Emoji semantics/pragmatics: investigating commitment and lying

Benjamin WeissmanNAACL 2022

Sure, here's a blog-style summary of the academic paper "Emoji semantics/pragmatics: investigating commitment and lying":

TL;DR

This paper explores whether emojis can convey speaker commitment and be used to lie, finding that emojis can indeed function this way, but their effectiveness depends heavily on their meaning agreement and conventionality. Emojis with high agreement, like concrete objects, are more likely to commit speakers and enable lying, while expressive emojis show more variability.

What Problem Does This Paper Address?

While emoji use is ubiquitous, the precise nature of their semantic and pragmatic contributions remains underexplored. Specifically, this research investigates two key aspects: first, whether speakers are committed to the meanings they convey through emojis, and second, whether it's possible to lie using emojis. Understanding these aspects is crucial for a comprehensive grasp of how emojis function in communication, especially as they become increasingly integrated into digital discourse.

Methods

The study employed a two-experiment design.

  • Experiment 1 (Commitment and Deniability): Participants were presented with question-answer pairs where the answer was a single emoji. They then provided two ratings on a 1-7 scale: one for direct commitment ("Have I committed to saying X?") and another for deniability ("Could I convincingly deny that I said X?"). These ratings were combined into a single commitment score.
  • Experiment 2 (Lying): This experiment built upon Experiment 1 by adding context to the question-answer emoji pairs. This context established whether the emoji response was true or false, and participants rated how likely it was that the response constituted a lie. A comparison was also made between emoji and word responses in this lying context.

Both experiments utilized pre-tested emoji sets, categorized into "expressive" (facial expressions, gestures) and "non-expressive" (objects, foods, animals), with varying degrees of meaning agreement.

Datasets

The study utilized pre-tested emoji meaning agreement data collected from two surveys. These surveys involved participants typing the meaning of various emojis, and the results were used to calculate an "agreement proportion" for each emoji. From these, sets of expressive and non-expressive emojis were selected for the experiments, with a focus on those exhibiting a range of meaning agreement.

Key Insights and Ideas

  • Emojis can commit speakers: The research strongly suggests that emojis can indeed commit speakers to the meanings they convey, similar to how verbal utterances do.
  • Lying with emojis is possible: The study provides evidence that emojis can be used to deceive, fulfilling the criteria for lying when a false meaning is intended.
  • Meaning agreement is key: The degree to which an emoji's meaning is consistently understood across a population (meaning agreement) is a critical factor in its ability to convey commitment and enable lying.
  • Non-expressive emojis are more direct: Emojis representing concrete objects or foods tend to have higher meaning agreement and, consequently, lead to higher commitment and lie ratings.
  • Expressive emojis are more nuanced: Emojis depicting facial expressions or emotional states show greater variability in meaning agreement and thus exhibit a wider range of commitment and lie attributions.
  • Conventionality matters: Emojis that are more conventionalized or closer to being "lexicalized" (having a fixed, agreed-upon meaning) are more effective in establishing speaker commitment and facilitating deception.

Key Results

  • Experiment 1: Non-expressive emoji yielded significantly higher commitment ratings than expressive emoji. Furthermore, for expressive emoji, higher pre-test meaning agreement correlated with higher commitment ratings.
  • Experiment 2: Lie ratings for emoji were significantly lower than for words. This difference was more pronounced for expressive emoji compared to non-expressive emoji. Crucially, for expressive emoji, higher commitment ratings from Experiment 1 led to higher lie ratings in Experiment 2 when the context indicated a false response.

Limitations and Future Work

The authors acknowledge that the context-sensitivity of emoji interpretation is a significant area for future research, as context heavily influences both lie ratings and emoji understanding. They also suggest exploring multi-emoji sequences and their pragmatic functions. The paper proposes that further work could lead to a taxonomy of semantic/pragmatic commitment across different modalities, and the authors envision real-world scenarios, like emoji-related scandals, as fascinating case studies for examining deniability and commitment.

Summary updated

BibTeX
@inproceedings{weissman2022emoji,
  author = {Benjamin Weissman},
  title = {{Emoji semantics/pragmatics: investigating commitment and lying}},
  booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2022},
}

Loading related content...

Figure