pith. sign in

arxiv: 2605.17203 · v1 · pith:QWLZ4W7Mnew · submitted 2026-05-17 · 💻 cs.CY

Beyond Model Readiness: Institutional Readiness for AI Deployment in Public Systems

Erika Fille Legara , Elmo Domino Jose , Paula Joy Martinez This is my paper

Pith reviewed 2026-05-19 23:28 UTC · model grok-4.3

classification 💻 cs.CY
keywords institutional readinessAI deploymentpublic sectorresponsible AIdeployment frameworkoperational barrierseducation technology
0
0 comments X

The pith

Public AI systems often stall at deployment because institutions lack operational, data, oversight, fiscal, and regulatory readiness.

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

The paper argues that many AI projects in public systems reach technical viability but cannot advance because the receiving institution lacks the structures and capacities needed for responsible use. Drawing on two cases from a large public education system—an image-based anthropometric screening tool and a speech-analysis system for early learning risk—the authors show how institutional factors, not model performance, blocked broader rollout. They propose the Institutional Alignment Readiness framework as a practical tool for resource-constrained public settings to diagnose these gaps. The framework is meant to work alongside existing model-focused evaluations and to guide decisions on whether to stop, pilot, or expand a system.

Core claim

We introduce Institutional Alignment Readiness (IAR), a five-dimensional framework for assessing deployment readiness in public systems. The framework is designed for resource-constrained settings, where gaps between technical viability and responsible deployment are most acute. It is grounded in two anonymized operational cases from a large public education system: an image-based anthropometric screening tool and a speech-analysis system for early learning risk identification. Both reached technically viable stages but could not advance to broader rollout for institutional rather than technical reasons. We use these cases to motivate a practical readiness framework covering institutional, 1

What carries the argument

Institutional Alignment Readiness (IAR), a five-dimensional framework that evaluates the receiving institution across operational compatibility, data ecosystem maturity, human oversight capacity, fiscal sustainability, and regulatory alignment.

Load-bearing premise

Two anonymized cases from a single large public education system provide a sufficient basis for a general framework that applies across other public-sector domains and resource-constrained settings.

What would settle it

A public institution successfully deploying an AI system while failing to meet one or more of the five IAR dimensions would show the framework is not necessary for deployment success.

Figures

Figures reproduced from arXiv: 2605.17203 by Elmo Domino Jose, Erika Fille Legara, Paula Joy Martinez.

Figure 1
Figure 1. Figure 1: From artifact-level evaluation to deployment readiness. Existing AI evaluation tools assess whether a model or dataset is technically suitable for its intended use; IAR adds a second layer that assesses whether the receiving institution is ready to use the system responsibly at the intended deployment stage. Note: Although presented sequentially, IAR dimensions such as Data Ecosystem Maturity are relevant … view at source ↗
Figure 2
Figure 2. Figure 2: Deployment trajectories of the two operational cases. Both projects followed a similar arc from stakeholder-defined need to early development, the emergence of institutional constraints, and a current stage of internal validation or pilot. In the anthropometric screening system, the main bottlenecks concerned approvals, data representativeness, and referral capacity; in the speech-based risk identification… view at source ↗
read the original abstract

Many public-sector artificial intelligence systems fail not at the point of model development, but at the point of deployment. Systems that perform well in internal testing may still stall because the receiving institution lacks the approvals, data arrangements, human oversight, operational capacity, fiscal continuity, or legal clarity needed for broader rollout. Existing responsible AI and model evaluation frameworks are valuable, but they primarily assess models, datasets, and developer-side processes, not the readiness of the institution that must use the system in practice. We introduce Institutional Alignment Readiness (IAR), a five-dimensional framework for assessing deployment readiness in public systems. The framework is designed for resource-constrained settings, where gaps between technical viability and responsible deployment are most acute. It is grounded in two anonymized operational cases from a large public education system: an image-based anthropometric screening tool and a speech-analysis system for early learning risk identification. Both reached technically viable stages but could not advance to broader rollout for institutional rather than technical reasons. We use these cases to motivate a practical readiness framework covering institutional and operational compatibility, data ecosystem maturity, human oversight capacity, fiscal sustainability, and regulatory alignment readiness. IAR is designed to complement, not replace, established AI evaluation tools. It assesses the receiving institution rather than the artifact alone and supports staging decisions such as no-go, pilot-only, or readiness for broader deployment.

