Digital Accountability in Collaborative Public Governance in Times of Crisis: Analysing the Debate in a Polarized Social Forum

From Collaborative Governance …

Coordination between governments and their counterparts is primarily structured through formal contracts, written agreements, decision-making procedures, and negotiation regimes. These formal documents include each partner's responsibilities, deliverables, expectations for performance, rewards, and reporting requirements. However, partnership goals are often shifting and fuzzy, making it difficult for partners to estimate what the other partner expects (Lee and Ospina, 2022). According to research, informal mechanisms like long-lasting partnerships, reciprocity, and mutual trust fill the gaps left by formal contractual agreements (Koliba et al., 2011). Functioning partnerships are characterized by mutual respect and influence, a careful balance between synergy and autonomy, and equal participation in decision making and transparency (Forrer et al., 2010).

Taking the lens of accountability, collaborative governance arrangements have been described as polycentric complex networks of multilateral accountability ties, in contrast to a monocentric system of government with a clear chain of command and control (Koliba et al., 2011; Lee and Ospina 2022). In the same vein, Emerson et al. (2011, p. 15) argue that the ‘internal authority structure of collaborative institutions tends to be less hierarchical and stable, and more complex and fluid, than those found in traditional bureaucracies’. Building on these insights, Ansell et al. (2023) recently identified a new role for public bureaucracies, which act like collaborative platforms to facilitate shifting coalitions of actors to come together in purpose-built arenas and then scaffold their interaction. Overall, the literature agrees that collaborative arrangements often lack accountability (Ansell et al., 2023; Koliba et al., 2011).

… To Digital Era Governance

A debate on ‘digital era governance’ has emerged in parallel to the collaborative governance debate (Dunleavy et al., 2006). Digital era governance encompasses three components (Margetts and Dunleavy, 2013). First, to reverse the fragmentation from earlier decentralization measures, ‘reintegration’ is facilitated by joining up and de-siloing public sector processes. As reintegration also stresses the aspect of genuine partnership working between different organizations, it forms a clear link to collaborative governance. A second feature is ‘needs-based holism’, centring on user needs (i.e., creating client-focused structures by redesigning services from a client perspective and putting one-stop processes in place). Third, and most relevant to this study, ‘digitalization’ means adapting the public sector to embrace electronic service delivery wherever possible (Margetts and Dunleavy, 2013). From insights into the diffusion of public sector reform paradigms it is known that ‘new ideas and concepts layer and sediment above the existing ones, rather than replace them’ (Hyndman et al., 2014, p. 388; see also Torfing et al., 2020). With this, collaborative governance is presumably complemented, rather than substituted, by digital era governance.

The understanding of public accountability is also affected by the advent of digital era governance (Schillemans et al., 2013; Vanhommerig and Karré, 2014). Scholars have conceptualized developments in the area as ‘digital accountability’ (Agostino et al., 2022b, p. 152). For example, with respect to the ‘digitalization’ component, there are calls for more research on social media, as they will gain importance in the future for government communication (Torfing et al., 2020). Below, the implications of collaborative governance and digital era governance for public accountability are discussed.

Accountability in Collaborative and Digital Era Governance in Times of Crises

Accountability has conceptual underpinnings in agency theory (Bovens, 2007). Bovens (2007, p. 463) notes that ‘modern representative democracy could be described as a concatenation of principal–agent relationships’. Public accountability has historically largely been about control in the sense of limiting the discretion of governments through compliance with tightly drawn regulations and rules. In essence, government accountability is concerned with maintaining control over non-elected officials (as agents), who are empowered by the delegation of sovereign authority and act in the name of the people and their representatives (as principals; Forrer et al., 2010).

