Anfonwch hwn fel neges destun: Can large language model agents simulate human trust behaviors?