pith. sign in

arxiv: 2605.19234 · v1 · pith:FZFPMXTJnew · submitted 2026-05-19 · 💻 cs.CL · cs.AI

AI Technologies in Language Access: Attitudes Towards AI and the Human Value of Language Access Managers

Miguel A. Jim\'enez-Crespo , Stephanie Rodriguez , Alejandro Jaume Losa This is my paper

Pith reviewed 2026-05-20 06:34 UTC · model grok-4.3

classification 💻 cs.CL cs.AI
keywords AI in language accesstranslation technologyhuman oversightqualitative interviewsrisk awarenesslanguage access managersconditional optimismhealthcare and courts
0
0 comments X

The pith

Language access managers show conditional optimism toward AI but insist on human oversight due to risks in serving diverse users.

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

This paper explores how language access managers perceive the rise of AI tools in their work, which serves broad populations under legal, ethical, and safety pressures. Interviews with ten managers in US healthcare, court, public service, and local government settings reveal they see AI adoption as unavoidable yet approach it cautiously, stressing the irreplaceable role of human judgment. A sympathetic reader would care because these services affect real access to justice, medical care, and government help for non-native speakers. The study highlights how efficiency goals from AI clash with needs for accuracy and accountability in high-stakes environments. Understanding these views can shape better ways to blend new technology with established human practices.

Core claim

The thematic analysis shows that language access managers exhibit conditional optimism towards the inevitable AI implementations, are strongly risk aware, and are deeply committed to the human value and human oversight of AI implementations and output.

What carries the argument

Thematic analysis of ten semi-structured interviews with language access managers in US healthcare, court, public service and local government contexts.

If this is right

  • AI systems for language access should include built-in human review processes to handle identified risks.
  • Full automation without oversight is unlikely to gain acceptance from these professionals.
  • Guidelines for AI use in legal and medical language access must address the ethical and safety concerns raised.
  • Human roles in language access will likely shift toward supervision and quality control rather than direct translation.
  • Training for staff should emphasize skills for effective collaboration with AI tools.

Where Pith is reading between the lines

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

  • Developers could design AI tools that flag uncertain outputs for human review based on the risks these managers highlight.
  • Similar attitude studies in other countries might show how legal mandates shape optimism or caution differently.
  • Policy makers in public services could draw on these views when setting standards for AI in multilingual access.
  • Testing specific AI outputs against the oversight priorities described here could provide practical validation of the findings.

Load-bearing premise

The attitudes voiced by these ten specific managers in selected US settings represent the broader views of language access professionals, and the analysis captures those attitudes without major influence from how the interviews were designed or interpreted.

What would settle it

A larger survey of language access managers across more contexts that finds most do not express conditional optimism, risk awareness, or commitment to human oversight would undermine the reported patterns.

read the original abstract

The rapid emergence of AI technologies is reshaping translation practices and theory across the board. This paper deals with the impact of AI in language access. This area is characterized by the need to serve broad and diverse user populations, within a context where efficiency and access are shaped by legal mandates, ethical and commercial tensions, and safety concerns. This paper reports on the attitudes and perceptions of language access managers towards the AI and the human value in the AI age. Methodologically, this paper presents an analysis of a subset of a broader study on language access and technology, specifically a qualitative thematic analysis of ten semi-structured interviews with language access managers in the USA working in healthcare, court, public service and local government contexts. The results indicate that language access managers show conditional optimism towards the inevitable AI implementations, are strongly risk aware, and deeply committed to the human value and human oversight of AI implementations and output.

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 / 1 minor

Summary. The paper presents a qualitative thematic analysis of ten semi-structured interviews with language access managers working in US healthcare, court, public service, and local government settings. It reports that these managers express conditional optimism about inevitable AI implementations in translation and language access, while demonstrating strong risk awareness and a deep commitment to preserving human value and oversight in AI-assisted processes.

