Problem-Space Transformation
Problem solving is commonly modeled as search within a representation of a task. In real design, engineering, organizational, and market settings, however, the task is rarely given as a complete and privileged object. Agents construct operational problem spaces from partial evidence, goals, commitments, representations, and assumptions; they revise these spaces while acting; and they often obtain large gains not by searching harder, but by transforming the space in which search and action occur. This paper proposes Problem-Space Transformation (PST) as an agent-relative framework linking three processes: epistemic revision, structural transformation, and search/execution. The framework explicitly rejects the need for a unique 'true problem' while retaining objective constraint through consequences, evidence, and intervention. It introduces a non-exhaustive taxonomy of transformation operators, including separation, identification, mediation, constraint shaping, temporal relocation, agency relocation, observability, inversion, reduction, self-similar design, and graded commitment. It also proposes vector-valued complexity accounting: a transformation may reduce search cost while increasing coordination, failure impact, or change cost; complexity is not conserved, but frequently relocates. Four cases illustrate the framework: reformulating exactly-once delivery as exactly-once business effect, self-similar design as reduction of independent rules with correlated-risk trade-offs, product vision as collective problem construction rather than direct customer transcription, and trial batches as actions that are simultaneously pragmatic and epistemic. The paper positions PST relative to problem-space theory, reflective framing, problem-solution co-evolution, epistemic action, wicked problems, causal reasoning, and TRIZ. The contribution is a unifying operational vocabulary and a set of testable propositions rather than a claim that its individual mechanisms are new.
Authors
- Alexey POLOVINKIN (ORCID: https://orcid.org/0009-0009-4307-6498)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24429324
- Primary Topic
- Design Education and Practice
- Type
- preprint