EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Truthfulness-constrained generative AI for job applications: a framework for ethical candidate empowerment in digital labor markets

Stephen Santhosh, Gabriela Jennifer

Frontiers in Computer Science · 17.07.2026 · 8

DOI: 10.3389/fcomp.2026.1892734 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Computer Science, online erschienen am 17.07.2026. Verfasst von Stephen Santhosh und 1 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Generative artificial intelligence tools are increasingly used by job seekers to produce application materials, raising concerns about credential fabrication and its downstream effects on hiring ecosystem integrity. Employers have responded with rigid applicant tracking system (ATS) configurations that, according to recent estimates, filter out qualified candidates who do not exactly match automated criteria. We propose a theoretical model we call the fabrication–filtration spiral , in which candidate fabrication prompts stricter employer filtering, which in turn invites more fabrication. We offer this as a hypothesis, not a demonstrated effect. The consequences fall disproportionately on vulnerable populations: immigrants in visa-dependent job searches, career changers, and candidates from non-traditional backgrounds. In this Perspective, we propose that candidate-facing AI tools should be designed with explicit architectural truthfulness constraints rather than operating as unconstrained optimization engines. Drawing on a patent-pending system and empirical observations from a deployed job analysis platform, we describe a three-component framework: (1) a tiered guidance engine that advises candidates when not to apply, (2) a resume content generator constrained to reframe only verified experience, and (3) a narrative generator governed by a hard truthfulness rule prohibiting the invention of facts absent from the candidate's resume. We situate this framework alongside recent work on proportionality constraints in AI for psychosocial support, arguing that bounded optimization—where AI systems operate within explicit ethical limits—is a generalizable design principle for human-facing AI in high-stakes contexts.

Zitieren

APA

Santhosh, S., & Jennifer, G. (2026). Truthfulness-constrained generative AI for job applications: a framework for ethical candidate empowerment in digital labor markets. Frontiers in Computer Science, 8. https://doi.org/10.3389/fcomp.2026.1892734

Vancouver

Santhosh S, Jennifer G. Truthfulness-constrained generative AI for job applications: a framework for ethical candidate empowerment in digital labor markets. Frontiers in Computer Science. 2026;8. doi:10.3389/fcomp.2026.1892734

BibTeX
@article{Santhosh2026,
  author = {Santhosh, Stephen and Jennifer, Gabriela},
  title = {Truthfulness-constrained generative AI for job applications: a framework for ethical candidate empowerment in digital labor markets},
  journal = {Frontiers in Computer Science},
  year = {2026},
  volume = {8},
  doi = {10.3389/fcomp.2026.1892734}
}
Kostenlos starten →