Algorithmic Accountability · Information Science
Notice is not enough.
A decision must be open to investigation.
I’m Kerry Thornhill, an independent researcher in algorithmic accountability and a graduate student at the University of North Texas. I developed the Necessary Informational Conditions (NIC) framework, which specifies what an affected person must be able to investigate in order to contest an institution’s automated determination. The first paper was published as a preprint in October 2026.
University of North Texas
Anuradha and Vikas Sinha College of Artificial Intelligence and Advanced Analytics
MS in Interdisciplinary Studies · In progress
Computer and Information Science and Engineering · Information Science
Research in algorithmic contestability
Twenty-five years on the institution’s side.
A research program on the other.
My research examines an asymmetry I know from the institution’s side. For twenty-five years I was the party that controlled the grounds, kept the records, and answered to the regulator.
I have held that position as a licensed fiduciary broker under the Texas Real Estate License Act, as owner of a motor carrier regulated concurrently by FMCSA, PHMSA, and FDA, and as chief executive of a manufacturer subject to both FDA food-safety and over-the-counter drug regimes.
Since 2008 I have also practiced knowledge engineering: encoding expert underwriting judgment into rule-based screening systems. That work is the same passage from governing standard to implemented rule that my research now interrogates.
The question I bring from that record is precise: when a fallible computational assessment acquires authority over a person’s options, what must remain investigable for that person to challenge it?
The experience behind the researchAn explanation is
only a beginning.
Transparency, explanation, and a human in the loop can each leave a determination effectively incontestable. My work replaces them with a more demanding measure: investigative capability.
Necessary informational conditions
A six-function register, a test of conditional necessity, and a five-part delivery architecture specifying what information must enable when contestability is required or promised. Published preprint.
02Relational institutional answerability
Why institutions exercising consequential private power through computational systems owe affected people informed challenge and consequential reconsideration. Companion paper, preprint expected January 2027.
03Testing the framework
A preregistered program of omission, substitution, and portability tests, with a linked study of when reliance on automated advice stays contributory and when it becomes preemptive.
03 / Accountability, then inquiry
Accountable for the record.
Now studying who can question it.
I have carried personal accountability under strict-liability safety regulation, statutory supervisory responsibility as a designated broker, and executive responsibility for multi-category FDA compliance. I know from the inside how institutions build, document, and defend their determinations.
I founded and direct the DG Humana Institute, a 501(c)(3) scientific and educational nonprofit that publishes interdisciplinary scholarship on human dignity, agency, and the future of work under automation and serves as a home for like-minded scholars. I am pursuing a Master of Science in Interdisciplinary Studies at the University of North Texas, where my research joins philosophical analysis, institutional case studies, and empirical design.
Follow the trajectory04 / What the work is for
Research others can
examine and use.
The NIC framework gives affected people and their representatives a basis for identifying what to investigate and request, and gives institutions and researchers a basis for designing and evaluating the routes through which those inquiries proceed.
Its necessity claims are bounded to stated conditions and built to be tested: a preregistered program can merge, split, or retire its categories. The planned contestability toolkit (affected-party and reviewer interfaces, a case generator, a grounds-to-disposition record, scoring tools, and domain modules for credit, public benefits, and health care) is designed for portability testing in a held-out setting.
Planned outputs also include open research materials, documented methods, synthetic cases, and teaching modules in experimental design, qualitative coding, and open science, for researchers and students. The published preprint, public outlines, and worked example on Contestability are available now; the toolkit, empirical studies, and teaching modules are in development.
Explore the worked example