You Said it? How Mis- and Disinformation Tweets Surrounding the Corona-5G-conspiracy Communicate Through Implying Lynn de Rijk, Radboud University, Netherlands l.derijk@student.ru.nl ABSTRACT 2 METHODS This paper aims to investigate if implied meaning plays a role Using the dataset compiled for the MediaEval 2020 FakeNews in mis-/disinformation tweets and what linguistic cues might Task [3], a qualitative analysis of 130 tweets was performed, signal this. A qualitative analysis of 130 mis-/disinformation coding for direct and indirect SA’s in Atlas.ti (version 8.4.5). tweets regarding the corona-5G-conspiracy using Speech Act The coding process was iterative and additional codes were Theory, shows that often meaning is implied by leaving out added based on patterns found in the data, such as recurring coherence markers, putting the words in someone else’s linguistic features, like coherence markers (e.g., ‘no mouth through citing and ambiguous phrasing/punctuation. meaningful connectives’) and certain SA’s (e.g., ‘citing’). Some codes were mutually exclusive (such as ‘Indirect SA: None’ 1 INTRODUCTION with other Indirect SA’s), where other codes were not (e.g., Systems that automatically recognize mis-/disinformation are tweets were often coded for multiple direct SA’s, see Example challenged by certain basic communicative features, such as 1). Only tweets supporting a conspiracy (e.g., corona is a sarcasm or implied meaning. It is therefore relevant to find coverup for 5G deaths or 5G causes corona) were included in out if indirect communication plays a role in mis- the result section, as tweets with a different stance were /disinformation and how large this role is. Speech Act Theory rarely found (n = 7). Furthermore, only tweets that were not [1] can be used as a framework to index implied meaning. It part of a thread were analyzed, to ensure that the context did distinguishes three forces in every utterance: 1) locution, not differ in regard to one-to-many vs. one-to-one interaction. what is said literally, 2) illocution, what the utterance does The data was coded iteratively until a saturation point was and 3) perlocution, what happens as a result. For example, in: reached within these inclusion criteria (n = 90). The final code “Peter, you are standing on my foot”, the locutionary force is list can be found in Appendix 1. Indirect SA’s were only coded asserting this state of affairs. The illocution, would generally for when these were the primary communicative force. For be requesting that this Peter lifts his foot, now that he is made example, the tweet below was coded for ‘Indirect SA: aware. The perlocutionary act would then be that Peter concluding’, as the implied meaning is most likely the primary indeed places his foot elsewhere. meaning. Without the implication the assertions made are However, Micheal Geis argues that that the illocutionary force simply loose statements. of a Speech Act (SA) is mostly dependent on the context, not Anyone else curious about the majority of deaths in china seem linguistic cues [2]. Geis gives the example of a teacher asking a to be the same areas they rolled out their stand alone 5G just a student if they can solve a quadratic equation vs. the same couple months ago. Verry few deaths being reported in other question being asked by a fellow student. The first would areas in comparison. #5G #CoronavirusOutbreak #COVID19 count as a request for information (Do you need help?), where #5gamechanger (Example 1) the second is more likely a request for action (I don’t understand, help me.) The importance of context is what The user asks a question in the first sentence, evidenced by makes recognizing illocutionary force difficult for an the syntactic structure of the sentence (‘direct SA: asking’), automated system. even though they did not use punctuation. This is followed by When dealing with data from a platform such as Twitter, two assertions (‘direct SA: asserting’). The intended relation however, this becomes less of an issue, since in one-to-many between the three sentences is not made explicit through communication every member of the audience is addressed coherence markers such as meaningful connectives (‘no relatively equally and the general context is the same for each meaningful connectives’). The last sentence does have a lexical tweet (Twitter). This exploratory study aims to find if indirect cue phrase (phrases that show the relation between sentences SA’s play a role in mis-/disinformation tweets and if so, if or the attitude of the speaker, e.g., ‘in my opinion’ or, in this there are linguistic features identifiable that can capture tweet, ‘in comparison’), that shows the relation between areas indirect SA’s. the user wishes to point out. As the causal relation is not made explicit, the act of concluding that these assertions are causally related is an indirect SA. 3 RESULTS Copyright 2020 for this paper by its authors. Use permitted under Users seem to employ a couple of strategies to avoid outright Creative Commons License Attribution 4.0 International (CC BY 4.0). claiming there is a conspiracy, often communicating through MediaEval’20, December 14-15 2020, Online MediaEval’20, December 14-15 2020, Online M. Larson et al. implication (see Table 2). First, they often omit connectives, 4 DISCUSSION leaving the relation between sentences implicit (e.g., Example This study found that there are linguistic cues identifiable that 1 and Table 3). Second, they tend to cite others (Table 2), such capture indirect SA’s, such as omission of certain connectives as citing the headline of an article they then link to, or use and lexical cue phrases. A possible explanation for these other means to put a middleman between themselves and findings is that users might (subconsciously) try to what is said, as can be seen in Example 2: circumvent the forewarning effect. This effect has been studied I’ve been reading a few posts from ppl I know about how the extensively in psychology and suggests that forewarning is a #CoronavirusOutbreak is because of 5G trials and Wuhan was factor that causes resistance to persuasion [4]. Using the place that first rolled this out and therefore is seen to be connectives and lexical cue phrases helps reader used as a biologocal warfare weapon... 