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 / 2 minor

Summary. The manuscript argues that public-sector AI systems frequently fail at deployment rather than model development due to institutional shortcomings, and introduces the Institutional Alignment Readiness (IAR) framework—a five-dimensional construct covering institutional/operational compatibility, data ecosystem maturity, human oversight capacity, fiscal sustainability, and regulatory alignment—to assess receiving institutions in resource-constrained settings. The framework is motivated by two anonymized cases from a single large public education system (image-based anthropometric screening and speech-analysis for early learning risk), both of which reached technical viability but stalled for non-technical reasons. IAR is positioned as complementary to existing model-focused responsible AI tools and intended to inform staging decisions such as no-go, pilot-only, or broader deployment.

Significance. If the framework can be operationalized and tested more broadly, it would address a genuine gap by shifting focus from model performance to institutional capacity in public AI deployments, particularly in under-resourced contexts. The practical orientation toward staging decisions and the grounding in real operational examples are strengths that could make the work useful for practitioners, though its value hinges on demonstrating transportability beyond the motivating cases.

major comments (2)
  1. [Abstract / Motivating Cases] Abstract and motivating cases section: the framework is derived from only two anonymized cases within a single public education system, with no cross-sector examples, derivation process, or test of exhaustiveness provided. This makes the claim of applicability to broader public systems (e.g., healthcare allocation or infrastructure) a load-bearing assumption that requires explicit discussion of domain adaptations or limitations to support the central contribution.
  2. [IAR Framework Description] Framework presentation: the five dimensions are introduced without explicit mapping to barriers observed in the two cases, inclusion/exclusion criteria, or validation metrics, leaving the comprehensiveness of the construct conceptually motivated but not empirically anchored in the manuscript.
minor comments (2)
  1. [Framework] Clarify potential overlaps between dimensions (e.g., how 'institutional/operational compatibility' differs from 'human oversight capacity') to improve usability for practitioners.
  2. [Discussion] Add a brief limitations subsection discussing the education-domain origin and any steps taken to anonymize or generalize the cases.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the boundaries of our contribution. We respond to each major comment below and indicate the revisions we will make to address the concerns about scope and empirical anchoring.

read point-by-point responses

  1. Referee: [Abstract / Motivating Cases] Abstract and motivating cases section: the framework is derived from only two anonymized cases within a single public education system, with no cross-sector examples, derivation process, or test of exhaustiveness provided. This makes the claim of applicability to broader public systems (e.g., healthcare allocation or infrastructure) a load-bearing assumption that requires explicit discussion of domain adaptations or limitations to support the central contribution.

    Authors: We agree that the motivating cases are limited to two anonymized examples from a single public education system. In the revised manuscript we will add a dedicated 'Scope, Limitations, and Domain Adaptations' subsection. This will explicitly discuss how the five IAR dimensions may require modification when applied to other sectors (e.g., adjusting regulatory-alignment considerations for healthcare data-protection regimes or infrastructure procurement rules) and will trace the derivation of each dimension to the concrete institutional barriers observed in the two cases. While we cannot introduce new cross-sector empirical cases or conduct a formal exhaustiveness test within the scope of this revision, the added discussion will replace the current load-bearing assumption with a more cautious and transparent statement of intended applicability. revision: yes

  2. Referee: [IAR Framework Description] Framework presentation: the five dimensions are introduced without explicit mapping to barriers observed in the two cases, inclusion/exclusion criteria, or validation metrics, leaving the comprehensiveness of the construct conceptually motivated but not empirically anchored in the manuscript.

