Persuasion detection in conversation

Citation: Henry T. Gilbert, I. V. (2010) Persuasion detection in conversation. Naval Postgraduate School (RSS)
Internet Archive Scholar (search for fulltext): Persuasion detection in conversation
Download: http://edocs.nps.edu/npspubs/scholarly/theses/2010/Mar/10Mar Gilbert.pdf
Tagged: Computer Science (RSS) persuasion (RSS), online argumentation (RSS), negotiation (RSS)

Summary

This Master's thesis draws from Cialdini's 6 key principles of persuasion in order to create an annotated corpus of persuasion.

Availability

The corpus is available from Dr. Joel D. Young, Naval Postgraduate School Department of Computer Science.

Annotation

The corpus was drawn from 37 police transcripts, which three annotators tagged, in two rounds. In the first annotation, each turn was treated as an utterance. Persuasive elements--reason, reciprocity, commitment and consistency, social proof, scarcity, liking, authority, and scarcity--were tagged. "If there were multiple kinds of persuasion in an utterance, annotators were asked to rank them in order of importance."

Cohen's Kappa was used to evaluate agreement. Two main problems arose in the first round: "reason" was difficult to annotate and subsequently removed. Meanwhile, an "other" category was added, and category definitions were improved. Further, hostage taggers (not just police negotiators) made persuasive utterances; in the second round these were more consistently tagged.

The second round used improved examples and category examples (see pages 36-44).

Results

In transcripts, only 5-20% of the utterances were persuasive.

Selected References

• Cialdini, R. B. (2001). Influence: Science and practice. Boston: Allyn and Bacon.
• Cialdini, R. B., Vincent, J. E., Lewis, S. K., Catalan, J., Wheeler, D., & Darby, B. L. (1975). Reciprocal concessions procedures for inducing compliance: The door-in-the-face technique. Journal of Personality and Social Psychology, 31 , 206-215.
• Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement Vol.20, No.1 , 37-46.
• Lin, W.-H., Wilson, T., Wiebe, J., & Hauptmann, A. (2006). Which side are you on? Identifying perspective at the document and sentence levels. Proceedings of the Conference on Natural Language Learning (CoNLL).

Theoretical and Practical Relevance

Chapter 2 provides powerful examples of the principles of persuasion, drawn from police negotiation transcripts and social psychology experiments.

The author suggests that the corpus could be used to train machine learning algorithms for persuasion detection; see his colleague's master's thesis, Machine learning techniques for persuasion detection in conversation which did that.

Additional suggestions are to look for correlations between dialogue acts and persuasion attempts.