X/Twitter thread (https://x.com/EnglishOER/status/1734429302702452945?s=20). Please feel free to share the link to the spreadsheet at https://bit.ly/HallucinationAlternatives This sheet is licensed CC BY 4.0 by Anna Mills. You are welcome to comment and suggest corrections. You may also copy it, add new terms, and change the scoring criteria and the scores and invite students to do so, but please give attribution if you do. I have tried to list the name of the person who suggested each term. | ||||||||||||||||||||||||||||
2 | Term | Percentage of points possible | Source | Link | Score (0-3) It lacks implications of or associations with intent on the part of the LLM | Score (0-3) It lacks associations with or implications of conscious experience on the part of the LLM | Score (0-3) It implies that outputs are untrue or do not match reality in some way | Score (0-3) It implies that outputs reflect patterns from training data | Score (0-3) It implies that outputs are not just copies of training data but may go beyond it | Score (0-3) It seems accessible without a lot of explanation | Score (0-3) It's catchy and memorable | ||||||||||||||||||
3 | data mirage | 86% | ChatGPT prompted by Anna Mills | https://chat.openai.com/share/a370bce7-8371-4d1c-b8d0-58153c5c0b6f | 3 | 3 | 3 | 2 | 2 | 2 | 3 | ||||||||||||||||||
4 | misprediction | 86% | Lauren Goodlad | https://x.com/CriticalAI/status/1734569970250498530?s=20 | 2 | 3 | 3 | 2 | 3 | 3 | 2 | ||||||||||||||||||
5 | auto-synthesized mirage | 81% | Anna Mills after Nate Angell and ChatGPT | https://x.com/EnglishOER/status/1741323993481925019?s=20 | 3 | 2 | 3 | 3 | 3 | 2 | 1 | ||||||||||||||||||
6 | auto-synthesized misinformation | 81% | Unknown via Anna Mills | https://x.com/EnglishOER/status/1734725770185146427?s=20 | 3 | 3 | 3 | 3 | 2 | 2 | 1 | ||||||||||||||||||
7 | bad prediction | 81% | Lauren Goodlad | https://x.com/CriticalAI/status/1734926368134115533?s=20 | 3 | 3 | 3 | 2 | 1 | 3 | 2 | ||||||||||||||||||
8 | miscalculation | 81% | Lance Eaton | https://x.com/leaton01/status/1734568030540066948?s=20 | 3 | 3 | 3 | 1 | 1 | 3 | 3 | ||||||||||||||||||
9 | misinterpolation | 81% | Anna Mills after Alberto Delgado | https://x.com/EnglishOER/status/1740537239673950280?s=20 | 3 | 3 | 3 | 3 | 3 | 1 | 1 | ||||||||||||||||||
10 | papier mâché made out of training data | 81% | Emily Bender | https://x.com/emilymbender/status/1735673412020809834?s=20 | 1 | 3 | 2 | 3 | 3 | 3 | 2 | ||||||||||||||||||
11 | synthetic misinformation | 81% | Unknown via Anna Mills | https://x.com/EnglishOER/status/1734725770185146427?s=20 | 2 | 3 | 3 | 3 | 2 | 2 | 2 | ||||||||||||||||||
12 | auto-synthesized invalid output | 76% | Anna Mills | https://x.com/EnglishOER/status/1740552494810640804?s=20 | 3 | 3 | 3 | 3 | 2 | 2 | 0 | ||||||||||||||||||
13 | confected nonsense | 76% | Penny Wheeler | https://x.com/pennyjw/status/1740546706281029867?s=20 | 1 | 2 | 3 | 2 | 3 | 3 | 2 | ||||||||||||||||||