😬😬 #coronavirus comprehension in informative texts, but in persuasive texts (Example 2) can build up the reader’s resistance, since they recognize In this tweet, connectives and lexical cue phrases (because of, more easily when they are being persuaded [4]. Apart from therefore) are used to make author intent clear, but the user the omission of connectives and lexical cue phrases, the puts the words in the mouths of ‘ppl I know’. ambiguity of certain tweets also points to this explanation. Alternatively, the omission of certain words might also be a Table 1: Frequency of most used direct SA’s result of the affordances [5] of Twitter, where the limited Code (not mutually Number of tweets Percentage of the characters per tweet might incentivize users to leave out exclusive) (n = 90) dataset words they deem unnecessary. However, this does not seem Direct SA: to be the case, as one would expect that emojis would be used - asking 28 32.2% quite often, since they leave room for ambiguity while - asserting 69 76.7% simultaneously only taking up one character space. As seen in - citing 24 26.6% Table 1, this is not the case. Additionally, emojis are often used - describing 29 32.2% decoratively as well, such as arrows or bullet points, without making use of their potential for ambiguity. Furthermore, Table 2: Distribution of indirect SA’s (suggesting, citing linked article titles is an interesting practice with these concluding, inviting, describing) affordances in mind, as it leaves little room for the user’s own Code Number of tweets Percentage of view. The limited space Twitter affords gives weight to the (n=90) the dataset chosen citation, which in turn creates the implication that the Indirect SA 59 65.6% citation is important/true/relevant. No indirect SA 31 34.4% Lastly, leaving things ambiguous and citing others, could point to a user orientation to distance themselves from conspiracy- Table 3: Distribution of linguistic features thinking – a way for users to keep plausible deniability for their support of what is said. It should be noted though, that Code (not mutually Number of tweets Percentage of grammatical or punctuation ambiguity might also be a result exclusive) (n = 90) the dataset of users not being native English-speakers or simply No meaningful 52 57.7% inattentiveness or oversight by the user. connectives Similar to [6]’s findings on Indonesian hoax data, I found that Lexical cue phrases: SA Theory can be useful to analyze mis-/disinformation online - Opinion 26 28.9% - Relation 19 21.1% data. Where they focused on direct SA’s, finding that assertive, Emoji use 12 13.3% directive and expressive SA’s were most common. In this study, indirect SA were also considered, showing that communication is often indirect in mis-/disinformation Third, users employ ambiguous phrasing. This can be seen in tweets. Future research could compare these findings to non- Example 1, where the question is not clearly stated, since a conspiracy-tweets, to shed some light on which explanation question mark is omitted and instead an assertion follows provided here is more plausible and to show if the found immediately. It is thus phrased initially as a question, but the linguistic features might aid in distinguishing information question itself does not seem of much importance. In Example from mis-/disinformation. It would also be useful to see if 2, this strategy can also be seen in ‘😬😬’. It is left to the reader machine-learning might have already been able to distinguish to infer what the target of the emoji is; is it the entire prior between the two regardless of picking up on the indirect SA’s. statement (how awkward that people I know say these things) or only the part describing a possible relation between 5G and ACKNOWLEDGMENTS corona (there might be a relation between corona and 5G)? I thank Martha Larson for her input and critical eye. I also Depending on the interpretation, the meaning of the tweet thank John Keates and Zhengyu Zhao for help with pre- changes completely, even with regard to the user’s stance. processing. FakeNews: Corona virus and 5G conspiracy L. de Rijk REFERENCES [1] J. R. Searle, 1965. What is a Speech Act? In M. Black (Ed.), Philosophy in America, pp. 221-239. Allen and Unwin. [2] M. L. Geis, 1995. Speech acts and conversational interaction. Cambridge University Press. [3] K. Pogorelov, D. T. Schroeder, L. Burchard, J. Moe, S. Brenner, P. Filkukova and J. Langguth 2020. FakeNews: Corona Virus and 5G Conspiracy Task at MediaEval 2020. In Working Notes Proceedings of the MediaEval 2020 Workshop. [4] J. Kamalski, L. Lentz, T. Sanders and R. A. Zwaan, 2008. The Forewarning Effect of Coherence Markers in Persuasive Discourse: Evidence from Persuasion and Processing. Discourse Processes, 45(6), 545-579. [5] I. Hutchby, 2001. Technologies, Texts and Affordances. Sociology, 35(2), 441-456. [6] A. A. Iswara and K. A. Bisena, 2020. Manipulation and Persuasion through Language Features in Fake News. RETORIKA: Jurnal Ilmu Bahasa, 6, 26-32. 3