AASHTO Journal - Q&A Podcast: AI, Accountability, and Accreditation

The latest episode of the AASHTO re:source Q&A podcast delves into the concerns posed by artificial intelligence or AI in the construction materials testing and inspection industry – especially in terms of how AI-generated policies, procedures, and records can be seriously flawed.

[Above image by AASHTO]

AASHTO re:source is a technical service program offered by the American Association of State Highway and Transportation Officials that provides services and tools through three major programs: the Laboratory Assessment Program or LAP, the Proficiency Sample Program or PSP, and the AAP.

This episode of the Q&A podcast – a podcast series originally launched in September 2020 – discusses a growing problem in the construction materials testing and inspection space: organizations submitting policies, procedures, and records that look AI-generated and haven’t been understood or even reviewed by a competent lab professional.

On the podcast, Brian Johnson and Kim Swanson – accreditation program director and communications manager, respectively, for AASHTO re:source – explained that a major concern with using AI is the legal and ethical boundary around laboratory standards derived from AASHTO and ASTM requirements.

Those standards come with copyright and user agreements that restrict creating derivative works, they explained, so using AI to generate a checklist, internal exam, procedure, template, or record “based on the standard” can cross that line quickly, noted AASHTO re:source, and even asking an AI platform to interpret the standard in a way that produces new work product can create exposure.

Johnson and Swanson also explain in this podcast episode how to spot questionable AI-generated documentation, with the main giveaway often nonsense detail: irrelevant equipment lists, overly long procedures stuffed with generic filler, or records turned into multi-tab spreadsheets that make a simple check nearly impossible.

To listen to the full episode, click here.