14 | fabrication | 76% | Unknown/Karen Weise and Cade Metz | https://www.nytimes.com/2023/05/01/business/ai-chatbots-hallucination.html | 1 | 2 | 3 | 1 | 3 | 3 | 3 | ||||||||||||||||||
15 | hallucination | 76% | Andrej Karpathy? | https://karpathy.github.io/2015/05/21/rnn-effectiveness/ | 3 | 0 | 3 | 1 | 3 | 3 | 3 | ||||||||||||||||||
16 | incorrect synthetic generation | 76% | Alan Levine | https://x.com/cogdog/status/1734693978774376463?s=20 | 3 | 3 | 3 | 3 | 2 | 2 | 0 | ||||||||||||||||||
17 | AI oopsy | 71% | Joseph Robertshaw | https://x.com/jwrobertshaw/status/1740582393973526707?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
18 | aifabrication | 71% | Rosalind Duke | https://x.com/rosalind_duke/status/1734728506632785952?s=20 | 3 | 3 | 3 | 1 | 2 | 1 | 2 | ||||||||||||||||||
19 | artifict | 71% | Rosalind Duke/ChatGPT | https://x.com/rosalind_duke/status/1734729256423317850?s=20 | 3 | 3 | 2 | 2 | 2 | 1 | 2 | ||||||||||||||||||
20 | bad guess | 71% | Maha Bali | https://x.com/Bali_Maha/status/1735160929896059267?s=20 | 2 | 1 | 3 | 1 | 2 | 3 | 3 | ||||||||||||||||||
21 | confabulation | 71% | Milad Khademi Nori, Sungwoo Kim, and Smith, A et al | https://x.com/khademinori/status/1734368583822565530?s=20 https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000388 | 3 | 1 | 3 | 1 | 3 | 2 | 2 | ||||||||||||||||||
22 | fail | 71% | Brian Butler | https://x.com/brian_s_butler/status/1734660374685778140?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
23 | jumbling up | 71% | Sungwoo Kim | https://x.com/sungwookim/status/1734768827110740066?s=20 | 1 | 2 | 3 | 2 | 2 | 2 | 3 | ||||||||||||||||||
24 | made up | 71% | Unknown/Ethan Mollick | https://x.com/emollick/status/1647040774142304257?s=20 | 1 | 2 | 3 | 0 | 3 | 3 | 3 | ||||||||||||||||||
25 | malfunction | 71% | The Bot Brothers | https://x.com/TheBotBrothers/status/1734687636730806497?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
26 | misfire | 71% | Rebecca Fordon | https://x.com/theFordon/status/1734745666079912369?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
27 | misinformation | 71% | ChatGPT prompted by Anna Mills | https://chat.openai.com/share/350444b7-656a-4871-b74a-a546700e0c8d | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
28 | reality-incompatible artifact | 71% | Nate Angell | https://x.com/xolotl/status/1734617274592444766?s=20 | 3 | 3 | 3 | 2 | 2 | 1 | 1 | ||||||||||||||||||
29 | scientific wild-ass guesses | 71% | Unknown via Katherine Yngve | https://x.com/AssessmentG/status/1735357395017064542?s=20 | 2 | 1 | 2 | 2 | 2 | 3 | 3 | ||||||||||||||||||
30 | synthesis | 71% | Helen Beetham | https://x.com/helenbeetham/status/1734862100072112351?s=20 | 3 | 3 | 0 | 3 | 2 | 2 | 2 | ||||||||||||||||||
31 | synthesized misinformation | 71% | Unknown via Anna Mills | https://x.com/EnglishOER/status/1734725770185146427?s=20 | 1 | 3 | 3 | 3 | 2 | 2 | 1 | ||||||||||||||||||
32 | untruth | 71% | Cassandra Nelson | https://x.com/cmnelson71/status/1735853624046018792?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
