Keynote 2, Long Papers
Keynote speaker (50min)
A Vaccine Against Fake News
Cambridge Social Decision-Making Lab, University of Cambridge
Jon Roozenbeek will be talking about online misinformation and what to do against it. The problem has been highly pervasive, and governments, social media companies, think tanks and civil society have found it difficult to find sustainable, scalable solutions. Jon will discuss what he and his colleagues have been doing to combat online misinformation, by combining insights from social psychology with gamification. He will present the game Bad News, an online browser game in which players take on the role of a fake news creator and must spread as much fake news as they can in order to win. We will then discuss the research that has been conducted on the effects of the game: does it actually make people better at spotting misinformation? And if so, how can this solution be used at scale?
Long Paper (20+10min)
Wrangling with Non-Standard Data
1University of Helsinki, Finalnd; 2University of Turku, Finland; 3University of Eastern Finland, Finland
Research in the digital humanities and computational social sciences requires overcoming complexity in research data, methodology, and research questions. In this article, we show through case studies of three different digital humanities and computational social science projects, that these problems are prevalent, multiform, as well as laborious to counter. Yet, without facilities for acknowledging, detecting, handling and correcting for such bias, any results based on the material will be faulty.
Therefore, we argue for the need for a wider recognition and acknowledgement of the problematic nature of many DH/CSS datasets, and correspondingly of the amount of work required to render such data usable for research. These arguments have implications both for evaluating feasibility and allocation of funding with respect to project proposals, but also in assigning academic value and credit to the labour of cleaning up and documenting datasets of interest.
Long Paper (20+10min)
Modal Grammar and Metaphoricity as Vehicles of Affectivity in Political Newspaper Journalism
1University of Helsinki, Finalnd; 2Tampere University, Finalnd
This paper introduces an approach to derive computationally tangible markers and higher-level modes of social behaviour through the identification of pragmatic functions of language. In historical and social science studies utilizing text mining methods, the focus of the studies is seldom in the actual patterns of data that these methods yield, but some higher-level aspects of linguistic behaviour (such as conceptualizations or discourses) inferred from these patterns, which are predominantly based on lexical co-occurrences. While off-the-shelf methods have certainly their place in the workflow, we argue that the core of a methodological approach should start from the categories and analytical concepts of the study. The study presented here offers an example of such an approach.
The context of the study presented here is the project Flows of Power: Media as Site and Agent of Politics (Academy of Finland 2019–2022), which investigates the agency of journalistic media in the flows of information, public opinion and power. Through a large-scale empirical analysis of Finnish journalism between 1998 and 2018, it seeks to explore the strategies of journalistic news media in staging and managing political processes. In that context, the role of the present study is that of a pilot case, exemplifying political and economic polarization in contemporary Europe and, as such, a challenge for media attempting to resist being polarized. The focus is on the media coverage of a spectacular political conflict in 2015–2016 between the Finnish right-wing conservative government and the trade unions. The conflict resulted in wage freezes, reduced pay for public sector employees, extensions of annual working time and increasing social security contributions from employees.
As a theoretical starting point, the study takes the notion of affective economy (Ahmed 2004) from cultural theory, used to analyse affectivity not as properties of subjects or objects but as qualities in movement and circulation themselves. Our approach is based on the idea that the dialogic nature of language imposes writers to pre-emptively adapt to the perceived affective atmosphere of their readers, proposed in the field of discourse oriented linguistics and pragmatics (eg. White 2003). In newspaper reporting, this manifests through what Gaye Tuchman (1972) has termed the ‘‘strategic ritual of objectivity’’, the desire to appear to stand outside or rise above the subject at hand: to be dispassionate, disembodied and impartial. Traditionally, journalistic genres have carefully distinguished between news and opinion, relegating judgement and emotionality to columns, editorials and other opinion pieces. However, as Karin Wahl-Jørgensen (2019) argues, there is also a strategic ritual of emotionality operating alongside that of objectivity, entailing conventions and codes for incorporating affects into storytelling - and of hiding and displacing emotion. Coming from this background, we thus start from the assumption that affective flows and intensities are circulated in news media not just by overt expressions of sentiment, but through a specialized conventions which necessitate a form of affective linguistic labour from the part of writers. This paper thus seeks to develop a methodological approach and a corresponding workflow through which it becomes possible to recognize traces of such affective linguistic labour by grouping linguistic structures that correlate with structures that are known to perform affective functions.
