Promises and lies: can observers detect deception in written messages

Jingnan Chen, Daniel Houser

We design a laboratory experiment to examine predictions of trustworthiness in a novel three-person trust game. We investigate whether and why observers of the game can predict the trustworthiness of hand-written communications. Observers report their perception of the trustworthiness of messages, and make predictions about the senders' behavior. Using observers' decisions, we are able to classify messages as "promises" or "empty talk." Drawing from substantial previous research, we hypothesize that certain factors influence whether a sender is likely to honor a message and/or whether an observer perceives the message as likely to behonored: the mention of money; the use of encompassing words; and message length. We find that observers have more trust in longer messages and "promises"; promises that mention money are significantly more likely to be broken; and observers trust equally in promises that do and do not mention money. Overall, observers perform slightly better than chance at predicting whether a message will be honored. We attribute this result to observers' ability to distinguish promises from empty talk, and to trust promises more than empty talk. However, within each of these two categories, observers are unable to discern between messages that senders will honor from those that they will not.