KU
Student / school
THE FRAMEWORK
A simple language for a consequential act.
The Hukumu Project rests on three terms. Together they describe where AI judges people, and how those judgments are recorded inside an organization and in public.
WHAT WE TRACK
Scope
The Hukumu Project tracks one thing: when AI or an algorithmic system is used to judge an identifiable person. 'Judge' means classify, sort, rank, score, recommend, flag, approve, deny, or otherwise assign a status or outcome to someone. We do not take on the wider field of AI risk, such as bias in training data generally, job displacement, misinformation, or model safety. Those matter, but they are out of scope unless they surface inside a specific judgment event. Our job is transparency and discoverability of judgments already being made, not policing what organizations decide. Judgments about organizations are outside the initial scope.
We record the judgment, not the merits.
A domain can define how to record an AI recommendation about a promotion or a clinical priority without deciding which promotion criteria or clinical choices are acceptable.
01 / KU
KUwhere AI judges
A KU is a context in which an AI or machine-learning system makes a judgment about a person. Judging covers far more than yes-or-no decisions. It includes classifying, ranking, scoring, predicting, recommending, flagging, selecting, and warning against. KUs follow the roles people play: student, employee, patient, customer, member, donor, constituent, citizen. A single person lives inside many KUs at once.
02 / DAB
DABthe internal record
A DAB is the record an organization keeps of an AI judgment it made. It answers what was judged, by which system, when, and with what result. DABs let an organization see its own AI decision-making and govern it. A DAB is permissioned. The person who was judged, and other authorized parties, can access it through secure, identity-verified channels.
03 / DIB
DIBthe public record
A DIB is the public, discoverable disclosure a domain requires. It is ordinarily about the system and its aggregate outcomes, not an identifiable person's case. It makes AI judgment discoverable without exposing individuals.
ONE JUDGMENT. TWO RECORDS.
An AI judgment
DABInternal record
DIBPublic record
IN PRACTICE
A worked example
A university uses a model to flag students at risk of academic probation.
DAB
The internal record: full inputs, model version, score, and which staff member reviewed it. Kept by the university and available to auditors or regulators.
DIB
The public record: that an automated flagging process exists for this purpose, its general criteria, and aggregate statistics. Published where the affected student, or anyone, can find it, without exposing case-level details.
One judgment event, two records, two different disclosure rules, both filed under one KU domain.
THE RULES
Deciding what goes where
Which judgments must be recorded, and which records become public, will differ by domain. A hiring screen and a hospital triage tool raise different questions. Each domain working group will set its own rules, published as open standards.
THE ORIGIN
Why cuneiform
The name
Hukumu is the Swahili word for judgment, from the Arabic ḥukm. As a verb it means to judge, or to pass judgment upon. Hidden inside it is KU: Hu-KU-mu.
The signs
KU · DAB: two readings of one sign
KU
An attested Sumerian sense is “to place, to lay.” We extend it to mean placing a person in a role before a judgment is made. This is a deliberate extension, not a literal translation.
DAB
The dab₅ reading of the same sign, meaning “to seize, hold, take charge of.” The Old Babylonian compound di dab₅ (“case” + dab) means “to render a verdict,” so DAB has real judicial usage.
DIB
A separate sign. “Dib” can be a variant of dab (“to seize”), but we use its other sense, “to pass, to pass in review (at an inspection),” which is what a public record does.
An honest note
Sumerian is a language isolate, and Hukumu comes from a Semitic root by way of Swahili. The pairing is a branding choice, not a claim of shared origin. We chose it because AI judgment can feel technical, frightening, and controversial, and three short, ancient words for an old human practice, keeping honest records of decisions, make it easier to remember and talk about.
Sources: Sumerian Lexicon, J. Halloran · ePSD2, Oracc
TAKE PART
Get involved
Whether you've been judged by an algorithm, build one, regulate one, or fund work like this, there's a place for you.