Accountability and collaborative governance

Figure 1 maps the six elements of accountability according to the seminal definition by Bovens (2007). Two elements of his definition deserve particular attention in collaborative governance arrangements, namely the ‘actors’ (1) and the ‘forum’ (2). With respect to the ‘actor’, Bovens (2007) refers to the ‘problem of many hands’, that is, who are the actors. While in traditional public administration, accountability lies with the government as one single actor (as agent) producing a single account, Hansen et al. (2022a) explain that accountability in collaborative governance ‘entails both multiple formats or types of accounts and multiple actors–inside and outside the collaboration–involved in the provision and reception of accounts’. This ‘double multiplicity’ is likely to produce various challenges; for example, the multitude of account-giving formats risks leading to ‘fuzzy accountability’ (Hansen et al., 2022a, p. 1158). In a similar vein, Lee and Ospina (2022) map the accountability components in the transition from bureaucratic to collaborative governance (Table 1). According to the authors, bureaucratic governance focuses on rules, processes, and command-and-control hierarchies. Elected officials make policy decisions, which bureaucrats implement in an efficient way, and accountability relationships between elected officials and bureaucrats is based on a hierarchy of authority. In contrast, Lee and Ospina (2022) find that challenges for accountability in collaborative governance arrangements stem from the shift from a bilateral to a multilateral perspective, requiring consideration of not only vertical but horizontal relationships and a focus on informal and implicit standards. They observe a development from control and audit issues to trust building on the side of the actors (as agents) who are collaborating for service delivery. Nonetheless, Ansell et al. (2022, p. 166; see also Hyndman and McConville, 2018) hold that ‘an accountable cocreation process stands a fair chance of creating a virtuous circle … [H]aving to give accounts can push and help a network or partnership to do better, which will enhance the support from external actors’.

Table 1. Accountability components in the transition to collaborative governance (Lee and Ospina, 2022, p. 65)

	Traditional Public Administration	Market-based Governance	Collaborative Governance
Theory base/paradigm	Bureaucracy	New Public Management	Public Value Governance; Networked Community Governance
Accountability relationship	Elected official-administrator	Government-private organization	Multilateral relationships between multiple participating organizations
	➔ More participants involved; from bilateral to multilateral ➔
Attribute of relationship	Hierarchy of authority (vertical)	Principal-agent relationship (vertical) or partnership (horizontal)	Mixed (vertical and horizontal)
	➔ Not only vertical relationships but also horizontal relationships ➔
Accountability standards	Administrative procedures, regulations, codes of ethics	Contracts, performance measures, mutual trust and durable relationships	Public value, mutual trust and durable relationships, performance measures
	➔ Not only formal/explicit standards but also informal/implicit standards ➔
Accountability challenges	Bureaucratic control, process standardization, tradeoff between accountability-types	Performance measurement, monitoring and sanctioning, risk sharing	Management of tensions and paradoxes, public value assessment
	➔ From trade-offs to paradoxes, from control/audit issues to trust building ➔

As accountability is a relational concept, a central question concerns the ‘forum’ (2) to which accountability is rendered. Here, Bovens (2007, p. 455) points towards what he refers to as the ‘problem of many eyes’. Similarly, Koliba et al. (2011, p. 211) refer to different ‘accountability frames’. Both studies discuss frames and critical stakeholders for each frame, for example, in political accountability—elected representatives, in legal accountability—courts, in administrative accountability—auditors, in professional accountability—professional peers, in market accountability—shareholders/owners and consumers, and in social accountability—citizens, (associations of) clients, interest groups, charities, and the public at large/civil society. Koliba et al. (2011) propose that collaborative governance arrangements will likely draw on a combination of some or all of the outlined accountability frames, ‘ultimately creating hybrid accountability regimes’ (p. 214). Nevertheless, this might prove difficult in practice, as different accountability frames may collide and place conflicting expectations on the actors in the arrangement, leading to accountability tensions (Lee, 2022). The study by Koliba et al. (2011), a case study of a collaborative governance initiative to mitigate the impact of Hurricane Katrina in the US, illustrates the confusion of project partners as a consequence of multiple accountability frames.

