Topic hub
An analysis of AI through procedures: who sets criteria, bears the cost of error, and gives people a route to challenge a decision.
Position
AI and society is treated as an analysis of procedure, accountability, and the consequences of error, not as a catalogue of product promises. When a system affects case order, access to a resource, or an assessment of a person, it must be possible to name the owner of a criterion, the route to challenge it, and the cost of a mistake for different people. Average effectiveness does not answer those questions. The texts gathered here examine where rules are hidden in data, instructions, or an interface and how they can be made visible to someone affected by an output. They do not assume automation is the default solution.
The practical question is not whether a system is intelligent enough, but whether a person affected by it has an understandable procedure, a way to correct information, and a real route to a person. If those conditions cannot be named, the process needs changing before deployment.
A question about AI’s effect on an institution does not start with model capability but with procedure. Who selects a criterion, admits sources, defines an exception, and decides when an output may be used? Those answers are part of social design rather than a technical detail. Without them, automation can only make implicit rules operate faster.
Every classification or priority criterion has an author, even when it is hidden in data or a model. An institution should be able to name the role that approves that criterion, reviews its change, and answers a challenge. Saying that an algorithm made the decision is not enough. That account obscures responsibility instead of showing where it can be meaningfully questioned.
An error has different meaning for a person affected by an outcome, an employee performing a procedure, and an organisation maintaining a system. Evaluation therefore cannot end with average effectiveness. It needs to consider who bears the cost of a false rejection, delay, data exposure, or inability to explain a result. Those costs determine whether automation is acceptable at a particular point at all.
A person affected by an output needs an understandable route to ask, correct, and appeal, while a worker needs to know how to stop an action without bypassing the system. A challenge mechanism is not an addition after deployment; it shapes requirements for source records, versions, decisions, and retention. This topic remains an analysis of accountability, not a promise of an automatic product.
In practice
Model capability does not decide how an institution changes. Procedures do: who selects a criterion, admits sources, bears the cost of error, and whether a person affected by an outcome has a route to challenge it. This page does not turn that question into a commercial service promise.
Contact here is for describing a research question or process, not for automatic product matching.
BY-UA and the source-aware material lead to examples of public work with an argument.
Projects
Research / Open and Applied Research
A case study of public material and editorial method: sources, claims, and their presentation status are kept distinct. It does not attribute a formal role in the external project to this site’s author.
This site’s case study presents an editorial method in which a source, a claim, and a description of its status are separate elements.
live demo
Start here
A long time series can reveal structural change, but only with explicit definitions, sources, and a clear boundary between history and scenario.
Reading
An essay on cheap applied intelligence, the reorganisation of work, and the value of intent, judgment, relationships, and new tasks.
Read the essay →An original analysis by Sergei Audzeichyk of questions raised by Yuval Noah Harari’s lecture about agency, institutions, language, and human judgment.
Read the original analysis →