Visible scientists in digital communication environments: An analysis of their role performance as public experts on Twitter/X during the Covid-19 pandemic

Abstract

In response to significant societal challenges, there is a growing demand for scientists to actively engage in public discussions. The recent Covid-19 pandemic led to the sudden visibility of certain scientists, necessitating them to extend their roles beyond research and actively communicate with the general public. Online platforms allow for direct engagement but increase the challenge by interconnecting different public arenas. Our case study examined the role performance of visible virologists on Twitter/X in six different time periods during the pandemic in Germany (N = 1003). Findings indicate that they often express calls to action, and voice their own views, but seldom address uncertainty. Differences over time were found in their use of different types of statements, scientific jargon and emotional language. They also utilised the opportunities of direct communication, performing roles like watchdogs or advocates, highlighting the importance for scientists to reflect on their roles as communicators.

1. Introduction

Given the many challenges societies are currently facing, there is a growing demand for scientists to directly engage with different non-scientific actors (Calice et al., 2023). Hence, in recent years, there has been increased attention to scientists as public communicators (Dudo and Besley, 2016). While scientists often remain invisible, there are some exceptional situations (e.g. the Covid-19 pandemic) in which they become visible to the broader public (Joubert et al., 2023; Peters, 2021). Due to their prominence, those scientists can shape public discussions (Fahy, 2017; Joubert et al., 2023). Although the term ‘visible scientist’ originally referred to scientists who were prominent in mass media (Goodell, 1977), scientists today can also become publicly visible via social media (Olesk, 2021). Social media have become significant platforms for disseminating scientific information and engaging with the public (Su et al., 2021), to which laypeople have increasingly turned for scientific information (WiD, 2023). However, few studies have directly considered visible scientists’ online communication in public discourses on socio-scientific issues (Sadler et al., 2007), which are real-world problems of social relevance and controversiality informed by science (Taddicken and Krämer, 2021).

Both the rise of digital communication environments and socio-scientific issues might ‘shape and challenge professional roles and norms that underlie their communication practices’ (Brüggemann et al., 2020: 3). We understand roles as expectations, attitudes and behaviours guiding individuals in certain situations when performing a specific task (Turner, 2001). There is a lack of empirical research on scientists’ roles in society. Previous studies are mostly limited to analyses of scientists’ self-reported role perceptions (e.g. Horst, 2013; Roedema et al., 2021), whereas research on (visible) scientists’ role performance in digital communication environments is scarce (e.g. Biermann et al., 2023; Walter et al., 2017). As scientists have the possibility to bypass mass media’s gatekeeper function, social media has changed how scientists can communicate with non-scientific actors (Roedema et al., 2021). Therefore, a significant question in the field of science communication is how visible scientists communicate on those platforms and – as very different social media platforms exist, bringing together different audiences, using different functionalities and communication patterns – at the boundaries of different online public arenas (Lörcher and Taddicken, 2017).

In our case study, we aim to provide deeper insights into visible scientists’ role performance as public experts in digital communication environments during the Covid-19 pandemic in the German context. Against the pandemic’s backdrop, we can explore how suddenly visible scientists communicate on social media during a global crisis garnering immense public attention. Scientists unexpectedly had to transgress their roles as researchers and simultaneously deal with the public, the media and politicians (Peters, 2021). This applied particularly to virologists, who also communicated through social media channels (Utz et al., 2022). To examine their role performance as public experts, we chose the microblogging platform Twitter/X, which recently changed its name to ‘X’, as it has been widely used by scientists (Zhang and Lu, 2022), and has enabled them to communicate in new ways with various actors (Brossard and Scheufele, 2022).

2. Literature review

Scientists engaging in digital communication environments

Recent developments on social media platforms such as Twitter/X have demonstrated the benefits of overcoming the rapid transience of platform-related research. A theoretical approach that seeks to broaden the platform-specific perspective is the concept of online public arenas (Schmidt, 2013). Here, online public arenas are seen as a combination of situational practice and supra-situational structuredness. The underlying software architecture (e.g. inherent affordances) affects communication behaviour in several ways (Schmidt, 2013). Thus, the constitution of online public arenas depends on communication technologies as well as communication modes, and different arenas exist with different communication barriers (e.g. level of expertise), inherent norms, intended audiences or communication goals (Lörcher and Taddicken, 2017; Schmidt, 2013). Following this, online arenas can be differentiated, such as the expert arena, mass media arena, or discussion arena, where, for example, experts discuss within their scientific community, journalists explain to the public, or citizens share their concerns. Forms of presentation and information vary depending on the arena (Lörcher and Taddicken, 2017; Schmidt, 2013).

The interconnection and integration of different public spheres on social media can lead to an intertwining of different online public arenas. On social media, communication can be quickly transferred from one online public arena to another online public arena, creating a complex network of intertwining arenas (Schrape, 2021). Posts from experts in the expert arena, for example, can be picked up directly by the public and thus enter the discussion arena and vice versa. Hence, online public arenas are not equivalent to platforms; rather, they can be found on a single platform. This is particularly true for Twitter/X, where actors from various arenas encounter and share different content. Twitter/X flattens multiple audiences into one – a phenomenon known as ‘context collapse’ (Marwick and boyd, 2010) – but offers the same functionalities and underlying communication structures of different arenas within the platform context. As different online public arenas intertwine, scientists may also reach unintended audiences. Twitter/X functions as an expert arena for many researchers to share their research and connect with their scientific networks (Collins et al., 2016; Costas et al., 2020); meanwhile, its architecture allows the public to read along and even to actively engage in those conversations. Recently, scientists have also increasingly been using the platform to engage in broader public discussions (Della Giusta et al., 2021).