Data Collection and Sample

This study focuses on the highly relevant case of CTAs to mitigate the spread of COVID-19. CTAs were regarded as ‘digital helpers that should play a decisive role on the way back to normality after the coronavirus pandemic’ (Amann et al., 2021, p. 11). Given both high expectations of CTAs as a digital tool and the rich constellation of actors involved in delivering CTAs, they provide a curious object of investigation bridging digital accountability and collaborative governance. We are particularly interested in the topics that were debated by the forum and how the debate evolved over time. This research embraces a qualitative research strategy (Patton, 2015), and concentrates on the app's own ‘digital habitat,’ that is, online communication on social media, more specifically on the X platform, following the form of ‘netnography’ (Jeacle, 2021; Kozinets, 2002, 2015). Scholars of accounting and accountability have shown growing interest in utilizing social media for research in the recent decade (Nerantzidis et al., 2024). Social media platforms facilitate human interaction (Bellucci et al., 2019). They constitute speaking arenas that can act as social accountability fora where debates take place (Neu et al., 2019). It has been observed that social media platforms have the potential to overcome the administrative barriers associated with traditional accountability channels (Jeacle and Carter, 2014).

Our research approach can be predominantly described as ‘passive netnography’ (Jeacle, 2021). Netnography was initially developed in marketing research and defined as ‘a new qualitative research methodology that adapts ethnographic research techniques to the study of cultures and communities emerging through computer-mediated communications’ (Kozinets, 2002, p. 62). Passive netnography is the monitoring and observation of archival data, that is, ‘communications and postings of online members made before the researcher enters the community, but which is still available to view’ (Jeacle, 2021, p. 91). It can be differentiated from ‘active netnography’, where the researcher contributes to a continuous real-time conversation, for example, by placing a controversial statement in an online forum as a stimulus to start a discussion. A third form of netnography, that our approach also slightly touches upon, is based on ‘the reflexive field notes produced by researchers themselves during their observation of the online community’ (Jeacle, 2021, p. 92). This was achieved by constant discussion and reflection on the coding process by the researchers, as the coding was done with only a few months delay during the observation period.

Tweets are characterized by their transparency—with being ‘public by default’—and concise statements, as a tweet cannot exceed 280 characters (‘X’—previously called ‘Twitter’, 2022). X is nowadays also regarded as a key communication and public health dissemination tool during outbreak situations, enabling insights into how people respond to certain events and topics such as infectious disease outbreaks, especially during the peak of an event (Ahmed et al., 2019). On the one hand, research has found that data posted on social media sites represents the actual feelings, ideas, and words of a person (Saqib et al., 2023). On the other hand, social media data appears divided. Hong and Kim (2016, p. 777) argue that ‘[t]he “echo chambers” view focuses on the highly fragmented, customized, and niche-oriented aspects of social media and suggests these venues foster greater political polarization of public opinion’. Although X data may not be representative of the whole population, it allows following a sampling strategy that targets maximum variation in gathering data about social opinion (Patton, 2015).

Considering X as a forum for public debate, our focus was on debate amongst citizens over the entire lifespan of the collaborative governance initiative. This type of debate is also particularly interesting in the case of a CTA, as such an innovation requires (data) co-production from citizens (Polzer and Goncharenko, 2022). Multiple search queries on X were performed to collect the data, using combinations of organizational actors involved in the initiative (Austrian Red Cross, Ministry of Health, Uniqa, and Accenture) and the string ‘stopp corona’. Applying and benefitting from the ‘Twitter Academic Access’ scheme (that allowed full archive search delivering 500 tweets per search request) for this study, we collected data in multiple rounds (in intervals of about every month) via the application programming interface (API) of the Postman platform and exported the output to an Excel spreadsheet. In addition, we cross-checked the number of tweets with the search function of ‘tweet count’.

Along with the main text of a tweet, the data collection also included parameters such as user name, full name, date and time of the tweet, and tweet ID. Overall, initially 1,084 tweets were collected. During the analysis, seven tweets that did not relate to the Stopp Corona app were excluded (e.g., when a tweet was a response to a tweet by the Red Cross but commented on an entirely unrelated issue), bringing the total sample down to 1,077 (Table 3). The debate in the ‘forum’ was complemented by the accounts given by the ‘actors’ involved in the consortium. Here, we identified 295 tweets for the period of observation. Table 3 offers a breakdown of the 164 tweets that Red Cross Austria made as the organization that was leading the consortium.

