How can automated software remain accountable?
I explore evidence, reversibility, human authority, and the difference between a generated suggestion and a justified decision.
Ideas under construction
I write about software, testing, cognition, identity, and the systems people build around them. These papers are independent personal research: public, provisional, and open to revision when better evidence arrives.
SEC-01
About this site
I'm Eric Mumford. This is a public notebook for research questions that interest me and for the designs, counterarguments, and reading trails that grow from them.
The work is undertaken autonomously, without organizational sponsorship, and is separate from professional work. It uses public sources and personal reasoning and contains no confidential material.
Some entries are complete working papers. Others are clearly labeled research notes in development. Neither label implies peer review or a finished conclusion.
SEC-02
01
02
03
04
Questions I keep returning to
A changing map of interests, not a product catalog
What changes when AI helps design tests?
I examine where generative tools may help, what they fail to observe, and how established testing patterns can contain their uncertainty.
How do identity and memory persist through change?
I compare ideas from philosophy, cognitive science, and software models without treating an analogy as proof.
Can different systems share a useful language?
I look for small common models that help methods communicate while preserving the differences that make each method useful.
SEC-03
How the work is presented
Each paper starts in ordinary language, defines necessary technical terms, distinguishes observation from interpretation, and names what could change the conclusion.
References support the lineage of an idea; they do not transfer authority to my design. Architecture diagrams describe relationships. They are not evidence that a system has been built, deployed, or proven.
The archive uses one author-declared date for each current record and omits reader-facing document versions. A date records the author's recollection of origin and publication; it is not independent archival proof.