From a normative perspective, science communication should be evidence-based to help laypeople make sound decisions, particularly during crises with various uncertainties (Burns et al., 2003). On social media, scientists can function as firsthand sources of evidence, as they can directly provide information ‘out of the lab’ (Szczuka et al., 2024). Meanwhile, Twitter/X, with its standard 280-character limit, is a ‘temptation for scientists’ to communicate beyond any evidence base (Brossard and Scheufele, 2022: 614). Scientific evidence is often defined as a ‘criterion of evaluation and comparison’ (Guenther et al., 2019: 41), referring to scientific facts and the use of data and factual claims (Flemming et al., 2017). As the pandemic’s context allows for the analysis of the prevalence of evidence in a situation marked by high uncertainty and initially sparse factual evidence (Lu et al., 2021), it is interesting to examine visible scientists’ evidence practices. Various types of evidence have been differentiated in the literature (Hoeken and Hustinx, 2009; Hornikx, 2005). When referenced, they can support the evidence base for a given issue (Kinnebrock and Bilandzic, 2023). We focus on three types of scientific evidence: statistical evidence representing a ‘numerical summary of a series of instances’ generated from scientific methods (Hoeken and Hustinx, 2009: 94); evidence through studies and their corresponding methods (Kessler, 2016); expert evidence based on the testimony of experts, characterised by their special competences and knowledge (Hoeken and Hustinx, 2009). Taken together, our first research question therefore asks:

• RQ1. About which topics, actors and evidence did visible scientists communicate on Twitter/X during the Covid-19 pandemic?

Moreover, as both the digital communication environment and socio-scientific issues may influence how scientists communicate (Brüggemann et al., 2020; Zhang and Lu, 2022), we examine the tonality and emotionalisation of visible scientists’ tweets. Tonality can provide insight into the presentation of content, which does not necessarily include emotive language, whereas emotionalisation refers to how the message itself is presented (Heiss et al., 2019). Social media has been described as including more emotionalisation than other channels as everyone can contribute to discussions (Stieglitz and Dang-Xuan, 2013). However, neutral messages are expected from scientists rather than emotionalised ones (Zhang and Lu, 2022). Yet, empirical research on emotionalised content in science communication is scarce (Huber and Aichberger, 2020; Lidskog et al., 2020).

Scientists are assumed to use scientific jargon, which can be defined as a language that is recognisable to scientists, but not to the broader public (Simis-Wilkinson et al., 2018). Hence, it can be seen as a barrier to laypeople’s comprehension (Wicke and Taddicken, 2021). Research based on qualitative interviews has found that scientists close their communications using jargon (Rödder, 2012). Thus, the use of scientific jargon on social media might be seen as an indicator of the scientists’ intended audience during the pandemic (Simis-Wilkinson et al., 2018). Furthermore, it is unclear how scientists address uncertainty in their social media communication in the context of socio-scientific issues. Uncertainty is inherent to the nature of science (Popper, 1959) and a central characteristic of scientific research, with scientists equipped to close research gaps, generate research questions and discuss the constraints of their work (Guenther and Ruhrmann, 2016). However, when addressing the public, science communicators often hesitate to describe the uncertainties inherent in scientific research (Gustafson and Rice, 2019). Most content analyses report an underrepresentation of scientific uncertainty in the media (e.g. Guenther et al., 2019). When scientists communicate their knowledge publicly, they must decide whether to communicate uncertainty (Peters and Dunwoody, 2016). Particularly during the pandemic, communicating scientific findings was challenging; while their rapid dissemination was crucial, they were often only preliminary (Fleerackers et al., 2022; Hendriks et al., 2023). It is questionable how visible scientists dealt with this in their communication. Overall, our second research question asks:

• RQ2. How did visible scientists communicate in terms of tonality, emotionalisation, scientific jargon and uncertainty on Twitter/X during the Covid-19 pandemic?

Role concepts

Although there are various typologies in the literature describing scientists’ roles in interactions with the public and politics (e.g. Bauer and Kastenhofer, 2019; Pielke, 2007), they have barely been empirically studied (Fähnrich and Lüthje, 2017; Roedema et al., 2021). However, role concepts offer a useful structure for capturing visible scientists’ communication behaviour on social media. As outlined, roles can be described as attitudes, expectations, behaviours and values that guide individuals in certain situations when performing a specific task (Turner, 2001). Two main strands of role approaches can be differentiated. According to the functionalist approach, roles are conceptualised as an expected attitude and behaviour of individuals occupying a certain position in society (Biddle, 1979; Merton, 1957). Meanwhile, interactionist role concepts define roles as dynamic rather than fixed sets (Turner, 2001). Individuals adopt and perform roles during social processes (Van der Horst, 2016). They can serve to analyse behaviour and are always linked to its performative side (Mellado, 2020).

In journalism studies, four role dimensions have been differentiated: Role conception focusses on the purposes of the profession; role enactment assesses how roles are put in practice, with an emphasis on an individual and role perception deals with the (perceived) expectation of the audience (Mellado, 2020; Mellado et al., 2016). Moreover, role performance concentrates on actual behaviour and practices, which might differ from the other three normative dimensions (Mellado et al., 2016). As publicly communicating actors in a changing media environment, parallels can be assumed between communicating scientists and journalists in the context of socio-scientific issues (Brüggemann et al., 2020). Hence, the adaptation of the journalistic role concept seems to be appropriate. We focus on visible scientists’ performed roles by analysing their communication on social media.

On social media, roles might overlap in practice, considering that the platforms are characterised by scientists’ engagement with various actors in different contexts (Roedema et al., 2021). We assume that visible scientists, like journalists, tend to redefine their roles through social media activities (Mellado and Alfaro, 2020). Without traditional gatekeepers, they must decide on their own how they want to participate in discourses online (Roedema et al., 2021), which requires new competences and dealing with their roles. Scientists might actively engage in discourses, by expressing their own views and calling to action. The provision of information and personal views can become intertwined (Scheufele, 2022). As the borders of science and journalism increasingly blur, journalistic practices might be adopted by scientists in their social media communication (Brüggemann et al., 2020). They can fulfil roles traditionally associated primarily with journalists, such as watchdogs or advocates for certain measures (Taddicken and Krämer, 2021). Hence, scientists’ role performance on social media might collide with their traditional role expectations to present research results rather than advocate certain policies (Walter et al., 2017).