Regarding forms of digital public accountability investigated, this research sits at the intersection between ‘dynamic’ and ‘citizen-initiated’ forms of accountability (Schillemans et al., 2013, see above). The availability of the app during the two years allowed citizens to engage with and scrutinize the digital tool on an ongoing basis. Both citizens who adopted and tested the app and those who did not download it due to various concerns found the means to express opinions on the online social forum (X). On the one hand, the covered tweets include conversations between actors from the consortium and citizens (dynamic accountability). Here, tweets by lead actors were debated on and replied to. On the other hand, we observed that citizens initiated debates by, for instance, asking questions to lead actors and scrutinizing the app's performance, thus holding the consortium accountable (citizen-initiated accountability).

Coding and Analysis

The tweets were subsequently imported into the Nvivo 1.6.1 software for content analysis. Each tweet forms one unit of analysis. Content analysis was performed manually by the team of two researchers. This approach contrasts studies that use ‘big data’ approaches to analyze debates on social media (e.g., performing automated counting of emotive words or URL link extraction) but that ‘are not particularly well-suited to “deep readings”’ (Neu et al., 2019, p. 45) and grasping the actual meanings of contents.

During the process of coding, following previous research (e.g., Hyndman et al., 2019), positive and negative statements were differentiated, that is, if a tweet took a critical or endorsing stance towards the CTA. For achieving qualitative rigour in the inductive research approach, the coding process was inspired by the method proposed by Gioia et al. (2013), although the themes are not rooted in the exisiting literature. Table 4 provides an overview of the data structure.

Initially, in closely paraphrasing the statements in the tweets, first order concepts were defined. A tweet message could contain multiple statements and thus refer to multiple first order concepts. In total, 1,675 codes (859 positive (51%); 816 negative (49%)) were assigned in this step. The results from this coding were 233 positive and 324 negative first order concepts (first column in Table 4). Here, the coding aimed at identifying the debated topics.

Subsequently, the concepts were aggregated into second order themes (20 positive and negative themes each; second column in Table 4). From of these themes, four thematic dimensions (again each positive and negative) were distilled: ‘content’, ‘context’, ‘governance of collaboration’, and ‘co-production’ (third column in Table 4).

In addition, the period of investigation was divided into seven temporal brackets (Langley, 1999). Each bracket was grouping around significant events over the lifespan of the CTA (Table 3). For example, the ‘initial launch and dissemination’ (bracket 1) was differentiated from the ‘discussion about EU interoperability’ (bracket 4), and emerging ‘new/alternative “tickets to freedom”’, such as vaccinations (bracket 5). This bracketing allows the reconstruction of the temporal dynamics in diachronic data.

Themes and Thematic Dimensions in the Debate

Overall, the social accountability forum scrutinized not only the immediate content of the ‘CTA as a service’ but also issues that were not directly related to the digital public health innovation. As shown in Figure 2, about half of the coded statements focused on the actual content of the CTA (45% positive and 55% negative). However, this means that the debate also evolved around other thematic dimensions—about one-fifth of statements addressed the wider context of the government's mitigation of the COVID-19 pandemic (22% positive and 19% negative). A further substantial part of the statements (17% positive and 18% negative) focused on governance issues relating to the collaboration initiative. Finally, 16% (9%) of positive (negative) statements related to the co-production of the public service, that is, the engagement of citizens and the potential and encountered barriers in doing so. Interestingly, the differences in the distribution of themes and thematic dimensions that take a positive and a negative stance are not very large from an overall perspective, except for statements endorsing or criticizing the co-production of the digital service.

A Temporal Perspective on Accountability

A second intention of this study is to shed light on how the overall debate on social media evolved over the lifespan of the initiative. As shown by the number of tweets per temporal bracket in Table 3, over 90% of the conversation happened during brackets 1 to 3. This is mirrored by the temporal distribution of tweets on the ‘Stopp Corona’ app by Red Cross Austria as the leading involved actor (Table 3). This section thus focuses on these three first brackets for a meaningful presentation of the calculated percentages from the coded data and to prevent outliers potentially appearing important beyond measure.