    Authors: We accept that the current presentation would be strengthened by tighter empirical linkage. We will revise the framework section to include a mapping table that directly connects each dimension to the specific non-technical barriers that halted deployment in the anthropometric-screening and speech-analysis cases. We will also articulate the inclusion criteria used to select the five dimensions, explaining how they emerged from the observed gaps between technical viability and institutional capacity. As the paper presents a conceptual framework rather than a psychometrically validated instrument, we will not add quantitative validation metrics; instead we will note this as an important direction for future empirical work while clarifying the qualitative grounding already present in the cases. revision: yes

Circularity Check

0 steps flagged

No circularity in IAR framework derivation from case observations

full rationale

The paper introduces the IAR five-dimensional framework by synthesizing observations from two anonymized operational cases in a single public education system, where technically viable AI tools failed to deploy due to institutional factors. The dimensions are explicitly motivated by the specific barriers encountered in those cases rather than being defined in terms of themselves, fitted to parameters, or derived via equations that loop back to inputs. No self-citations, uniqueness theorems, or ansatzes are invoked as load-bearing elements; the framework is presented as a qualitative, practical complement to existing model-focused tools. The derivation chain is self-contained as an empirical-to-conceptual mapping without reduction by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The paper introduces the IAR framework as a new construct resting on domain assumptions about the nature of deployment failures and the representativeness of the two education cases.

axioms (1)
  • domain assumption Institutional factors beyond model performance are the dominant reason public AI systems fail to reach broader rollout.
    This premise justifies creating a separate institutional assessment layer rather than extending existing model-evaluation tools.
invented entities (1)
  • Institutional Alignment Readiness (IAR) framework no independent evidence
    purpose: To assess institutional deployment readiness for AI in public systems
    Newly proposed five-dimensional construct motivated by the two cases.

pith-pipeline@v0.9.0 · 5774 in / 1247 out tokens · 57038 ms · 2026-05-19T23:28:29.232006+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    An Artificial Intelligence Maturity Model for the Public Sector: A Design Science Approach , journal =

    Richard Dreyling and Juhani Lemmik and Tanel Tammet and Ingrid Pappel , doi =. An Artificial Intelligence Maturity Model for the Public Sector: A Design Science Approach , journal =. 2024 , pages =

    work page 2024

  2. [2]

    2023 , type =

    Artificial Intelligence Risk Management Framework (AI RMF 1.0) , institution =. 2023 , type =

    work page 2023

  3. [3]

    2023 , number =

    Information technology --- Artificial intelligence --- Management system , institution =. 2023 , number =

    work page 2023

  4. [4]

    Big Data & Society , volume =

    Angèle Christin , title =. Big Data & Society , volume =. 2017 , doi =

    work page 2017

  5. [5]

    Algorithmic realism: expanding the boundaries of algorithmic thought , year =

    Green, Ben and Viljoen, Salom\'. Algorithmic realism: expanding the boundaries of algorithmic thought , year =. doi:10.1145/3351095.3372840 , booktitle =

    work page doi:10.1145/3351095.3372840

  6. [6]

    1995 , publisher=

    Tinkering toward Utopia: A Century of Public School Reform , author=. 1995 , publisher=

    work page 1995

  7. [7]

    Reading Research Quarterly , volume=

    Introduction to Response to Intervention: What, why, and how valid is it? , author=. Reading Research Quarterly , volume=

    work page

  8. [8]

    Advances in Neural Information Processing Systems , pages=

    Hidden technical debt in machine learning systems , author=. Advances in Neural Information Processing Systems , pages=

    work page

  9. [9]

    1980 , publisher=

    Street-level Bureaucracy: Dilemmas of the Individual in Public Services , author=. 1980 , publisher=

    work page 1980

  10. [10]

