Abstract
This paper introduces thought experimentation with LLM agents as a method for theory elaboration. The method helps render the logical implications of a verbally articulated theory as observable patterns within a controlled computational environment. It elaborates Dimov and Güneştepe's (2024) theory of opportunity capital by taking advantage of the narrative competence of recently developed LLMs (Ramoglou et al., 2026). The thought experiment operationalises central elements of the theory: opportunity capital as dialogical calibration between entrepreneur and investor, and the persistence of a commitment-legitimacy gap. It also reveals model sensitivity in the method, with different trajectories produced when Claude Sonnet and Gemini Flash models enact the entrepreneur, investor, and evaluator roles. Computational thought experimentation opens new avenues for elaborating theory where core mechanisms are dialogical, processual, and difficult to observe in practice.
| Original language | English |
|---|---|
| Article number | e00635 |
| Number of pages | 9 |
| Journal | Journal of Business Venturing Insights |
| Volume | 25 |
| Early online date | 26 May 2026 |
| DOIs | |
| Publication status | Published - 30 Jun 2026 |
Data Availability Statement
Code will be made publicly available.ASJC Scopus subject areas
- Business and International Management
- Management of Technology and Innovation
Fingerprint
Dive into the research topics of 'Giving texture to theory: A computational thought experiment on opportunity capital with LLM agents'. Together they form a unique fingerprint.Cite this
- APA
- Standard
- Harvard
- Vancouver
- Author
- BIBTEX
- RIS