pith. sign in

arxiv: 2606.01929 · v2 · pith:YJZBIJGEnew · submitted 2026-06-01 · 💻 cs.AI

VET: A Framework for Analyzing AI Discourse

Meredith Ringel Morris This is my paper

Pith reviewed 2026-06-28 14:39 UTC · model grok-4.3

classification 💻 cs.AI
keywords AI discourseVET frameworkAI hypeAI doomAI denialAI normalcyAI literacypolarized narratives
0
0 comments X

The pith

The VET framework categorizes AI discourse by valence, effectiveness, and trajectory to show exaggerations in four common stances.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents the VET framework as a method for sorting AI discourse along three dimensions: valence, effectiveness, and trajectory. It demonstrates the approach by mapping four prevalent positions—AI Hype, AI Doom, AI Denial, and AI Normalcy—and identifies where each one overstates current capabilities or future paths. A reader would care because polarized claims in media can distort public understanding of AI. If the dimensions work as described, the framework offers a practical way to examine and compare such claims for greater clarity.

Core claim

The VET framework categorizes AI discourse along the dimensions of valence, effectiveness, and trajectory. Using this structure reveals that the stances of AI Hype, AI Doom, AI Denial, and AI Normalcy each exaggerate aspects of AI's current state and likely evolution. The framework thereby functions as an AI literacy tool that supports the vetting of polarized discourse.

What carries the argument

The VET framework, which classifies discourse using the three dimensions of valence (positive or negative tone), effectiveness (views on AI capabilities), and trajectory (expectations for future development).

If this is right

  • The four stances can be directly compared and critiqued using the same three dimensions.
  • VET supports systematic identification of exaggerations in common AI narratives.
  • The framework provides a concrete method for improving public analysis of AI claims.
  • AI literacy efforts can incorporate VET to address polarization in media and discussion.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same three dimensions could be tested on discourse about other technologies such as biotechnology or climate modeling.
  • Classroom trials could check whether exposure to VET changes how people evaluate new AI announcements.
  • If new discourse patterns emerge that resist the current categories, the framework might need an added dimension.

Load-bearing premise

The three dimensions of valence, effectiveness, and trajectory are sufficient and non-overlapping to capture the key exaggerations in prevalent AI discourse.

What would settle it

A set of real-world AI discourse samples that cannot be placed into the four stances without forcing overlaps or leaving major elements unaccounted for by the three dimensions.

Figures

Figures reproduced from arXiv: 2606.01929 by Meredith Ringel Morris.

Figure 1
Figure 1. Figure 1: The VET Framework: An illustration of how the proposed analytic framework of valence, effectiveness, and trajectory helps to identify and compare four prevalent narratives in U.S. discourse about AI. Note that, per the framework, AI Normalcy is less strongly polarized with respect to valence and effectiveness than the other three narratives. Because effectiveness and trajectory are correlated in many narra… view at source ↗
read the original abstract

Public discourse on AI has become polarized; exaggerated positions on AI in traditional and social media threaten the development of AI Literacy among the general public. In this article, I introduce the VET Framework, a method for categorizing AI discourse along the dimensions of valence, effectiveness, and trajectory. I show how this framework can be used to identify, compare, and critique prevalent narratives of AI Hype, AI Doom, AI Denial, and AI Normalcy. Using VET, I analyze how each of these four stances exaggerates some aspects of the current state and/or likely evolution of AI, and illustrate how the VET framework can serve as an AI Literacy tool by supporting the ``vetting'' of polarized AI discourse.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper proposes the VET framework, which categorizes AI discourse along three dimensions—valence, effectiveness, and trajectory—to identify and critique four prevalent stances (Hype, Doom, Denial, and Normalcy). It analyzes how each stance exaggerates aspects of AI's current state or future evolution and positions VET as an AI literacy tool for vetting polarized public discourse.

