Understanding of linguistic politeness

Communication is an act of cooperation between two or more partners (Grice, 1975), and people normally assume that their communicators would speak in a way that is maximally truthful and informative. Imagine two people in a conversation, Tom and Sally, who are talking about pets they own. Tom says to Sally: “I have two dogs.” Assuming that Tom is a cooperative speaker, and that Sally also believes that he is, Sally will imagine the the true state of the world is such that Tom owns at least two dogs, which makes Tom’s utterance truthful. Furthermore, she will infer that Tom owns two dogs, but not three or more, assuming Tom is informative and would have used a greater number term otherwise. Now imagine the following situation: Tom and Sally are talking about a presentation that Sally just gave, and Sally asked for Tom’s opinion on her talk. Tom says to Sally: “It was okay.” We can apply the same analysis as the previous one in this situation: if we assume Tom has some scale on his mind that ranges from ‘terrible’ to ‘excellent,’ and the term ‘okay’ covers the positive part of the scale starting from the midpoint to the ‘excellent’ end, then we might infer that Tom meant to indicate that Sally’s presentation was “above average but not excellent.” But is that the farthest we can infer? For example, if Sally’s performance was exactly average, would Tom necessarily use the word ‘average’ instead of ‘okay’ to be more informative and exact? Is it also possible that Tom used the word ‘okay’ even though Sally’s performance was actually below average? In that case, should we conclude that Tom was not a cooperative speaker? But people do share a strong intuition to maximize praise for their conversational partner, even by telling white lies and compromising truthfulness (DePaulo & Kashy, 1998). Even children show tendency to tell white lies and think of white lies as better than malicious lies or even truths in situations in which the listener is harmed by truth-telling (Talwar, Murphy, & Lee, 2007; Bussey, 1999). Past literature has explained this phenomenon of ‘linguistic politeness’ in terms of people’s want to maintain each other’s ‘face’: every human’s wants to maintain positive public selfimage and freedom from impositions. Going back to the example above, speakers maximize praises for the listener in order to save the listener’s positive face, or self-esteem. Leech (1983) furthered this idea and proposed Politeness Principle (PP), which includes Maxim of Approbation: speakers must maximize praise and minimize dispraise for the listener. Despite the rich literature on related phenomena of linguistic politeness observed across different situations and cultures, there has thus far been few formal models for the common phenomenon of linguistic politeness. The goal of the current paper is to propose a generative model that looks at mechanisms of polite utterance generation and of listeners’ inferences based on the utterances about speakers’ underlying goals.