In his categorisation of scientists in public, Peters (2021) differentiates the role of ‘public experts’ from two other roles scientists may take in public: When scientists popularise their research, they act in a ‘teacher role’, in meta-discourses about science and technology and the science–society relationship they take a ‘stakeholder role’. Peters states that the role of ‘public experts’ combines the role of scientists as ‘public communicators’ and the role of ‘public advisors’, as scientists are expected to apply their knowledge to explain and solve problems of society at large, crossing the boundary of science (Peters, 2021). Unclear is, how scientists fulfil their roles as public experts. Against the backdrop of science-policy relations, Pielke (2007) distinguished four ideal-typical scientists’ roles: These are the role of ‘pure scientist’, who focuses solely on research; the ‘science arbiter’, who answers factual questions of issues relevant to decision-makers without communicating his or her own preferences; the ‘issue advocate’, who advocates the implementation of a specific policy option and the ‘honest broker’, who aims to clarify and broaden the policy alternatives for decision-making without advocating a particular solution. As ideal types, Pielke’s (2007) categories represent a continuum from strictly limiting choices to broad options and are not mutually exclusive. He states that these four different roles ‘do reflect that scientists face practically meaningful choices in how they act in the context of policy and politics’ (Pielke, 2007: 1). Given that scientists have been key figures in public discourse about the virus at the boundaries of science, politics and society (Safford et al., 2021), the typologies can serve as a basis for classifying visible scientists’ role performance as public experts.

The few studies that have empirically analysed scientists’ roles have mainly focussed on their perceptions of their role in policy advice (e.g. Cologna et al., 2021; Turnhout et al., 2013). Studies on motives for public engagement can provide further starting points. Scientists primarily aim to inform and excite laypeople about scientific findings (Dudo and Besley, 2016). They see a responsibility to provide knowledge to the broader public and to policymakers (Getson et al., 2021). A study of scientists’ goals of public engagement showed that scientists see their contribution mainly in convincing the public and policymakers to include science in their decisions (Besley et al., 2020). Roedema et al. (2021) found that scientists felt a huge responsibility to communicate with the public online, and mostly saw their role in educating and raising awareness. Moreover, scientists’ intended roles did not correspond to their actual interactions with the public, leading to frustration by scientists in their public engagement online (Roedema et al., 2021). Overall, these studies highlight the fact that scientists perceive that they should play an active role in public discourse. In doing so, they seem to view their role primarily as ‘knowledge brokers’ (Meyer, 2010; Walter et al., 2017).

However, such studies remain narrow in only dealing with role conception and perception (Mellado et al., 2016), as self-identified roles might differ from actual performed roles (Fahy and Nisbet, 2011). Regarding scientists’ communication behaviour on social media, a study showed that scientists expressed more negative emotions in tweets addressed to political actors or journalists than in tweets to their peers (Walter et al., 2019). In controversial fields, scientists also use social media to express their opinions to the wider public (Biermann et al., 2023; Jahng and Lee, 2018; Walter et al., 2017). Walter et al. (2017) found that scientists on Twitter/X also called on others to act; scientists’ roles appear to have broadened as scientists have started to assume roles traditionally related to political advocates or journalists. These initial studies on scientists’ social media communication suggest that their communication on social media partly deviates from traditional scientists’ role perceptions. Moreover, although scientists are assumed to adopt traditional journalistic roles (Brüggemann et al., 2020; Taddicken and Krämer, 2021), there is a lack of empirical research exploring scientists’ interaction with journalistic content on social media. In general, the importance of social media in the distribution and discussion of news media coverage is growing (Kümpel, 2022). References to news media have both information and interpretation functions (Jakob, 2022). Media coverage can be enhanced or depreciated, and users thus act as gatekeepers (Singer, 2014). Therefore, elucidating how visible scientists refer to the media can provide further insights into their role performance.

Overall, our aim is to use the role concept as an analytical framework and make communicative roles empirically measurable via visible scientists’ communication behaviour by examining the nature of the statements, prevalence of calls to action and media references in their tweets. Thus, our third research question is:

• RQ3. How and to what extent did visible scientists’ role performance in terms of statement types, calls to action and media references differ on Twitter/X during the Covid-19 pandemic?

By analysing tweets from visible virologists during the Covid-19 pandemic, we concentrated on scientists who have only recently come into the public eye and quickly gained high visibility (Peters, 2021). Hence, it is worth analysing their communication behaviour over time to make statements about possible changes in their role performance. From this point of view, reaching out to non-scientific audiences was not necessarily a professional task for most virologists before the pandemic.

3. Methods

Identifying key scientists and periods

We conducted a manual content analysis of visible German scientists’ tweets (original tweets and quotes) posted during the Covid-19 pandemic. We developed a multi-stage procedure to identify the key visible scientists from the field of virology on Twitter/X (Supplemental material A). As discipline-specific expertise is a prerequisite, we first reviewed the websites of German virology research institutes. We applied different selection criteria (Lavazza and Farina, 2020): academic tenure (completed PhD), affiliation to an official research organisation and publication activity in peer-reviewed journals, identifying 311 scientists. To ensure that scientists represented ‘visible scientists’ in the discourse (Goodell, 1977; Joubert et al., 2023), we focus on virologists with more than 10,000 followers. Thus, we operationalise the analytic concept of visibility (Goodell, 1977), as scientists who were visible on social media during the Covid-19 pandemic. ¹ Of the 64 scientists with Twitter/X accounts, seven fulfilled the criteria of 10,000 followers at the time of the first data collection (August 2021). By scanning the list of accounts they followed, we identified and added another virologist who was not affiliated with a German research institute but had a German-speaking background and met all the above criteria. As we were interested in the communication behaviour over time, we needed to determine different periods over the course of the pandemic. We focussed on profound political events (e.g. lockdowns), as we assumed that scientists are encouraged to share their knowledge and engage with the public in times of high public awareness. Six periods from the start of the pandemic in March 2020 to the Omicron wave in 2022 were considered, each lasting 14 days (see Table 1 for dates).

