Adaptive Self–Human Selection: A Unified Framework for Dynamic Switching Between Autonomous Model Selection and Human Selection

Large language models and artificial intelligence systems increasingly perform selection-related functions, including selecting answers, candidates, strategies, actions, resources, recommendations, solutions, and decision alternatives. Such selection may be performed autonomously through a model's internal mechanisms or directly through human judgment and intervention. Although existing research has investigated automated decision-making, algorithmic recommendation, selective prediction, human-in-the-loop decision-making, human–AI collaboration, and adaptive delegation, these approaches frequently focus on particular decision tasks, interaction arrangements, or domain-specific selection mechanisms. A general, mechanism-independent architecture that treats autonomous model selection and human selection as distinct but dynamically combinable operational configurations across arbitrary selection functions remains insufficiently formalized. This paper introduces Adaptive Self–Human Selection (ASH-Select), a unified theoretical framework grounded in the Universal Self–Human Model (USHM). The framework establishes three fundamental operational configurations: Self-Selection (S), in which an artificial intelligence system independently selects among available alternatives through its internal mechanisms; Human Selection (H), in which a human individual or group performs the selection; and Self–Human Selection (S–H), in which model-based and human-based selection mechanisms contribute to the same selection process. Unlike architectures that permanently assign selection authority to either an artificial system or a human operator, ASH-Select introduces adaptive configuration selection, dynamic mechanism switching, complementary selection, and structural reconfiguration. The proposed architecture distinguishes the selection of an object or alternative from the higher-level selection of the mechanism responsible for performing that selection. It further develops a context-sensitive allocation model incorporating selection uncertainty, candidate complexity, model capability, human expertise, expected decision quality, operational cost, risk, reversibility, and authority constraints. A formal representation is presented for candidate evaluation, selection policies, mechanism transitions, joint selection, and adaptive responsibility allocation. The framework also establishes theoretical propositions, application scenarios, governance requirements, and an empirical evaluation methodology. To the best of our knowledge, this paper is the first to formally articulate the specific mechanism-independent, function-general architecture developed here that unifies autonomous model selection and human selection through adaptive selection, switching, combination, and reconfiguration. This priority claim concerns the particular theoretical integration and architecture presented in this paper, rather than the broader established fields of automated decision-making, algorithmic selection, human-in-the-loop systems, or human–AI collaboration. The proposed framework provides a conceptual foundation for intelligent systems that dynamically determine not only what should be selected, but also whether the selection should be performed by an artificial intelligence system, a human, or a coordinated combination of both.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24359938
Primary Topic
Human-Automation Interaction and Safety
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Adaptive Self–Human Selection: A Unified Framework for Dynamic Switching Between Autonomous Model Selection and Human Selection

Jincheng Zhang
Zenodo (CERN European Organization for Nuclear Research)
Human-Automation Interaction and Safety
preprint

Adaptive Self–Human Selection: A Unified Framework for Dynamic Switching Between Autonomous Model Selection and Human Selection

Jincheng Zhang
preprint en

Abstract

Large language models and artificial intelligence systems increasingly perform selection-related functions, including selecting answers, candidates, strategies, actions, resources, recommendations, solutions, and decision alternatives. Such selection may be performed autonomously through a model's internal mechanisms or directly through human judgment and intervention. Although existing research has investigated automated decision-making, algorithmic recommendation, selective prediction, human-in-the-loop decision-making, human–AI collaboration, and adaptive delegation, these approaches frequently focus on particular decision tasks, interaction arrangements, or domain-specific selection mechanisms. A general, mechanism-independent architecture that treats autonomous model selection and human selection as distinct but dynamically combinable operational configurations across arbitrary selection functions remains insufficiently formalized. This paper introduces Adaptive Self–Human Selection (ASH-Select), a unified theoretical framework grounded in the Universal Self–Human Model (USHM). The framework establishes three fundamental operational configurations: Self-Selection (S), in which an artificial intelligence system independently selects among available alternatives through its internal mechanisms; Human Selection (H), in which a human individual or group performs the selection; and Self–Human Selection (S–H), in which model-based and human-based selection mechanisms contribute to the same selection process. Unlike architectures that permanently assign selection authority to either an artificial system or a human operator, ASH-Select introduces adaptive configuration selection, dynamic mechanism switching, complementary selection, and structural reconfiguration. The proposed architecture distinguishes the selection of an object or alternative from the higher-level selection of the mechanism responsible for performing that selection. It further develops a context-sensitive allocation model incorporating selection uncertainty, candidate complexity, model capability, human expertise, expected decision quality, operational cost, risk, reversibility, and authority constraints. A formal representation is presented for candidate evaluation, selection policies, mechanism transitions, joint selection, and adaptive responsibility allocation. The framework also establishes theoretical propositions, application scenarios, governance requirements, and an empirical evaluation methodology. To the best of our knowledge, this paper is the first to formally articulate the specific mechanism-independent, function-general architecture developed here that unifies autonomous model selection and human selection through adaptive selection, switching, combination, and reconfiguration. This priority claim concerns the particular theoretical integration and architecture presented in this paper, rather than the broader established fields of automated decision-making, algorithmic selection, human-in-the-loop systems, or human–AI collaboration. The proposed framework provides a conceptual foundation for intelligent systems that dynamically determine not only what should be selected, but also whether the selection should be performed by an artificial intelligence system, a human, or a coordinated combination of both.

Zenodo (CERN European Organization for Nuclear Research)
Human-Automation Interaction and Safety
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon