This study examines the feasibility of utilizing artificial intelligence AI-based virtual surveys in Choice Experiments to analyze consumers’ attribute values for alternative meat, aiming to complement the high time and economic costs associated with traditional human-subject surveys. Focusing on alternative meat attributes including production method, quality grade, country of origin, and price, we compare choice behaviors and marginal willingness to pay estimates across three response types: human responses to a human-centric design, AI responses to the same design, and AI responses to a large-scale AI-specific design. Multinomial logit estimation reveals that AI responses reliably replicate the directional hierarchy of human preferences, consistently favoring conventional livestock, domestic origin, and higher quality. However, a statistical discrepancy emerges in valuation tasks for less-preferred attributes, indicating that human risk aversion may not be fully captured in the AI data. These findings suggest that while AI-based virtual surveys can be useful for pre-testing experimental designs and identifying preference rankings, their application to actual monetary valuation tasks requires careful calibration and cautious interpretation to prevent data distortion.

Abstract
Introduction
Materials and Methods
    1. Analytic model
    2. Survey design
Results and Discussion
    1. Estimation results
    2. Discussion and implications
References

• Dong Hyun Koo(Master’s student, Department of Agricultural Economics, Kyungpook National Univ., Daegu 41566, Korea)
• Se Hyuk Kim(Research Professor, Kyungpook Anchor Business Division; Researcher, Institute of Regional Development, Kyungpook National Univ., Sangju 37224, Korea)
• Seung Gyu Kim(Professor, Department of Food and Resource Economics; Board member, Institute of Regional Development, Kyungpook National Univ., Daegu 41566, Korea) Corresponding author