Table 1.

Proportion of individual actor groups on all actors over time (n = 869), cross-tabulation of scientific evidence and period in visible scientists’ tweets (n = 891), and cross-tabulation of evidence types and period in visible scientists’ tweets including evidence (n = 257).

Actor type	Period 1		Period 2		Period 3		Period 4		Period 5		Period 6		Total
	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Science	100	49.0	89	38.4	98	49.5	36	43.9	54	46.2	23	63.9	400	46.0
Media	42	20.6	78	33.6	50	25.3	25	30.5	29	24.8	10	27.8	234	26.9
Politics	30	14.7	16	6.9	24	12.1	11	13.4	19	16.2	1	2.8	101	11.6
Practical experts	9	4.4	35	15.1	14	7.1	6	7.3	10	8.5	0	0	74	8.5
Laypeople	11	5.4	7	3.0	4	2.0	2	2.4	3	2.6	1	2.8	28	3.2
Other	12	5.9	7	3.0	8	4.0	2	2.4	2	1.7	1	2.8	32	3.7
No application of the chi-square test due to multiple coding
Scientific evidence
Not present	96	76.2	117	67.2	173	72.7	99	76.2	99	69.7	50	61.7	634	71.2
Present	30	23.8	57	32.8	65	27.3	31	23.8	43	30.3	31	38.3	257	28.8
χ² = 8.91, df = 5, p = .113
Evidence types
Statistical	5a, b	16.7	7b	12.3	28a	43.1	11a. b	35.5	12a, b	27.9	14a	45.2	77	30.0
χ² = 17.74, df = 5, p < .001, Cramer’s V = .263
Studies	8a	26.7	20a	35.1	18a	27.7	9a	29.0	19a, b	44.2	24b	77.4	98	38.1
χ² = 26.94, df = 5, p < .001, Cramer’s V = .312
Expert	21a	70.0	40a	70.2	36a, b	55.4	17a, b	54.8	20a, b	46.5	9b	29.0	143	55.6
χ² = 20.30, df = 5, p < .001, Cramer’s V = .281

Note. Period 1: 16 March 2020–29 March 2020; Period 2: 28 October 2020–11 November 2020; Period 3: 16 December 2020–03 January 2021; Period 4: 03 March 2021–16 March 2021; Period 5: 11 November 2021–25 November 2021; Period 6: 28 December 2021–11 January 2022.

Each different subscript letter denotes a subset of categories whose column proportions differ significantly from each other at the 0.05 level (Bonferroni correction).

Sample and measures

The data were tracked by account for the identified scientists and periods by a crawler based on the social media analytics framework (Stieglitz et al., 2018) that utilises Twitter’s open-source library. Thus, we were able to collate all Twitter/X activity for each scientist. We analysed all original tweets and quotes (N = 1003) posted by the eight scientists in the derived periods (M = 167.17, SD = 64.14). Each scientist posted on average 125.35 tweets (SD = 109.8) (Supplemental material A). To capture visible scientists’ communication behaviour in detail, we developed a standardised codebook containing formal categories (e.g. period, type of tweet) as well as deductive and inductive categories.

• Topics and actors. We coded the overarching theme of each tweet. We differentiated between 10 different topics (e.g. ‘Political Measures and Claims’) with subcategories; the topics were determined inductively from the material and referred to different societal subsectors. We captured the actors mentioned in the tweets from a list of actor types (e.g. science, media and politics). We defined the mentioned actors as all named persons or groups, organisations or institutions referred to.

• Evidence. In addition to the prevalence of scientific evidence in scientists’ tweets, we coded for three types of evidence: statistical evidence, studies evidence and expert evidence based on previous content analyses (Guenther et al., 2019; Kessler, 2016). For each type, we captured its prevalence.

• Tonality, emotionalisation and emotions: To examine the tonality, we coded the tone of each tweet by its overall impression. We differentiated between negative, neutral and positive tonality. In addition, we captured emotionalisation. A tweet was coded as emotionalised when using emotional expressions. The indicators of emotionalised tweets were exaggerations, figurative language or explicit expressions of emotions. Moreover, we coded the prevalence of discrete emotions. Drawing on previous content analyses capturing emotions (Huber and Aichberger, 2020), we created a list of four discrete emotions (joy, fear, anger and sadness) and coded their presence.

• Uncertainty. Uncertainty was coded when the tweets included scientific uncertainty. Indicators were statements about the uncertainty or provisional nature of scientific results, uncertainty due to a lack of representativeness, missing data or a lack of reliability or validity (Ruhrmann et al., 2015).

• Scientific jargon. To make statements about comprehensibility, we differentiated between tweets using scientific jargon (e.g. human cardiomyocytes, in vitro cellular toxicity), tweets using jargon that became common usage during the pandemic (e.g. R-value, PCR test) and tweets without scientific jargon.

• Statement types and calls to action. Drawing on previous research on communication behaviour, five different statement types were deduced (Jahng and Lee, 2018; Taddicken et al., 2019; Walter et al., 2017). The presence of one statement type did not exclude the presence of others. Information was coded when a tweet predominantly provided factual information. Information-seeking was prevalent when a search for information was recognisable. We coded appeals when a serious request for a certain behaviour or measure was made. Tweets coded as expressions of opinions involved expression of scientists’ own views, feelings or judgments. This category referred to both positive and negative judgements. By contrast, tweets coded as criticism always had negative connotations and involved stating the criticism of individuals, institutions or measures. Furthermore, the category calls to action captured whether scientists explicitly called on others to act. When a call to action was prevalent, we also coded for its addressee (e.g. public, political actors).