Figure 2 presents the distribution of the thematic dimensions and Table 5 shows the distribution of thematic concepts over the temporal brackets. Regarding the former, a relatively stable pattern can be observed over time, with two exceptions: first, looking at the positive and neutral dimensions, issues of co-production seem to have gained nuanced prominence during the third bracket, potentially reflecting frustration due to the second (and long) lockdown in autumn and early winter 2020, which the app did not help to prevent. An interpretation is that the app was potentially regarded as a ‘ticket to freedom’ from this lockdown and thus calls to participate and promotion of the CTA were more intense during bracket 3 than during previous ones. Second, at the same time (especially in comparison with bracket 2), the governance constellation received enhanced scrutiny on the negative side (during the third bracket). An explanation for this could be that during this bracket, disillusionment was possibly starting to form due to the low uptake of the innovation, also spurred by perceived inadequacies in governance constellation and communication efforts by the consortium. This discontent was then voiced in bracket 3 in the accountability debate.

Table 5 breaks down the distribution of second-order themes over the temporal brackets (separately for positive and negative statements). The results reveal a number of developments. First, the debate about the app's functionality started on a relatively low level in bracket 1 and then intensified over time. In contrast, during the first bracket, issues such as privacy (positive) and transparency (negative) were of particular concern in the online community. An explanation for this development is that experience with the app had only begun to form and, thus, themes other than functionality were more prominent at the beginning of the initiative.

Second, indicators for this starting formation of experience with the Austrian CTA—and repercussions for the social accountability debate—can be seen at several points. First, bracket 2 brought about more intense examination of the approach of other countries. Here, comparisons of the ‘Stopp Corona’ app were made with CTAs in other countries (both positive and negative), predominantly with its German counterpart (which was a government-led innovation). A second indicator for the accumulation of knowledge with the app relates to the fact that over time, the debate about the potential effectiveness of the innovation intensified, that is, the CTA could play in mitigating the pandemic was better understood from an overall perspective.

Another observation relates to the scrutiny of the actors that were part of the governance constellation. This debate was negatively in tone right from the beginning, except in bracket 2, where less criticism targeted the constellation compared to other themes. This result contrasts communication as a polarized theme. While a substantial number of tweets praised the communication, there was, at the same time, persistently strong criticism. This disapproval of the consortium's communication peaked in bracket 3, possibly fuelled by frustration at the extended lockdown during this temporal bracket and cumulative discontent with the ambiguity of actors’ statements. These observations were evident in tweets referring to, for instance, vague and ambiguous communication by the consortium, including too much noise, as well as critiques about the lack of good marketing of the app when it was needed the most (when pandemic cases and lockdown measures were on the rise).

Overall, a number of temporal dynamics can be identified, illustrating that the accountability debate was not static but embedded in developments in the wider context of the pandemic (and possibly even political events beyond). For instance, the focus of the debate changed when another wave hit due to a further mutation of the virus or experience with the app accumulated. The passage of time allowed users to make cross-references to other countries or to comment on the relationship of digital contact tracing with other measures to mitigate the effects of the pandemic. In the following section, we discuss the findings and outline the implications of the analysis.

Contributions

We started this scholarly inquiry with the observation that (digital) accountability in collaborative public governance initiatives (e.g., Hansen and Triantafillou, 2022a; Lægreid and Rykkja, 2022) and digital era governance (e.g., Agostino et al., 2022a; Bracci, 2023) is under-researched. This is particularly the case in the context of crisis and with respect to the ‘debate’ (Bovens, 2007) that takes place in the forum to which accountability is discharged. In addressing these gaps, the study makes several contributions—exploring the understudied digital accountability forum, the themes of the debate, and the evolution of the debate over time. Furthermore, in light of these aspects, the study also contributes conceptually to an enhanced understanding of accountability by introducing the concept of perceived expertise of actors at the ‘delivering’ (i.e., members of the consortium) and ‘receiving’ end (i.e., citizens) of digital public services.