    Arizona State Law Journal , volume=

    The Structural Consequences of Educational Privacy , author=. Arizona State Law Journal , volume=

    work page

  11. [11]

    The global landscape of AI ethics guidelines,

    Anna Jobin and Marcello Ienca and Effy Vayena , title =. Nature Machine Intelligence , volume =. doi:10.1038/s42256-019-0088-2 , pages =

    work page doi:10.1038/s42256-019-0088-2

  12. [12]

    URL https://cacm.acm.org/research/ datasheets-for-datasets/

    Gebru, Timnit and Morgenstern, Jamie and Vecchione, Briana and Vaughan, Jennifer Wortman and Wallach, Hanna and III, Hal Daum\'. Datasheets for datasets , year =. doi:10.1145/3458723 , journal =

    work page doi:10.1145/3458723

  13. [13]

    The Information Society , volume =

    Richard Heeks , title =. The Information Society , volume =

    work page

  14. [14]

    Model Cards for Model Reporting,

    Mitchell, Margaret and Wu, Simone and Zaldivar, Andrew and Barnes, Parker and Vasserman, Lucy and Hutchinson, Ben and Spitzer, Elena and Raji, Inioluwa Deborah and Gebru, Timnit , title =. 2019 , isbn =. doi:10.1145/3287560.3287596 , booktitle =

    work page doi:10.1145/3287560.3287596 2019

  15. [15]

    Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency , location =

    Raji, Inioluwa Deborah and Smart, Andrew and White, Rebecca N. and Mitchell, Margaret and Gebru, Timnit and Hutchinson, Ben and Smith-Loud, Jamila and Theron, Daniel and Barnes, Parker , title =. 2020 , isbn =. doi:10.1145/3351095.3372873 , booktitle =

    work page doi:10.1145/3351095.3372873 2020

  16. [16]

    and Boyd, Danah and Friedler, Sorelle A

    Selbst, Andrew D. and Boyd, Danah and Friedler, Sorelle A. and Venkatasubramanian, Suresh and Vertesi, Janet , title =. Proceedings of the Conference on Fairness, Accountability, and Transparency , pages =. 2019 , isbn =

    work page 2019

  17. [17]

    2019 , booktitle =

    Michael Veale and Irina Brass , title =. 2019 , booktitle =

    work page 2019

  18. [18]

    2021 , pages =

    Nithya Sambasivan and Shivani Kapania and Hannah Highfill and Diana Akrong, Praveen Paritosh and Lora M Aroyo , title =. 2021 , pages =

    work page 2021

  19. [19]

    Langley , title =

    P. Langley , title =. Proceedings of the 17th International Conference on Machine Learning (ICML 2000) , address =. 2000 , pages =

    work page 2000

  20. [20]

    T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980

    work page 1980

  21. [21]

    M. J. Kearns , title =

    work page

  22. [22]

    Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983

    work page 1983

  23. [23]

    R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000

    work page 2000

  24. [24]

    Suppressed for Anonymity , author=

    work page

  25. [25]

    Newell and P

    A. Newell and P. S. Rosenbloom. Mechanisms of Skill Acquisition and the Law of Practice. Cognitive Skills and Their Acquisition. 1981

    work page 1981

  26. [26]

    A. L. Samuel. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and Development. 1959

    work page 1959

  27. [27]

    Fostering Implementation of Health Services Research Findings into practice: a Consolidated Framework for Advancing Implementation Science , doi =

    Damschroder, Laura J and Aron, David C and Keith, Rosalind E and Kirsh, Susan R and Alexander, Jeffery A and Lowery, Julie C , month =. Fostering Implementation of Health Services Research Findings into practice: a Consolidated Framework for Advancing Implementation Science , doi =. 2009 , journal =

    work page 2009

  28. [28]

    2016 , organization =

    World Health Organization , title =. 2016 , organization =

    work page 2016