LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915578829078528 |
|---|---|
| author | Maier, Benjamin F. Aslak, Ulf Fiaschi, Luca Rismal, Nina Fletcher, Kemble Luhmann, Christian C. Dow, Robbie Pappas, Kli Wiecki, Thomas V. |
| author_facet | Maier, Benjamin F. Aslak, Ulf Fiaschi, Luca Rismal, Nina Fletcher, Kemble Luhmann, Christian C. Dow, Robbie Pappas, Kli Wiecki, Thomas V. |
| contents | Consumer research costs companies billions annually yet suffers from panel biases and limited scale. Large language models (LLMs) offer an alternative by simulating synthetic consumers, but produce unrealistic response distributions when asked directly for numerical ratings. We present semantic similarity rating (SSR), a method that elicits textual responses from LLMs and maps these to Likert distributions using embedding similarity to reference statements. Testing on an extensive dataset comprising 57 personal care product surveys conducted by a leading corporation in that market (9,300 human responses), SSR achieves 90% of human test-retest reliability while maintaining realistic response distributions (KS similarity > 0.85). Additionally, these synthetic respondents provide rich qualitative feedback explaining their ratings. This framework enables scalable consumer research simulations while preserving traditional survey metrics and interpretability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_08338 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings Maier, Benjamin F. Aslak, Ulf Fiaschi, Luca Rismal, Nina Fletcher, Kemble Luhmann, Christian C. Dow, Robbie Pappas, Kli Wiecki, Thomas V. Artificial Intelligence I.2.7; J.4 Consumer research costs companies billions annually yet suffers from panel biases and limited scale. Large language models (LLMs) offer an alternative by simulating synthetic consumers, but produce unrealistic response distributions when asked directly for numerical ratings. We present semantic similarity rating (SSR), a method that elicits textual responses from LLMs and maps these to Likert distributions using embedding similarity to reference statements. Testing on an extensive dataset comprising 57 personal care product surveys conducted by a leading corporation in that market (9,300 human responses), SSR achieves 90% of human test-retest reliability while maintaining realistic response distributions (KS similarity > 0.85). Additionally, these synthetic respondents provide rich qualitative feedback explaining their ratings. This framework enables scalable consumer research simulations while preserving traditional survey metrics and interpretability. |
| title | LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings |
| topic | Artificial Intelligence I.2.7; J.4 |
| url | https://arxiv.org/abs/2510.08338 |