Detecting Deception Strategies [Godsend for the 2016 Election Cycle]

Discriminative Models for Predicting Deception Strategies by Scott Appling, Erica Briscoe, C.J. Hutto.


Although a large body of work has previously investigated various cues predicting deceptive communications, especially as demonstrated through written and spoken language (e.g., [30]), little has been done to explore predicting kinds of deception. We present novel work to evaluate the use of textual cues to discriminate between deception strategies (such as exaggeration or falsifi cation), concentrating on intentionally untruthful statements meant to persuade in a social media context. We conduct human subjects experimentation wherein subjects were engaged in a conversational task and then asked to label the kind(s) of deception they employed for each deceptive statement made. We then develop discriminative models to understand the difficulty between choosing between one and several strategies. We evaluate the models using precision and recall for strategy prediction among 4 deception strategies based on the most relevant psycholinguistic, structural, and data-driven cues. Our single strategy model results demonstrate as much as a 58% increase over baseline (random chance) accuracy and we also find that it is more difficult to predict certain kinds of deception than others.

The deception strategies studied in this paper:

  • Falsification
  • Exaggeration
  • Omission
  • Misleading

especially omission, will form the bulk of the content in the 2016 election cycle in the United States. Only deceptive statements were included in the test data, so the models were tested on correctly recognizing the deception strategy in a known deceptive statement.

The test data is remarkably similar to political content, which aside from their names and names of their opponents (mostly), is composed entirely of deceptive statements, albeit not marked for the strategy used in each one.

A web interface for loading pointers to video, audio or text with political content that emits tagged deception with pointers to additional information would be a real hit for the next U.S. election cycle. Monetize with ads, the sources of additional information, etc.

I first saw this in a tweet by Leon Derczynski.

Comments are closed.