• Media reference and functions. This category captured references to traditional media (online, offline). As these references can have different functions (e.g. Jakob, 2022), we coded for the function that the media references had in the respective tweet (affirmation, appreciation, criticism and announcement). Inductively determined function categories were included in the analysis as binary variables.

Manual coding was conducted by four coders who underwent extensive training and adjustment processes until a mutual understanding of the categories was assured. Krippendorff’s alpha intercoder reliability was calculated on a random subset of 100 tweets. The coefficients varied between 0.7 and 1 across all variables showing overall acceptable reliability (Supplemental material B).

In total, 1003 tweets were analysed; 88.8% (n = 891) included a Covid-19 reference. This first finding indicates that the selected scientists’ overall Twitter/X communication mainly focused on the pandemic during our periods, as they communicated almost exclusively about the pandemic. Further analyses refer solely to tweets with reference to Covid-19. The Covid-19 sample (n = 891) comprised 514 original tweets and 377 quotes. The majority of the tweets were in German (82.2%); nearly, one-fifth were in English (17.7%) and one in French (0.1%). The analysis was carried out using descriptive statistics, recording frequencies and calculating cross-tabulations for the differences between the time periods. Where appropriate, we used chi-square analysis, which is often used to test statistical differences when comparing temporal variations (Huber and Aichberger, 2020; Peng et al., 2022). Since the cross-tabulations were larger than 2 × 2, a column proportion test (z-test) with the Bonferroni correction was performed to adjust the p-values to correct for multiple comparisons.

4. Results

RQ1: About which topics, actors and evidence did visible scientists communicate on Twitter/X during the Covid-19 pandemic?

Regarding RQ1, we first analysed the main topics and actors in the visible scientists’ tweets. To gain deeper insight into what these scientists were actually talking about, up to two main topics per tweet were coded (n = 1242). Unsurprisingly, the most common topic was ‘Research and Science’ (e.g. research on new variants) (29.4%), followed by ‘Political Measures and Claims’ (e.g. school closures) (25.1%). ‘Media’ (focus on media coverage of Covid-19) (16.5%) and ‘Prevention on Individual Level’ (non-pharmaceutical interventions, such as social distancing) (15.8%) occurred almost equally often, followed by ‘Spread and Growth’ dealing with infection figures (14%). Tweets about the social impact of the pandemic on society (‘Society’) accounted for 11.7%. The topic ‘Vaccination’ occurred nearly every tenth tweet (9.4%). Less prevalent were ‘Health Sector’ dealing with the capacity of hospitals (6.6%) and ‘Infection-related Topics’ (e.g. course of the disease) (6.5%). In 2.9% of all tweets, the visible scientists announced their media or public appearances; for example, in talk shows or press conferences (‘Announcements’). The prevalence of topics varied between the periods (Supplemental material C). It is noteworthy that during the last period, characterised by fewer posts overall, half of all tweets dealt with ‘Research and Science’ (50.2%), while ‘Media’ and ‘Political Measures and Claims’ were almost absent. However, overall, the results showed that visible scientists not only limited their communication to content on research and science, but often also addressed political measures, commented on media coverage and communicated about prevention measures at individual levels.

Moreover, 869 actors were mentioned in the analysed tweets (M = .97, SD = 1.14). The vast majority of all actors were from science (46%), followed by media (26.9%) (Table 1). Political actors accounted for 11.6% of all the actors. Practical experts from the health sector were slightly less present (8.5%). Our results show that visible scientists rarely mentioned actors from the general public in their tweets (3.2%). Although scientists most often mentioned actors from their field, actors from other fields accounted for more than half of all actors in their tweets. The results suggest slight differences in the frequency with which certain actors were mentioned at different time periods.

In summary, 28.8% of all visible scientists’ tweets contained scientific evidence. During different periods, the prevalence of evidence remained relatively stable (Table 1). The most frequently used type of evidence was expert evidence, which was present in more than half of the tweets including scientific evidence (55.6%), followed by evidence from studies (38.1%) and statistical evidence (30.0%). Thus, visible scientists most often used other experts and studies as sources of evidence in their tweets. Several chi-square tests with Bonferroni correction were performed to gain insight into the presence of different evidence types over time (Table 1). Expert evidence occurred significantly less frequently in the sixth period than in the first and second. Evidence from studies was significantly more visible in the last period than in the first to fourth. Statistical evidence was significantly more present in the third and last periods than in the second. Overall, our results indicate that visible scientists used different evidence types at different stages of the pandemic, as differences between the late and early periods were particularly visible. In times of knowledge production, when no published scientific findings exist, scientists seem to call on their peers as sources of evidence.

RQ2: How did scientists communicate in terms of tonality, emotionalisation, scientific jargon and uncertainty on Twitter/X during the Covid-19 pandemic?

To investigate how visible scientists communicated during the pandemic, we analysed the tonality of the tweets. Overall, tweets were mostly neutral (55.4%). The tonality of the remaining tweets was slightly more negative (24.7%) than positive (19.9%). The tonality of the selected scientists’ tweets fluctuated over time (Table 2). Positive tweets were more prevalent in periods 1 and 2 than in the last two periods. Tweets with negative tonality were significantly more prevalent in periods 3 and 5 than in period 1. In addition, we coded the prevalence of emotionalisation. Whereas tonality captured the general tone of the content and did not require the presence of emotive language, emotionalisation presupposed the presence of emotional keywords and hence referred to the way the message itself was presented. Almost, 30% of all tweets included emotionalisation (e.g. exaggeration, dramatisation) (29.5%). The visible scientists’ tweets included significantly fewer emotionalisation in the fourth and sixth periods than in the second period (Table 2). Thus, emotionalisation decreased over time; particularly at the beginning of the pandemic, the selected scientists integrated emotional elements in their tweets that were less visible in later periods. Furthermore, we investigated the prevalence of emotions in the visible scientists’ tweets. In 15.6% of all tweets, scientists explicitly mentioned discrete emotions (n = 139). The most frequent emotion was anger (9.5%) followed by joy (4.2%). By contrast, fear (0.9%) and sadness (1.5%) were almost absent in their tweets. Although discrete emotions were rather rare and the tonality was overall neutral, the tweets were not free from emotional elements.