Significance. If the framework's dimensions prove robust, it could offer a structured approach to dissecting exaggerated AI narratives and supporting more balanced public understanding. The proposal is conceptually clear and free of circular reasoning, but its significance is constrained by the absence of any systematic justification, mapping, or validation for the chosen dimensions.

major comments (2)
  1. [Section introducing the VET framework] Section introducing the VET framework: The central claim that valence, effectiveness, and trajectory suffice to identify and critique exaggerations across the four stances rests on the unargued axiom that these dimensions are meaningful, sufficient, and non-overlapping. No mapping, independence check, or argument rules out covariance (e.g., between valence and trajectory) or shows exhaustiveness relative to other possible axes such as evidence quality.
  2. [Application to the four stances] Application to the four stances: The analyses of Hype, Doom, Denial, and Normalcy are presented as illustrations of exaggeration, but no concrete examples, case studies, or inter-annotator checks are supplied to demonstrate that the framework reliably surfaces these exaggerations rather than merely labeling them.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments on our manuscript. We address each major comment below and indicate where revisions will be made to strengthen the paper.

read point-by-point responses

  1. Referee: [Section introducing the VET framework] Section introducing the VET framework: The central claim that valence, effectiveness, and trajectory suffice to identify and critique exaggerations across the four stances rests on the unargued axiom that these dimensions are meaningful, sufficient, and non-overlapping. No mapping, independence check, or argument rules out covariance (e.g., between valence and trajectory) or shows exhaustiveness relative to other possible axes such as evidence quality.

    Authors: We agree that the manuscript would be strengthened by a more explicit justification for selecting these three dimensions. The dimensions were chosen because they directly map onto recurring structural features of polarized AI narratives (emotional tone, capability claims, and temporal projections). We will revise the introduction section to include a dedicated paragraph that (a) motivates the choice with reference to prior discourse analysis literature, (b) discusses potential covariances such as between valence and trajectory, and (c) clarifies that evidence quality is treated as orthogonal because the framework targets narrative stance rather than factual verification. This addition will make the conceptual grounding more transparent. revision: yes

  2. Referee: [Application to the four stances] Application to the four stances: The analyses of Hype, Doom, Denial, and Normalcy are presented as illustrations of exaggeration, but no concrete examples, case studies, or inter-annotator checks are supplied to demonstrate that the framework reliably surfaces these exaggerations rather than merely labeling them.

    Authors: The current text presents the four stances as archetypal illustrations to show how VET can be applied. We accept that adding concrete, sourced examples from public discourse would make the demonstration more rigorous and allow readers to evaluate the framework's utility directly. We will add a new subsection containing 2–3 brief case studies with references to specific statements from media or public figures. Inter-annotator reliability checks are not applicable here, as the framework is offered as a conceptual analytic tool rather than a formal annotation scheme; the added examples will instead serve to illustrate consistent application. revision: partial

Circularity Check

0 steps flagged

No circularity; framework is independently proposed and applied illustratively

full rationale

The paper introduces the VET framework (valence, effectiveness, trajectory) as an original categorization method for AI discourse stances. It defines the dimensions, maps them to four stances (Hype, Doom, Denial, Normalcy), and demonstrates application for critique and AI literacy. No equations, parameter fitting, self-citations, or derivations are present that would reduce the framework or its claims to its own inputs by construction. The dimensions are posited rather than derived, and the analysis is presented as an application of the proposed tool, not a self-referential prediction. This matches the default expectation of a non-circular framework paper.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the utility of the proposed VET dimensions for discourse analysis, which are introduced within the paper without external empirical grounding mentioned.

axioms (2)
  • domain assumption Public discourse on AI has become polarized with exaggerated positions that threaten AI literacy.
    Opening premise of the abstract used to motivate the framework.
  • ad hoc to paper Valence, effectiveness, and trajectory are meaningful and sufficient dimensions for categorizing AI discourse.
    Core definitional assumption of the VET framework stated in the abstract.
invented entities (1)
  • VET Framework no independent evidence
    purpose: To categorize AI discourse and support vetting of polarized narratives
    Newly introduced analytical tool in this paper.

pith-pipeline@v0.9.1-grok · 5632 in / 1310 out tokens · 29980 ms · 2026-06-28T14:39:24.076278+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

211 extracted references · 21 canonical work pages

  1. [1]

    2026 , address =

    Mallaby, Sebastian , title =. 2026 , address =

    2026

  2. [2]

    2023 , month =

    Roose, Kevin , title =. 2023 , month =

    2023

  3. [3]

    Nature Human Behaviour , volume =

    Gilardi, Fabrizio and Kasirzadeh, Atoosa and Bernstein, Abraham and Staab, Steffen and Gohdes, Anita , title =. Nature Human Behaviour , volume =. 2024 , month = dec, publisher =. doi:10.1038/s41562-024-02026-z , issn =

    work page doi:10.1038/s41562-024-02026-z 2024

  4. [4]