Limitations and Further Research

As with all empirical studies, the chosen methodological approach has strengths and limitations. With respect to the strengths of analyzing the selected digital social forum, the results from the ‘context’ dimension related to accountability particularly highlighted a large variation of voiced opinions that text genres other than new media are probably unable to grasp. The polarization of the debate on social media (Hong and Kim, 2016) adds to this great variety. Moreover, it can be argued that in studying digital innovations, a research strategy that focuses on X as part of the ‘digital habitat’ of tech-affine persons makes the case also a ‘relevant’ one (Patton, 2015). Issues that the tech-savvy community raises, for example concerning the functionality and accessibility of a digital innovation, most likely apply to other societal groups as well. In addition, the chosen research approach can complement studies that use big data techniques for automated coding (Neu et al., 2019; Nerantzidis et al., 2023) by identifying more nuanced aspects from the empirical material through qualitative coding.

Nevertheless, there are several limitations of the present study that deserve to be mentioned. First, the analyzed debate in online fora is never representative of society overall and thus needs to be complemented by other research approaches. Second, as the analyzed sample is limited, a big data analytics approach could cover much more content, for example, analyzing the comments on newspaper articles.

Questions for further research to address include, for example, how the social accountability debate can be kept alive. In this context, it is relevant to know more about the participating social media users. For example, do most members enter the debate by responding to a tweet issued by the consortium, by re-posting a tweet or do they attempt to start a debate ‘from scratch’? Is it always the same members that respond to tweets by organizational actors? Are members contributing consistently over time or not? Moreover, are there differences in the debate between collaborative governance and other organizational arrangements for pandemic mitigation? Also, as expertise is a dynamic concept (Herling, 2000; Kotzee and Smit, 2016), how is it created, nurtured, and curated, but also destroyed in social accountability? Online fora allow answering these questions and tracing the debate between ‘actor(s) (1)’ and the ‘forum (2)’ (Figure 1), as participants are assigned unique identifiers.

Practical Implications

The results from this research have important implications for practitioners interested in the acceptance of innovations in an era of digital era governance. Often, in the design of digital innovations, the technical is privileged over the social context (Bailey and Barley, 2020). Similarly, Agostino et al. (2022a, p. 150) highlight that digitalization ‘needs to be understood as a holistic process, not just a technological problem, thus requiring a closer look at the behavioural, psychological, ethical and contextual factors involved’. The present study shows that a better understanding of digital technology adoption in the public sector in collaborative governance emerges from a focus on accountability considerations beyond account-giving by organizational actors (Andreaus et al., 2021) and by considering the dynamic debates over time in the forum (see also the recent synthesis by Agostino et al., 2022 on accountability issues in digital transformation). Policymakers should carefully monitor the public opinion that is expressed on social media, possibly in combination with monitoring debates around political accountability and expertise that take place in ‘traditional’ media (see, e.g., the study by Amann et al., 2021, that investigates the discourse around CTAs in newspapers). Here, we also see ample potential to engage with the communication literature (Vásquez, 2022) that, for example, stresses the importance of appropriate communication by governments (e.g., Reynolds et al., 2020), but at the same time considering the reactions at the receiving end of communication. In this context, Lutzky and Kehoe (2022) recently analyzed linguistic changes across time in a large-scale dataset of over 136,000 tweets posted by and addressed to the Irish airline Ryanair between August 2018 and July 2019, using a corpus linguistics approach. Page (2022) highlights that interactions in social media contexts are of multimodal character, that is, include audio-visual resources such as images and videos that represent huge resources for studying digital accountability—for example, when influencers are criticizing the actions of governments.

A second point for policymakers to consider is that the evaluation of the digital innovation does not happen in isolation, but within the ‘ecology’ of other (digital) tools, measures, and policies (Höllerer et al., 2020; Samuel et al., 2021). We suggest that such a relational perspective, that is, looking more holistically at the forum's reaction to pandemic mitigation, can help improve collaborative service delivery and create a ‘virtuous circle’ for such arrangements (Ansell et al., 2022). Such an understanding is possibly particularly important for innovations requiring citizen co-production in sensitive areas such as health.

Furthermore, as shown by the debate extending beyond the mere content of the digital public innovation, transparently showing the role and expertise of leading actors in a collaborative governance arrangement becomes critical to make them accountable. That may at times mean proactively responding to the prospective question, before it emerges, of why and how certain non-obvious actors get involved in the development and implementation of a digital co-production tool.