Table 2.

Cross-tabulation of tonality and period, emotionalisation and period, scientific jargon and period, and uncertainty and period in tweets from visible scientists (n = 891).

Tonality	Period 1		Period 2		Period 3		Period 4		Period 5		Period 6		Total
	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Negative	15a	11.9	48b, c	27.6	76c	31.9	22a, b	16.9	43b, c	30.3	16a, b, c	19.8	220	24.7
Neutral	75a, b	59.5	78b	44.8	122a, b	51.3	82a	63.1	81a, b	57.0	56a	69.1	494	55.4
Positive	36a	28.6	48a	27.6	40a, b	16.8	26a, b	20.0	18b	12.7	9b	11.1	177	19.9
χ² = 46.57, df = 10, p < .001, Cramer’s V = .229
Emotionalisation
Not present	79a, b	62.7	106b	60.9	167a, b, c	70.2	101a, c	77.7	108a, b, c	76.1	67c	82.7	628	70.5
Present	47a, b	37.3	68b	39.1	71a, b, c	29.8	29a, c	22.3	34a, b, c	23.9	14c	17.3	263	29.5
χ² = 22.53, df = 5, p < .001, Cramer’s V = .159
Scientific jargon
Not present	98a	77.8	137a	78.7	146b	61.3	78b	60.0	95a, b	66.9	25c	30.9	579	65.0
Common usage	12a	9.5	29a, b	16.7	61b, c	25.6	32b, c	24.6	36b, c	25.4	30c	37	200	22.4
Pure jargon	16a, b	12.7	8b	4.6	31a, b	13.0	20a, c	15.4	11a, b	7.7	26c	32.1	112	12.6
χ² = 82.20, df = 10, p < .001, Cramer’s V = .215
Uncertainty
Not present	120a	95.2	168a	96.6	200b	84.0	116a, b	89.2	138a	97.2	75a, b	92.6	817	91.7
Present	6a	4.8	6a	3.4	38b	16.0	14a, b	10.8	4a	2.8	6a, b	7.4	74	8.3
χ² = 32.55, df = 5, p = .001, Cramer’s V = .191

Note. Period 1: 16 March 2020–29 March 2020; Period 2: 28 October 2020–11 November 2020; Period 3: 16 December 2020–03 January 2021; Period 4: 03 March 2021–16 March 2021; Period 5: 11 November 2021–25 November 2021; Period 6: 28 December 2021–11 January 2022.

Each different subscript letter denotes a subset of categories whose column proportions differ significantly from each other at the 0.05 level (Bonferroni correction).

The prevalence of scientific jargon in scientists’ tweets can provide further insights into how visible scientists communicate on Twitter/X, and can be considered an indicator of the comprehensibility of their tweets. Most tweets did not include scientific jargon (65%). Tweets with scientific jargon accounted for 12.6% of the total tweets. In 22.4% of the cases, the tweets contained jargon that had passed into common usage, such as polymerase chain reaction (PCR) tests or incidence. Visible scientists used scientific jargon differently in the pandemic’s course (Table 2). In the first two periods, tweets without scientific jargon were significantly more frequent than in periods 3, 4 and 6. Noticeably, a high percentage of scientific jargon was used in the last period, when tweets with pure jargon were significantly more prevalent than in almost all other periods. Furthermore, scientific jargon that had passed into common usage was more prevalent in periods 3–6 than in the first one. Thus, scientific jargon clearly increased in visible scientists’ tweets during the last periods.

The extent to which visible scientists directly address uncertainty can shed light on how they deal with scientific uncertainty in their social media communication. Our results show that they rarely communicated scientific uncertainty in their tweets (Table 2). However, differences over time occurred. Particularly in the third and fourth periods, uncertainty was prevalent in visible scientists’ tweets. When scientific evidence was prevalent (n = 257), scientific uncertainty was communicated in 12.5% (n = 32) of cases. Meanwhile, 87.5% of all tweets with scientific evidence did not deal with uncertainty (n = 225).

RQ3: How and to what extent did visible scientists’ role performance in terms of statement types, calls to action and media references differ on Twitter/X during the Covid-19 pandemic?

To answer our third research question, we first analysed the use of statement types, which can be an indicator of visible scientists’ role performance on Twitter/X. Although visible scientists predominantly provided factual information (53%), they also frequently expressed opinions (37.6%), and even voiced criticism in nearly one-fifth of all tweets (17.5%). They used appeals less often (11.4%) and hardly sought information in their tweets (3.1%). Looking at the individual statement types over time (Figure 1), the results of a chi-square test with Bonferroni correction indicated that the provision of information increased, as factual information was significantly more visible in the last four periods than in the first two (χ² = 62.20, df = 5, p < .001, Cramer’s V = .264, p < .001). We found no correlations between period and information seeking (χ² = 4.49, df = 5, p = .481) and period and appeals (χ² = 6.20, df= 5, p = .287). However, our results show a significant relationship between the expression of opinion and period (χ² = 43.80, df = 5, p < .001, Cramer’s V = .222) and the expression of opinion was significantly less visible in periods 1 and 3–6 than in period 2. A significant relationship was also revealed between criticism and period (χ² = 14.74, df = 5, p = .012, Cramer’s V = .129). More specifically, our findings show that visible scientists voiced significantly more criticism in the second than in the fourth period, whereas the other periods did not differ significantly in terms of the frequency of criticism.