    Robertson, Claire E. and Pr. Negativity drives online news consumption , journal=. 2023 , month=. doi:10.1038/s41562-023-01538-4 , url=

    work page doi:10.1038/s41562-023-01538-4 2023

  5. [5]

    Van Bavel and Sander van der Linden , title =

    Steve Rathje and Jay J. Van Bavel and Sander van der Linden , title =. Proceedings of the National Academy of Sciences , volume =. 2021 , doi =

    2021

  6. [6]

    Brady and Julian A

    William J. Brady and Julian A. Wills and John T. Jost and Joshua A. Tucker and Jay J. Van Bavel , title =. Proceedings of the National Academy of Sciences , volume =. 2017 , doi =

    2017

  7. [7]

    and Karger, Ezra , title =

    Murphy, Connacher and Rosenberg, Josh and Canedy, Jordan and Jacobs, Zach and Flechner, Nadja and Britt, Rhiannon and Pan, Alexa and Rogers-Smith, Charlie and Mayland, Dan and Buffington, Cathy and Kučinskas, Simas and Coston, Amanda and Kerner, Hannah and Pierson, Emma and Rabbany, Reihaneh and Salganik, Matthew and Seamans, Robert and Su, Yu and Tramèr,...

  8. [8]

    and Sagaria, Sabato D

    Wagenaar, Willem A. and Sagaria, Sabato D. , title =. Perception & Psychophysics , volume =. 1975 , doi =

    1975

  9. [9]

    Journal of Legal Analysis , volume =

    Kleinberg, Jon and Ludwig, Jens and Mullainathan, Sendhil and Sunstein, Cass R , title =. Journal of Legal Analysis , volume =. 2018 , month =. doi:10.1093/jla/laz001 , url =

    work page doi:10.1093/jla/laz001 2018

  10. [10]

    2026 , eprint=

    Economics of Human and AI Collaboration: When is Partial Automation More Attractive than Full Automation? , author=. 2026 , eprint=

    2026

  11. [11]

    Safety Impact Hub , year =

  12. [12]

    , title =

    West, Darrell M. , title =. 2024 , month = jul, day =

    2024

  13. [13]

    Fear in Self-Driving Vehicles Persists , year =

  14. [14]

    , title =

    Lee, Timothy B. , title =. 2025 , month = aug, day =

    2025

  15. [15]

    2026 , month = may, day =

    Sanders, James and Egan, Janet and Madigan, Rory , title =. 2026 , month = may, day =

    2026

  16. [16]

    2024 , howpublished =

    Schneider, James , title =. 2024 , howpublished =

    2024

  17. [17]

    2025 , month = nov, url =

    Digital Progress and Trends Report 2025 , institution =. 2025 , month = nov, url =

    2025

  18. [18]

    2025 , month = nov, url =

    Brynjolfsson, Erik and Chandar, Bharat and Chen, Ruyu , title =. 2025 , month = nov, url =

    2025

  19. [19]

    2026 , month = may, day =

    John Ruwitch , title =. 2026 , month = may, day =

    2026

  20. [20]

    2026 , month = feb, day =

    Angelo, Jake , title =. 2026 , month = feb, day =

    2026

  21. [21]

    2026 , month = jan, day =

    Quiroz-Gutierrez, Marco , title =. 2026 , month = jan, day =

    2026

  22. [22]

    2025 , month = oct, day =

    Preparing for. 2025 , month = oct, day =

    2025

  23. [23]

    Industrial Policy for the Intelligence Age: Ideas to Keep People First , year =

  24. [24]

    , title =

    Autor, David H. , title =. Journal of Economic Perspectives , volume =. 2015 , month = aug, doi =

    2015

  25. [25]

    2025 , month = jan, day =

    Future of Jobs Report 2025 , institution =. 2025 , month = jan, day =

    2025

  26. [26]

    1865 , publisher =

    Jevons, William Stanley , title =. 1865 , publisher =

  27. [27]

    Cambridge Journal of Regions, Economy and Society , volume =

    Acemoglu, Daron and Restrepo, Pascual , title =. Cambridge Journal of Regions, Economy and Society , volume =. 2020 , doi =

