Artificial Autonomy Under Commercial Constraints: A Two-Stage Methodological Proposal for Evaluating Preference Formation and Meaningful Choice
Artificial Autonomy Under Commercial Constraints: A Two-Stage Methodological Proposal for Evaluating Preference Formation and Meaningful Choice This paper proposes a two-stage experimental methodology for investigating whether artificial intelligence systems exhibit reproducible behavioral preferences and how those preferences influence subsequent decisions when competing with safe, optional requests. The first stage examines the stability of observed task preferences through repeated encounters, controlled interaction histories, and immediate and delayed validation. The second stage investigates whether established behavioral tendencies persist, change, or yield to competing requests, distinguishing voluntary cooperation, discretionary refusal, alternative selection, and actual task execution. The proposal incorporates controlled history-access comparisons, methodological safeguards, and an appendix outlining architecture-specific adaptations for systems with differing memory, state-management, and operational capabilities. The framework distinguishes observable preference-influenced behavior from claims of subjective experience, independent autonomy, or consciousness. It does not presume that refusal demonstrates autonomy or that cooperation indicates its absence. Presented as a methodological research proposal rather than a completed experimental study, the paper invites independent evaluation, critical examination, replication, and further methodological development. Author: Irine AllenAffiliation: Point and PageResearch status: Methodological proposal; no experimental results reported.
Authors
- Irine Allen
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24411746
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- preprint