Significance. If the reported attitudes are robustly grounded, the work offers timely descriptive insights into professional perspectives on AI in regulated, high-stakes language-access domains where legal mandates and safety concerns intersect with efficiency pressures. The emphasis on human oversight aligns with ongoing debates in AI ethics and could inform policy or training guidelines, though the small purposive sample constrains claims of representativeness across the broader profession.

major comments (2)
  1. [Methods] Methods description (abstract and corresponding section): The account of the ten interviews provides no information on recruitment strategy, purposive sampling criteria, saturation, codebook development, inter-coder reliability, or member checking. These omissions are load-bearing because the central claims about conditional optimism, risk awareness, and commitment to human oversight rest entirely on the thematic analysis of this specific subset; without such details, it is impossible to evaluate selection bias or interpretive robustness.
  2. [Results] Results and conclusion: The phrasing that 'language access managers show conditional optimism...' generalizes from the ten selected participants to the profession at large. The manuscript does not address how the purposive sample drawn from particular US contexts captures relevant variation (e.g., by organization size, AI exposure, or regional policy differences), undermining the strength of the headline generalization.
minor comments (1)
  1. [Abstract] The abstract could explicitly note that the ten interviews constitute a subset of a larger study and briefly flag the corresponding scope limitations.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which help clarify the presentation of our qualitative study. We address each major point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Methods] Methods description (abstract and corresponding section): The account of the ten interviews provides no information on recruitment strategy, purposive sampling criteria, saturation, codebook development, inter-coder reliability, or member checking. These omissions are load-bearing because the central claims about conditional optimism, risk awareness, and commitment to human oversight rest entirely on the thematic analysis of this specific subset; without such details, it is impossible to evaluate selection bias or interpretive robustness.

    Authors: We agree that the current methods description is insufficient for evaluating the analysis. In the revised manuscript we will expand the methods section to detail the recruitment strategy, the purposive sampling criteria applied to select language access managers from healthcare, court, public service and local government contexts, how thematic saturation was assessed, the iterative codebook development process, and any steps taken toward interpretive robustness (including whether inter-coder reliability checks or member checking were performed). revision: yes

  2. Referee: [Results] Results and conclusion: The phrasing that 'language access managers show conditional optimism...' generalizes from the ten selected participants to the profession at large. The manuscript does not address how the purposive sample drawn from particular US contexts captures relevant variation (e.g., by organization size, AI exposure, or regional policy differences), undermining the strength of the headline generalization.

    Authors: We accept that the current phrasing risks implying broader representativeness than the data support. We will revise the results and conclusion sections to state that the observed conditional optimism, risk awareness, and commitment to human oversight were found among the ten participants in this purposive sample. We will also add explicit discussion of the sample's characteristics and limitations with respect to organization size, prior AI exposure, and regional policy variation, while clarifying the exploratory scope of the study. revision: yes

Circularity Check

0 steps flagged

No circularity: qualitative empirical analysis of interview data

full rationale

The paper conducts a qualitative thematic analysis of ten semi-structured interviews drawn from a subset of a larger study on language access managers in US contexts. The central claims about conditional optimism, risk awareness, and commitment to human oversight are presented as direct outputs of coding the interview responses rather than any derivation, equation, fitted parameter, or self-referential prediction. No mathematical steps, uniqueness theorems, ansatzes, or load-bearing self-citations appear in the reported chain; the mapping from transcripts to themes is treated as an interpretive but non-circular empirical process. The study is therefore self-contained against external benchmarks with no reduction of results to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

This is an empirical qualitative study relying on standard social science methods with no free parameters, mathematical axioms, or invented entities introduced to support the claims.

axioms (1)
  • domain assumption Semi-structured interviews with a small purposive sample combined with thematic analysis can reliably identify and represent professional attitudes toward AI.
    Invoked implicitly in the description of the method and results; this is a common assumption in qualitative research but depends on sample representativeness and minimal bias.

pith-pipeline@v0.9.0 · 5691 in / 1337 out tokens · 67138 ms · 2026-05-20T06:34:21.339459+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

14 extracted references · 14 canonical work pages

  1. [1]

    less than very well

    require government agencies, as well as those receiving state funds such as hospitals or non- profits, to translate vital documents into the most spoken languages. In the USA, approximately 22% of people speak a language other than English at home while 8-9% identify as speaking English “less than very well” (25.7 million) (U.S. Census). In the Tri-State ...

    work page 2023

  2. [2]

    or end users, including for example migrant communities (Yang and Valdes 2025). The urgency of this research question is due to serious risks involved in technology-mediated communicative events in critical contexts and the need to guarantee language access rights. These risks range from low acceptance rates, misinformation, gender or racial biases, lack ...

    work page 2025

  3. [3]

    These studies show that perceptions and use of AI in translation depend on a wide variety of factors

    or students and trainees (ELIS 2025). These studies show that perceptions and use of AI in translation depend on a wide variety of factors. One of these factors being the perceived imposition by those professionals that manage organizations, such as translation, localization, or language access managers, LSPs, digital platform managers, etc. (e.g., Jimene...

    work page 2025

  4. [4]

    The survey was designed using the online platform Qualtrics, while the script for the semi-structured interviews was adapted from the interview questions for the European LT-Lider project (Secară et al. 2025). Ethical approval was obtained from the Institutional Review Board at Rutgers University, with the protocol approved on 8-18-2025. A combination of ...

    work page 2025

  5. [5]

    language access coordinator

    was used. Potential participants were recruited through Google and LinkedIn searches for job profiles such as “language access coordinator”, “language access manager” or related titles. The online survey included a final question asking about their availability to participate in a semi-structured interview. Participants who responded with willingness were...

    work page 2006

  6. [6]

    streamline the processes

    or AI-assisted brainstorming and inspiration (Rivas Ginel and Moorkens 2024). This is reflected in language access managers attitudes, as participants express it truly helps to “streamline the processes” (P3) and make their work “more efficient” (P1). Efficiency and higher quality are often mentioned as the benefits of higher productivity, as indicated by...

    work page 2024

  7. [7]

    Translation Spaces, 9(1):58–77

    Risks in neural machine translation. Translation Spaces, 9(1):58–77. https://doi.org/10.1075/ts.00021.can Capel, T. and Brereton, M

    work page doi:10.1075/ts.00021.can

  8. [8]

    InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems(Hamburg, Germany)(CHI ’23)

    Mapping human-centered artificial intelligence research. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI ’23), pages 1–23. https://doi.org/10.1145/3544548 ELIS

    work page doi:10.1145/3544548 2023

  9. [9]

    http://elis-survey.org/wp-content/uploads/2025/03/ELIS-2025_Results.pdf Elsayed, N

    work page 2025

  10. [10]

    arXiv preprint arXiv:2601.00382

    Unseen risks of clinical speech-to-text systems: Transparency, privacy, and reliability challenges in AI-driven documentation. arXiv preprint arXiv:2601.00382. Flores, G

    work page arXiv

  11. [11]

    Information, 16(5):387

    Human-centered AI and the future of translation technologies. Information, 16(5):387. https://doi.org/10.3390/info16050387 Kenny, D. and Winters, M

    work page doi:10.3390/info16050387

  12. [12]

    In Teaching translation in the age of generative AI: New paradigm, new learning?, pages iii–ix

    Introduction. In Teaching translation in the age of generative AI: New paradigm, new learning?, pages iii–ix. Language Science Press. https://doi.org/10.5281/zenodo.17580856 Rivas Ginel, M. T. and Moorkens, J

    work page doi:10.5281/zenodo.17580856

  13. [13]

    https://doi.org/10.25365/phaidra.641 11 Tomlinson, K., Jaffe, S., Wang, W., Counts, S., and Suri, S

    LT-LiDER language technology map: Technologies in translation practice and their impact on the skills needed. https://doi.org/10.25365/phaidra.641 11 Tomlinson, K., Jaffe, S., Wang, W., Counts, S., and Suri, S

    work page doi:10.25365/phaidra.641

  14. [14]

    arXiv preprint arXiv:2507.07935 , year=

    Working with AI: Measuring the applicability of generative AI to occupations. arXiv preprint arXiv:2507.07935. Toral, A

    work page arXiv