(Expected) Using AI to Detect Bots in Surveys
Can we use GPT to validate if the survey is done by human or AI?
LLM
Conjoint
About
Initial thoughts of detecting bots on surveys using LLM.
Initial Ideas
- Attempt to use LLM to distinguish between humans responses and bot responses.
- This survey can be as generic as possible, for example, the apple conjoint surveys.
- We perform the survey on human participants and via LLM. Then the results will be embedded by GPT models to generate corresponding digital values.
- Once we have the embedded models for both human and LLM respondents, we can use them as templates to check for our public survey results such as Mechanical Turk.