Figure 1.

Prevalence of statement types in visible scientists’ tweets (n = 891) over time.

Percentages do not add up to 100 as multiple statement types could be coded.

Taken together, the visible scientists did not just provide information and disseminate knowledge but also extensively shared their own views and even criticism. The use of different statement types seems to be dynamic, as changes occurred over time. The results indicate a decrease in evaluative statements.

Moreover, we coded for explicit calls to action in the visible scientists’ tweets. Calls to actions were visible in one quarter of all tweets (25.3%). There were no significant differences between the individual phases (Table 3). These calls to action were mainly addressed to society in general (51.1%), followed by politics (35.6%). Calls to action were rarely addressed to science (8%) and even less to the media (2.2%). Hence, our results suggest that visible scientists advocate for the implementation of specific measures.

Table 3.

Cross-tabulation of calls to action and period and media references and period in visible scientists’ tweets (n = 891).

	Period 1		Period 2		Period 3		Period 4		Period 5		Period 6		Total
Calls to action	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Not present	87	69	127	73	178	74.8	110	84.6	102	71.8	62	76.5	666	74.7
Present	39	31	47	27	60	25.2	20	15.4	40	28.2	19	23.5	225	25.3
χ² = 9.94, df = 5, p = .077
Media references
Not present	89a, b	70.6	103b	59.2	185a, c	77.7	95a, b	73.1	112a, c	78.9	73c	90.1	657	73.7
Present	37a, b	29.4	71b	40.8	53a, c	22.3	35a, b	26.9	30a, c	21.1	8c	9.9	234	26.3
χ² = 34.79, df = 5, p < .001, Cramer’s V = .198

Note. Period 1: 16 March 2020–29 March 2020; Period 2: 28 October 2020–11 November 2020; Period 3: 16 December 2020–03 January 2021; Period 4: 03 March 2021–16 March 2021; Period 5: 11 November 2021–25 November 2021; Period 6: 28 December 2021–11 January 2022.

Each different subscript letter denotes a subset of categories whose column proportions differ significantly from each other at the 0.05 level (Bonferroni correction).

To gain deeper insight into how the selected scientists referred to the media, we examined media references (reference to specific news articles, media coverage) and their functions in more detail. Media references occurred in more than a quarter of all tweets (26.3%) (Table 3). In the second period, scientists referred significantly more often to the media than in the fifth and sixth periods, indicating a decrease in media references over the course of the pandemic. In summary, although our results indicate that the use of media references decreased over time, overall, media references were a common practice in visible scientists’ role performance as public experts on Twitter/X during the pandemic. To further investigate their media referencing practices, we examined the function of media references in the respective tweets. Multiple responses were coded. Scientists most frequently referred to the media to emphasise and appreciate a certain media contribution (41.9%), followed by affirmation of their own statements (31.6%). In 21.8% of all tweets referring to the media, scientists shared their own appearances. Less frequently, they referred to the media to criticise media coverage (11.5%). Thus, when visible scientists referred to the media, they primarily highlighted a particular media contribution.

5. Discussion

Overall, the science-related topic ‘Research and Science’ was most frequently mentioned in the scientists’ Twitter/X communications. However, other pandemic-related topics, such as politics, the media and society, were communicated as well. Visible scientists participated in broader public discourse, presumably transcending their role as broker of scientific information. It is worth discussing the extent to which this is also the consequence of the intertwining of different online public arenas, as content from different arenas diffuses on social media. In line with previous research (Walter et al., 2017), we found that while actors from science were by far the most mentioned actor group in visible scientists’ tweets, they also addressed actors from other fields, particularly the media and politics. Moreover, visible scientists did not limit their communication to sharing evidence-based information, as scientific evidence was not prevalent in the majority of tweets.

Regarding the affective aspects of visible scientists’ communication, the fluctuation in the tonality of scientists’ tweets illustrates that scientists’ communication behaviour was not static but subject to changes over time. With almost a third of visible scientists’ tweets being emotionalised, emotionalisation was even slightly more discernible than scientific evidence. As anger was the most frequently occurring discrete emotion in scientists’ tweets, visible scientists also seem to have used their social media channels to explicitly express their resentment and frustration, with, for example, government inaction. Hence, visible scientists’ communication was at least in part emotionally charged, especially at the pandemic’s start. Surprisingly, scientific uncertainty was seldom addressed in their tweets. Furthermore, the overall low frequency of jargon in their tweets suggests that scientists’ Twitter/X communication was mostly comprehensible to the broader public. However, the increase in scientific jargon in the last period of our analysis could indicate either that the selected scientists’ intended audience shifted again to their peers, or that they assumed that the public had become familiar with this kind of jargon. Moreover, there is a trend towards more factual information in the tweets of visible scientists over time.

Concerning visible scientists’ role performance as public experts, our findings emphasised that the scientists in our sample did not just provide information and disseminate knowledge but also extensively shared their own views and even criticism on Twitter/X during the pandemic. The changes over time in this context illustrate that their role performance is dynamic. We further found that calls to action were a common practice in their social media communication. Calls to action to the public may be due to the fact that they felt a strong motivation to engage directly with the public in the exceptional situation of the pandemic (Joubert et al., 2023). If we assume the use of calls to action as an indicator that visible scientists acted as ‘issue advocates’ (Pielke, 2007) during the pandemic, then the selected scientists seem to have used their expanded opportunities online to participate in decision-making processes by informally advocating for certain measures outside official policy advice. Furthermore, by appreciating and criticising certain media contributions, visible scientists acted as watchdogs of the media themselves.