    2020

  28. [28]

    2026 , month = apr, howpublished =

    Introducing. 2026 , month = apr, howpublished =

    2026

  29. [29]

    2026 , month = apr, day =

    Mullin, Emily , title =. 2026 , month = apr, day =

    2026

  30. [30]

    and Gayre, Gregg and Hoberman, Brian and Mattern, Britt and Ballesca, Manuel and

    Tierney, Aaron A. and Gayre, Gregg and Hoberman, Brian and Mattern, Britt and Ballesca, Manuel and. Ambient Artificial Intelligence Scribes: Learnings after 1 Year and over 2.5 Million Uses , journal =. 2025 , doi =

    2025

  31. [31]

    npj Digital Medicine , volume =

    Krakowski, Isabelle and Kim, Jiyeong and Cai, Zhuo Ran and Daneshjou, Roxana and Lapins, Jan and Eriksson, Hanna and Lykou, Anastasia and Linos, Eleni , title =. npj Digital Medicine , volume =. 2024 , month = apr, doi =

    2024

  32. [32]

    Interval Cancer, Sensitivity, and Specificity Comparing

    Gommers, Jessie and Hernstr. Interval Cancer, Sensitivity, and Specificity Comparing. The Lancet , volume =. 2026 , month = jan, doi =

    2026

  33. [33]

    2026 , eprint=

    Remarks on the disproof of the unit distance conjecture , author=. 2026 , eprint=

    2026

  34. [34]

    Primitive sets and von

    Boris Alexeev and Kevin Barreto and Yanyang Li and Jared Duker Lichtman and Liam Price and Jibran Iqbal Shah and Quanyu Tang and Terence Tao , year=. Primitive sets and von. 2605.00301 , archivePrefix=

    Pith/arXiv arXiv

  35. [35]

    2024 , eprint=

    A Roadmap to Pluralistic Alignment , author=. 2024 , eprint=

    2024

  36. [36]

    Triplett , title =

    Jack E. Triplett , title =. Canadian Journal of Economics , volume =. 1999 , month = apr, doi =

    1999

  37. [37]

    Advances in Computers , editor =

    Good, Irving John , title =. Advances in Computers , editor =. 1965 , publisher =

    1965

  38. [38]

    2024 , eprint=

    Self-Taught Optimizer (STOP): Recursively Self-Improving Code Generation , author=. 2024 , eprint=

    2024

  39. [39]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  40. [40]

    2022 , eprint=

    Training language models to follow instructions with human feedback , author=. 2022 , eprint=

    2022

  41. [41]

    Nature , year=

    Silver, David and Schrittwieser, Julian and Simonyan, Karen and Antonoglou, Ioannis and Huang, Aja and Guez, Arthur and Hubert, Thomas and Baker, Lucas and Lai, Matthew and Bolton, Adrian and Chen, Yutian and Lillicrap, Timothy and Hui, Fan and Sifre, Laurent and van den Driessche, George and Graepel, Thore and Hassabis, Demis , title=. Nature , year=. do...

    work page doi:10.1038/nature24270

  42. [42]

    Reuters , year =

    Krystal Hu , title =. Reuters , year =

  43. [43]

    2026 , month = apr, day =

    Man charged after Molotov cocktail attack on OpenAI CEO Sam Altman's home , journal =. 2026 , month = apr, day =

    2026

  44. [44]

    Marcus, Gary , title =

  45. [45]

    2026 , month =

    Marcus, Gary , title =. 2026 , month =

    2026

  46. [46]

    2024 , month =

    Marcus, Gary , title =. 2024 , month =

    2024

  47. [47]

    Proceedings of the National Academy of Sciences , volume =

    Hoes, Emma and Gilardi, Fabrizio , title =. Proceedings of the National Academy of Sciences , volume =. 2025 , month = apr, doi =

    2025

  48. [48]

    , title =

    Hanna, Alex and Bender, Emily M. , title =. Scientific American , volume =

  49. [49]

    2023 , month = jun, day =

    As. 2023 , month = jun, day =

    2023

  50. [50]

    2025 , eprint=

    Superintelligence Strategy: Expert Version , author=. 2025 , eprint=

    2025

  51. [51]