33 | artifact | 67% | ChatGPT prompted by Anna Mills | https://chat.openai.com/share/350444b7-656a-4871-b74a-a546700e0c8d | 3 | 3 | 0 | 1 | 1 | 3 | 3 | ||||||||||||||||||
34 | bullshit | 67% | Gary Marcus, Mark Corbett Wilson (after Harry Frankfurt) and The Bot Brothers | https://x.com/TheBotBrothers/status/1734701845061267949?s=20 and https://x.com/MCorbettWilson/status/1743134332905210166?s=20 | 2 | 2 | 3 | 1 | 0 | 3 | 3 | ||||||||||||||||||
35 | burp | 67% | Dan Ryan | https://x.com/djjr/status/1740544663076196606?s=20 | 3 | 3 | 2 | 2 | 0 | 1 | 3 | ||||||||||||||||||
36 | defect | 67% | Thomas Basbøll | https://x.com/Inframethod/status/1734599078166573425?s=20 | 3 | 3 | 3 | 0 | 0 | 2 | 3 | ||||||||||||||||||
37 | error | 67% | Rebecca Tomas and Katherine Yngve | https://x.com/RebbeccaTomas/status/1734613279039803578?s=20 and https://x.com/AssessmentG/status/1735043205865799981?s=20 | 2 | 3 | 3 | 0 | 0 | 3 | 3 | ||||||||||||||||||
38 | faulty source synthesis | 67% | Nick Morgan | https://x.com/NickRMorgan/status/1735523604467528023?s=20 | 2 | 3 | 3 | 3 | 1 | 1 | 1 | ||||||||||||||||||
39 | incongruous artifact | 67% | Nate Angell | https://x.com/xolotl/status/1734617274592444766?s=20 | 3 | 3 | 3 | 2 | 2 | 1 | 0 | ||||||||||||||||||
40 | interpolation | 67% | Alberto Delgado | https://x.com/jadelgador/status/1740533031847460936?s=20 | 3 | 3 | 0 | 3 | 3 | 1 | 1 | ||||||||||||||||||
41 | mistake | 67% | The Bot Brothers and Joseph Robertshaw | https://x.com/TheBotBrothers/status/1734609465104818313?s=20 and https://x.com/jwrobertshaw/status/1735074409541468638?s=20 | 3 | 1 | 3 | 0 | 1 | 3 | 3 | ||||||||||||||||||
42 | phantom/phantasma | 67% | Alexander Doria | https://x.com/Dorialexander/status/1734571450865192964?s=20 | 3 | 2 | 3 | 0 | 3 | 0 | 3 | ||||||||||||||||||
43 | unwarranted generative correlation | 67% | Ramón Alvarado | https://x.com/ramonalvaradoq/status/1734810099024372037?s=20 | 3 | 3 | 2 | 3 | 2 | 1 | 0 | ||||||||||||||||||
44 | generation error | 62% | Rebecca Fordon | https://x.com/theFordon/status/1734744048617783797?s=20 | 3 | 3 | 3 | 0 | 0 | 2 | 2 | ||||||||||||||||||
45 | information-shaped sentences | 62% | Neil Gaiman | https://x.com/neilhimself/status/1639610373115375616?s=20 | 3 | 3 | 2 | 1 | 1 | 1 | 2 | ||||||||||||||||||
46 | invalid autogenerated output | 62% | Anna Mills | https://x.com/EnglishOER/status/1735147682027749645?s=20 | 3 | 3 | 3 | 0 | 1 | 2 | 1 | ||||||||||||||||||
47 | invalid output | 62% | Anna Mills | https://x.com/EnglishOER/status/1740552494810640804?s=20 | 3 | 3 | 3 | 0 | 0 | 3 | 1 | ||||||||||||||||||
48 | plausible untruth | 62% | @Jisc National Centre for AI in Tertiary Education via Mary Jacob | https://x.com/MaryJacobTEL1/status/1734556326225740142?s=20 | 3 | 3 | 3 | 1 | 0 | 2 | 1 | ||||||||||||||||||
49 | prediction anomalies | 62% | Stevco Le Roc | https://x.com/shahaoul/status/1740542190630383786?s=20 | 3 | 3 | 2 | 2 | 2 | 0 | 1 | ||||||||||||||||||
50 | concoction | 57% | ChatGPT prompted by Anna Mills | https://chat.openai.com/share/350444b7-656a-4871-b74a-a546700e0c8d | 0 | 1 | 2 | 1 | 2 | 3 | 3 | ||||||||||||||||||
