Simulating Strategic Reasoning: A Digital Twin Approach to AI Advisors in Decision-Making

Dinithi N. Jayasekara, Qian Huang

Keywords: Digital Twin, Strategic Reasoning, AI Advisors

Abstract: This study investigates the feasibility of constructing and evaluating AI digital twins as advisors in strategic decision-making. Phase 1 focused on modeling the reasoning of a senior strategist (Participant A) through structured interviews, curated datasets, and prompt-based interactions with multiple large language models (LLMs). Results show high fidelity on simple tasks but significant gaps in complex reasoning.This discrepancy highlights the limitations of current LLMs in replicating nuanced strategic reasoning. To address this, we propose an evaluation framework that balances both potential and limitations where Potential corresponds to high accuracy in simple decision-making, and Limitations reflect reduced performance in complex, multi-step reasoning.

Submission Number: 200