Overall, our results indicate that visible scientists actively engaged in broader public discourses by voicing their opinions, calling politicians and the public to action, and classifying and evaluating journalistic content at the boundaries of different online public arenas. However, this did not exclude role performances associated with traditional roles; for example, the dissemination of factual information (Table 4).

Table 4.

Overview of traditional roles and (new) scientists’ roles on social media.

	Traditional roles	(New) Visible scientists’ roles on social media
Role performance	- Communication within science context - Neutral information source (facts) - No explicit calls to action - Orientation towards scientific criteria	- Communication beyond science context - Expression of opinions (evaluation) - Explicit calls to action - Orientation towards journalistic practices
Communication environment	Expert arena Indirect communication to public (mostly via mass media arena)	Intertwined online public arenas Direct communication to public (own decision about the specific manner)
Function	Educating, informing, explaining	Interpreting, critiquing, advocating
Examples	- Pure scientist (Pielke, 2007) - Knowledge brokers (Walter et al., 2017) - Objective expert (Fähnrich and Lüthje, 2017)	- Public expert (Fähnrich and Lüthje, 2017; Peters, 2021) - Issue advocate (Pielke, 2007) - Watchdog (Taddicken and Krämer, 2021)

6. Conclusion

We aimed to elucidate visible scientists’ role performance as public experts on social media in the context of socio-scientific issues by analysing visible German scientists’ communication behaviour on Twitter/X during the Covid-19 pandemic. Our study contributes to the debate on visible scientists’ roles and engagement in digital communication environments by shifting the focus from normative questions to an empirical research approach by adopting the journalistic role concept (Mellado, 2020; Mellado et al., 2016).

The findings of our case study reveal that visible scientists’ role performance as public experts in digital communication environments goes beyond the roles of ‘pure scientists’ (Pielke, 2007) or ‘knowledge brokers’ (Meyer 2010; Walter et al., 2017), crossing the boundary of science. In line with previous research, our results show that the selected scientists acted as ‘issue advocates’ (Pielke, 2007) for certain measures by calling to action and voicing their opinions (Jahng and Lee, 2018; Walter et al., 2017). Moreover, they also acted as ‘watchdogs’ of the media by referencing and evaluating specific media contributions. Hence, digital communication environments enable them to perform roles that are primarily associated with journalists. These roles seem to accompany the expanded opportunities online and the exceptional situation of the pandemic. Thus, this study empirically supports the assumption that digital communication environments and science issues with high public awareness lead to changes in scientists’ roles (Brüggemann et al., 2020; Taddicken and Krämer, 2021). Furthermore, our findings emphasise that the intertwining of different online public arenas on social media can lead to role conflicts: as particular visible scientists have to perform various roles simultaneously, as different online public arenas are accompanied by different modes of communication (Lörcher and Taddicken, 2017; Schmidt, 2013).

Moreover, visible scientists’ role performance was dynamic and fluctuated over time. In particular, at the pandemic’s onset, the selected scientists included emotionalisation, referenced the media and often expressed opinions, suggesting that the urgency of a crisis might prompt scientists to transcend their roles. In contrast, in the last period, their role performance was characterised by the use of scientific jargon, communication about research and the use of evidence from studies and statistics. These differences indicate a learning curve in visible scientists’ communication behaviour or a return to traditional roles and a withdrawal from the broader public discourse, as judgemental statements were less prevalent in the later periods. Thus, we conclude that new roles have not replaced old ones but that different roles coexist in intertwining online public arenas.

Our study had several limitations. As our analysis is limited to eight visible scientists in a specific field, the findings are not generalisable to all communicating visible scientists’ communication behaviour. The conclusions drawn from our analysis should be made cautiously, as each individual may have a significant impact and the sample size of each period varied. However, by applying a systematic multi-stage procedure to identify the most visible scientists, we ensured that we captured all virologists with a wide reach. Our manual coding can provide the basis for supervised machine learning for automated analyses, which can be helpful in examining scientists’ role performance in larger datasets. Whereas we focussed on analysing textual elements, the use of visual components embedded in tweets would be worth exploring. Another limitation is the category of actors in scientists’ tweets, as we only coded for the mentioned actors; we can only make statements about those actors whom scientists address in their tweets, but not about the interactions between the actors. Here, network analyses can provide deeper insights into the interconnection of actors from different online public arenas. Most importantly, the exceptional situation of Covid-19 must be considered when interpreting the results. It remains questionable whether the results are transferable to ‘normal’ times.

Nevertheless, the present study offers a starting point for further research. As roles are dynamic and negotiated in social processes (Van der Horst, 2016), studies on role perception (Mellado, 2020; Mellado et al., 2016) can be fruitful in showing the extent to which communicative roles meet the public’s expectations, which, in turn, are important for their evaluations (Wicke and Taddicken, 2021; Zhang and Lu, 2022). In this context, considering reactions to scientists’ performed roles is indispensable, as new roles are likely to be negotiated (Brüggemann et al., 2020). Furthermore, qualitative interviews with visible scientists could clarify the challenges of communicating and performing their roles in highly controversial discourses in digital communication environments. Therefore, comparing scientists’ role conceptions and performance can provide insights into the extent to which both role dimensions correspond (Roedema et al., 2021).

This study is among the first to analyse in depth visible scientists’ actual communication behaviour and role performance at the boundary of different online public arenas. Our findings highlight the importance of (visible) scientists reflecting on their own communicative roles (Lewenstein and Baram-Tsabari, 2022), especially in the realm of socio-scientific issues, which include not only scientific questions but also political, normative and ethical dimensions.

Author biographies

Kaija Biermann is a PhD candidate at the Institute for Communication Science at Technische Universität Braunschweig (Germany). Her research interests include science communication in digital communication environments with a special focus on the roles of scientists in online public discourses on socio-scientific issues.

Monika Taddicken is a Professor in Communication Science at the Technische Universität Braunschweig (Germany). Her research interests include science communication with a special focus on new media environments and user engagement.