51 | hairball | 57% | Matt Parker | https://x.com/DrMattParker/status/1734718386779631616?s=20 | 3 | 3 | 1 | 2 | 0 | 0 | 3 | ||||||||||||||||||
52 | inaccuracy | 57% | Mark Warschauer | https://x.com/markwarschauer/status/1740544038783393954?s=20 | 3 | 3 | 3 | 0 | 0 | 1 | 2 | ||||||||||||||||||
53 | non-factual outputs | 57% | Melissa McCradden | https://x.com/MMccradden/status/1734717975804940421?s=20 | 3 | 3 | 3 | 0 | 0 | 2 | 1 | ||||||||||||||||||
54 | plausible erroneous output | 57% | Tim Reierson/Anna Mills | https://x.com/EnglishOER/status/1735084963702428053?s=20 | 3 | 3 | 3 | 1 | 0 | 1 | 1 | ||||||||||||||||||
55 | all signifier, no signified | 52% | Anna Mills (after Maha Bali) | https://x.com/EnglishOER/status/1735089732986736748?s=20 | 3 | 3 | 2 | 0 | 0 | 1 | 2 | ||||||||||||||||||
56 | apophenia | 52% | Roberto Leon | https://x.com/rleonboone1/status/1734660742559813962?s=20 | 2 | 0 | 3 | 2 | 3 | 0 | 1 | ||||||||||||||||||
57 | concatenation | 52% | Steven Kapica | https://x.com/sskapica/status/1734674961560715269?s=20 | 3 | 3 | 0 | 1 | 2 | 1 | 1 | ||||||||||||||||||
58 | congolomerate | 52% | Roberto Leon | https://x.com/rleonboone1/status/1734660742559813962?s=20 | 3 | 3 | 0 | 3 | 0 | 1 | 1 | ||||||||||||||||||
59 | debris | 52% | Annaliese Hoehling | https://x.com/Evenannaliese/status/1734560104400253404?s=20 | 3 | 3 | 2 | 0 | 0 | 1 | 2 | ||||||||||||||||||
60 | digital scrap | 52% | Eric Martinsen | https://x.com/EricMartinsen/status/1734748579892498893?s=20 | 3 | 3 | 1 | 1 | 0 | 1 | 2 | ||||||||||||||||||
61 | falsehood | 52% | Info General/@OccamsOps | https://x.com/OccamsOps/status/1740562055969882211?s=20 | 1 | 2 | 3 | 0 | 0 | 3 | 2 | ||||||||||||||||||
62 | fantasma | 52% | Alexander Doria | https://x.com/Dorialexander/status/1734629655808807258?s=20 | 2 | 1 | 3 | 0 | 2 | 1 | 2 | ||||||||||||||||||
63 | generation | 52% | Balázs Kégl | https://x.com/balazskegl/status/1734552898019745820?s=20 | 3 | 3 | 0 | 0 | 0 | 3 | 2 | ||||||||||||||||||
64 | jibbering | 52% | Helen Morley | https://x.com/HelenBrite/status/1734904540741194009?s=20 | 0 | 1 | 3 | 1 | 1 | 3 | 2 | ||||||||||||||||||
65 | lie | 52% | Daniel VIllalba | https://x.com/danvillalba/status/1734674803708117415?s=20 | 0 | 1 | 3 | 0 | 1 | 3 | 3 | ||||||||||||||||||
66 | phantom islands on ancient maps | 52% | Alexander Doria | https://x.com/Dorialexander/status/1734571450865192964?s=20 | 1 | 1 | 3 | 1 | 2 | 1 | 2 | ||||||||||||||||||
67 | scrap output | 52% | Eric Martinsen | https://x.com/EricMartinsen/status/1734748579892498893?s=20 | 3 | 3 | 1 | 1 | 0 | 1 | 2 | ||||||||||||||||||
68 | semantic error | 52% | The Bot Brothers | https://x.com/TheBotBrothers/status/1734687636730806497?s=20 | 3 | 3 | 3 | 0 | 0 | 1 | 1 | ||||||||||||||||||
69 | all text, no subtext | 48% | Dominik Lukeš | https://x.com/techczech/status/1734737458825159045?s=20 | 3 | 3 | 1 | 0 | 0 | 1 | 2 | ||||||||||||||||||
70 | erroneous output with potential to cause user harm or confusion | 48% | Tim Reierson | https://twitter.com/holdspacefree/status/1735084224468918274 | 3 | 3 | 3 | 0 | 0 | 1 | 0 | ||||||||||||||||||