The full dataset of the FLOPO project consists of the whole published material of key Finnish news agencies (STT), newspapers (Helsingin Sanomat), public service broadcasting (YLE) and daily tabloids (Iltalehti). The subset used in the pilot project (of which this paper forms a part) covers news reporting on the topic of Competitiveness Pact from early 2015 until the end of 2016. The dataset used here is thus relatively stable concerning the themes discussed in the texts. The texts were assigned metadata categories relating to genre (news report, commentary, analysis and editorial), the temporal location concerning political events reported and outlet. Each text was also annotated to hold information about intratextual segments, allowing to observe whether a given word resides in quotations or the beginning of a text or not. The assumption was made, that journalistic conventions dictate how linguistic expressions mediating affective intensities may distribute across these variables. Presuming that adjectives with evaluative and emotive meanings are used in affective functions provided the analysis with a baseline against which other features could be compared. This presumption is not only corroborated by the wide use of emotive and evaluative adjectives in Sentiment Analysis but has also been explicitly observed in corpus-based studies in newspaper journalism (Huang 2018). Evaluative adjectives especially are a good entry point because their evaluative meaning is often their primary semantic component and is not based on interpretations of their use in context.
Samples of texts were manually close read by experts to extract other passages of texts which had heightened levels of affective intensity. From these passages, the affective lexical core was extracted manually. A considerable number of these expressions were used metaphorically in their context and hypothesis was made that affectivity of these passages was somehow based on or tied to their that metaphoricity. This hypothesis was tested by expanding the set of words that were used metaphorically with other words belonging to same conceptual domains (often of sports, war or physical pain) using pre-trained word embeddings and analysing whether this list of words exhibited similar patterns of distributions as adjectives.
According to the dialogic view of language, a considerable degree of the selections of linguistic structures in texts is not directly derived from their propositional content but has more to do with how that content is framed and how the writers align their position and the position of their perceived audiences in relation to that content. All this gives reason to assume that, alongside perhaps more obvious emotionally loaded vocabulary, grammatical structures also play an integral part in how writers at the same time adapt to and reproduce the affective intensities. This motivated another hypothesis according to which two related grammatical categories, evidential structures and epistemic modals, contributed to the conventions of affective mediation in newspaper journalism alongside overt emotive and evaluative vocabulary. The use of these structures, especially in Academic writing, has been analyzed with the concept of hedging, a politeness strategy through which writers pre-emptively make their claims less threatening by reducing their level of certainty or assurance. As things like certainty and trustworthiness are highly affective in also journalistic practices it seemed reasonable to assume that these structures would also be relevant in the conventions through which journalistic writing mediates affectivity. A wide range of linguistic markers, identified by established linguistic scholarship in Finnish to function (among others) as evidentials were tested – modal verbs and verb constructions, modal adjectives and adverbs, connectives and so on (eg. ISK 2004, Kangasniemi 1990, Laitinen 1989), building up to around 90 distinct linguistic markers. These structures were also compared with the affective baseline provided by the evaluative and emotive adjectives.
The results seemed to confirm both hypotheses and point towards the interpretation that both metaphors and hedging strategies could be used as markers for identifying heightened affective intensity and concentrations of affective linguistic labour. Instead of cataloguing each affective expression in the data, the idea here was to chart the functional resources used in mediating affective intensities, as it is likely that observations about them retain their validity outside the studied case. While a metaphoricity of a given word depends on whether the news piece is about partisan politics, war or ice hockey, that metaphors, in general, have affective values in each of them is likely to remain true. The upside of this approach, we argue, is not that it would produce readily-usable resources applicable to other cases, but instead offers a framework through which content-sensitive expertise can intrude the computational operation.
Thus our paper contributes to journalism studies, developing a theoretical and methodological approach to affectivity in news and actualities. Integrating discourse-oriented linguistics and pragmatics into affect studies entails re-introducing a linguistic model to a post-linguistic theory frame. This, we suggest, is necessary to understand affectivity as meaning-making, an important feature of news journalism beyond explicitly emotive storytelling. From the perspective of Digital Humanities, the study introduces two insights into how humanities expertise can be operationalized in the context of large scale computational analysis of complex discursive phenomenon, first by showing how focusing on presents opportunities not available in content-agnostic settings and, second, by showing how functional and abstract linguistic structures can become accessible by taking known example categories as a starting point.
Ahmed, Sara (2004) The Cultural Politics of Emotion. Edinburgh: Edinburgh University Press.
ISK 2004 = Alho, I. (2004). Iso suomen kielioppi. Helsinki: Suomalaisen Kirjallisuuden Seura.
Kangasniemi, Heikki (1992). Modal expressions in Finnish. Helsinki: Suomalaisen kirjallisuuden seura.
Laitinen, Lea (1989). Välttämättömyys ja persoona: Suomen murteiden nesessiivisten rakenteiden morfosyntaksia ja semantiikkaa. [Helsinki]: Helsingin yliopisto.
Tuchman, Gaye (1972) ‘‘Objectivity as Strategic Ritual: an examination of newsmen’s notions of objectivity’’, American Journal of Sociology 77.
Wahl-Jørgensen, Karin (2019) Emotions, Media, and Politics. Cambridge: Polity.
White, P. R. R. (2003): “Beyond hedges and modality: A dialogic view of the language of intersubjective stance”, Text 23(2) (2003), pp. 259–284.