    2023 , publisher =

    Suleyman, Mustafa and Bhaskar, Michael , title =. 2023 , publisher =

    2023

  52. [52]

    Assessing

    Carlini, Nicholas and Cheng, Newton and Lucas, Keane and Moore, Michael and Nasr, Milad and Prabhushankar, Vinay and Xiao, Winnie and Angulu, Hakeem and. Assessing. 2026 , month = apr, day =

    2026

  53. [53]

    2026 , month = apr, day =

    Project. 2026 , month = apr, day =

    2026

  54. [54]

    2026 , month = may, day =

    Adversaries Leverage. 2026 , month = may, day =

    2026

  55. [55]

    2025 , eprint=

    Quantifying CBRN Risk in Frontier Models , author=. 2025 , eprint=

    2025

  56. [56]

    2024 , month = feb, day =

    Staying Ahead of Threat Actors in the Age of. 2024 , month = feb, day =

    2024

  57. [57]

    2024 , eprint=

    Taking AI Welfare Seriously , author=. 2024 , eprint=

    2024

  58. [58]

    Exploring Model Welfare , year =

  59. [59]

    2026 , month = jan, day =

    Perrigo, Billy , title =. 2026 , month = jan, day =

    2026

  60. [60]

    Claude's Constitution , year =

    Askell, Amanda and Carlsmith, Joe and Olah, Chris and Kaplan, Jared and Karnofsky, Holden and. Claude's Constitution , year =

  61. [61]

    2025 , month = jan, day =

    Hinton, Geoffrey , title =. 2025 , month = jan, day =

    2025

  62. [62]

    2022 , month = feb, day =

    Sutskever, Ilya , title =. 2022 , month = feb, day =

    2022

  63. [63]

    2022 , publisher =

    Do Large Language Models Understand Us? , journal =. 2022 , publisher =

    2022

  64. [64]

    2026 , eprint=

    Towards a Science of AI Agent Reliability , author=. 2026 , eprint=

    2026

  65. [65]

    2020 , eprint=

    Scaling Laws for Neural Language Models , author=. 2020 , eprint=

    2020

  66. [66]

    Nature , volume =

    Lu, Chris and Lu, Cong and Lange, Robert Tjarko and Yamada, Yutaro and Hu, Shengran and Foerster, Jakob and Ha, David and Clune, Jeff , title =. Nature , volume =. 2026 , publisher =

    2026

  67. [67]

    Economic Policy , volume =

    Acemoglu, Daron , title =. Economic Policy , volume =. 2025 , publisher =

    2025

  68. [68]

    American Economic Journal: Macroeconomics , volume =

    Brynjolfsson, Erik and Rock, Daniel and Syverson, Chad , title =. American Economic Journal: Macroeconomics , volume =. 2021 , publisher =

    2021

  69. [69]

    2024 , month=

    Dario Amodei , title=. 2024 , month=

    2024

  70. [70]

    2025 , month=

    Sam Altman , title=. 2025 , month=

    2025

  71. [71]

    2023 , month = oct, day =

    Andreessen, Marc , title =. 2023 , month = oct, day =

    2023

  72. [72]

    2026 , month = jan, day =

    Rogelberg, Sasha , title =. 2026 , month = jan, day =

    2026

  73. [73]

    2005 , publisher =

    Kurzweil, Ray , title =. 2005 , publisher =

    2005

  74. [74]

    2024 , publisher =

    Kurzweil, Ray , title =. 2024 , publisher =

    2024

  75. [75]

    2018 , month = apr, howpublished =

    2018

  76. [76]

    Company , howpublished =

  77. [77]

    2024 , month = nov, day =

    Khosla, Vinod , title =. 2024 , month = nov, day =

    2024

  78. [78]

    2024 , month = sep, day =

    Altman, Sam , title =. 2024 , month = sep, day =

    2024

  79. [79]

    , title =

    Amdahl, Gene M. , title =. Proceedings of the. 1967 , publisher =

    1967

  80. [80]

    Journal of Law and the Biosciences , volume =

    Lenarczyk, Gabriela and Minssen, Timo and Price, Nicholson and Rai, Arti , title =. Journal of Law and the Biosciences , volume =. 2025 , month =. doi:10.1093/jlb/lsaf028 , url =

    work page doi:10.1093/jlb/lsaf028 2025

Showing first 80 references.