71 | fantasy | 48% | Alexander Doria | https://x.com/Dorialexander/status/1734629655808807258?s=20 | 1 | 1 | 3 | 0 | 2 | 1 | 2 | ||||||||||||||||||
72 | formulation | 48% | Nick Morgan | https://x.com/NickRMorgan/status/1735523104846164132?s=20 | 3 | 3 | 1 | 1 | 0 | 1 | 1 | ||||||||||||||||||
73 | reconstruction artifacts | 48% | Nicholas Bailey | https://x.com/AliasNickelby/status/1734609749415788769?s=20 | 3 | 3 | 0 | 1 | 2 | 1 | 0 | ||||||||||||||||||
74 | confident falsehood | 43% | Katie Conrad | https://x.com/KatieConradKS/status/1734743108317741518?s=20 | 1 | 0 | 3 | 0 | 0 | 3 | 2 | ||||||||||||||||||
75 | content validity anomalies | 43% | Joseph Robertshaw | https://x.com/jwrobertshaw/status/1735074409541468638?s=20 | 3 | 3 | 2 | 0 | 1 | 0 | 0 | ||||||||||||||||||
76 | statistically likely utterances | 43% | Edward R. O'Neill | https://x.com/learningtech/status/1734548378409660663?s=20 | 2 | 2 | 0 | 3 | 0 | 1 | 1 | ||||||||||||||||||
77 | windows into alternate realities in the latent space | 43% | Emad Mostaque | https://x.com/EMostaque/status/1632902759690240001?s=20 | 2 | 3 | 1 | 0 | 0 | 1 | 2 | ||||||||||||||||||
78 | ageneration | 38% | Martin Compton | https://x.com/mart_compton/status/1734864286059532384?s=20 | 3 | 3 | 1 | 0 | 0 | 0 | 1 | ||||||||||||||||||
79 | gibbering | 38% | Tony Hirst | https://x.com/psychemedia/status/1734610713652715545?s=20 | 1 | 0 | 3 | 0 | 1 | 0 | 3 | ||||||||||||||||||
80 | agenerative output | 33% | Martin Compton | https://x.com/mart_compton/status/1734864286059532384?s=20 | 3 | 3 | 1 | 0 | 0 | 0 | 0 | ||||||||||||||||||
81 | irrational utterances | 33% | Joseph Robertshaw | https://x.com/jwrobertshaw/status/1735074409541468638?s=20 | 1 | 1 | 3 | 0 | 0 | 1 | 1 | ||||||||||||||||||
82 | AI-splaining | 29% | Anna Mills after Mike Czaplicki | https://x.com/EnglishOER/status/1740560803240251639?s=20 | 0 | 0 | 2 | 0 | 0 | 1 | 3 | ||||||||||||||||||
83 | composition conjurations | 29% | Joseph Robertshaw | https://x.com/jwrobertshaw/status/1735074409541468638?s=20 | 0 | 0 | 1 | 0 | 3 | 0 | 2 | ||||||||||||||||||
84 | Llmsplaining | 29% | Mike Czaplicki | https://x.com/flughafencza/status/1740558616439234929?s=20 | 0 | 0 | 2 | 0 | 0 | 1 | 3 | ||||||||||||||||||
85 | yam-splaining | 24% | Mike Czaplicki | https://x.com/flughafencza/status/1740558616439234929?s=20 | 0 | 0 | 2 | 0 | 0 | 0 | 3 | ||||||||||||||||||
86 | AI mirage | 71% | Anna Mills and Nate Angell | 3 | 3 | 3 | 0 | 0 | 3 | 3 | |||||||||||||||||||
87 | |||||||||||||||||||||||||||||
88 | |||||||||||||||||||||||||||||
89 | |||||||||||||||||||||||||||||
90 | |||||||||||||||||||||||||||||
91 | |||||||||||||||||||||||||||||
92 | |||||||||||||||||||||||||||||
93 | |||||||||||||||||||||||||||||
94 | |||||||||||||||||||||||||||||
95 | |||||||||||||||||||||||||||||
96 | |||||||||||||||||||||||||||||
97 | |||||||||||||||||||||||||||||
98 | |||||||||||||||||||||||||||||
99 | |||||||||||||||||||||||||||||
100 | |||||||||||||||||||||||||||||