Title Responsibility to the Problem: An Explicit Standard for the Expert–AI System
Author(s) Laboratory of Information Systems
Affiliation New Euro Vision: exhibitions, marketing, research s.r.o.
Country Czech Republic
Contribution type conceptual paper; methodological paper
Thematic area • Theory, Methodology, and Evidence
Conference framework connection • Analytical Track AT-01 — provisional title to be defined• Analytical Track AT-02 — provisional title to be defined• Analytical Track AT-03 — provisional title to be defined• Analytical Track AT-04 — provisional title to be defined• Analytical Track AT-05 — provisional title to be defined• Analytical Track AT-06 — provisional title to be defined• Analytical Track AT-07 — provisional title to be defined
Abstract Discussions of artificial intelligence in science frequently ask who is responsible when AI produces an incorrect result. In its usual form, this question is too vague to guide scientific work. It does not identify the object of responsibility, the decision under evaluation, the evidence used, or the cognitive process that led to the result. Public criticism often isolates a curious or incorrect AI response while omitting the preceding interaction, including the original question, supplied information, intermediate corrections, verification attempts, and the human decision to accept the output.
This contribution reformulates responsibility for the specific work of the Laboratory of Information Systems and the Laboratory of Military Ecotones. The relevant unit of analysis is the expert–AI system. It includes the expert’s formulation of the problem, the evidence introduced, the AI system and version, the alternatives generated, the corrections made during the interaction, the verification procedure, and the final expert decision. Cognitive work may be joint, but final accountability remains human because a person decides whether a result will be accepted, published, or used.
The principal object of responsibility is the real problem of war-related environmental contamination. Institutional requirements, professional recognition, and career considerations may affect scientific work, but they cannot substitute for efforts to recognize the environmental process and improve decisions. A correct investigation of an isolated fragment is not condemned as irresponsible; it is classified as partially completed work when the relationships necessary to understand the wider process remain unexamined.
The proposed standard distinguishes five levels: personal cognitive responsibility, epistemic responsibility, process responsibility, decision responsibility, and practical responsibility. Personal cognitive responsibility begins with the expert’s genuine interest in understanding the object. Epistemic responsibility requires separation of observation, interpretation, hypothesis, and verified fact. Process responsibility makes the cognitive path reconstructable. Decision responsibility identifies the expert who accepts the conclusion. Practical responsibility evaluates the contribution to recognizing or resolving the problem.
Documentation occurs through an archived interaction record and a concise Logical Record of Expert–AI Problem Solving. The logical record identifies the problem, evidence, assumptions, AI configuration, alternatives, expert corrections, verification method, remaining uncertainty, and final decision. Responsibility is therefore evaluated through an explicit and correctable cognitive process rather than through rhetorical fear of AI assistance.
Key points • Responsibility for AI-assisted research cannot be evaluated from an isolated AI response.
• The relevant analytical unit is the complete expert–AI decision process.
• The principal responsibility of the researcher is to the real scientific problem.
• A valid but isolated disciplinary result may constitute partially completed work rather than a complete explanation.
• A concise logical record makes AI-assisted scientific reasoning open to verification and correction.
Keywords artificial intelligence; scientific responsibility; expert–AI system; war-related environmental contamination; cognitive process; logical record; verification; expert judgement; research transparency; problem resolution
Main discussion question What information must be preserved to evaluate whether an expert–AI system has produced a transparent, testable, correctable, and problem-relevant scientific decision?
OJS publication link https://pollution-diseases-ojs.org/index.php/pd/article/view/77
Note. Analytical TracksIn addition to the main thematic areas, the conference programme will include several cross-cutting analytical tracks. These tracks will be defined during the preparation of the programme, based on the submitted abstracts and the emerging links between presentations.At the preliminary stage, abstracts may be assigned to provisional analytical tracks marked as AT-01 to AT-07. Final track titles will be announced after the Scientific Committee has reviewed the submitted materials.