Expectation dissonance: the role of perceived negativity bias in enterprise social media in explaining accountability and support

Abstract

Purpose

Enterprise social media (ESM) are expressive spaces where users exchange emotional workplace communication. While some studies have explored how positive emotions may be contagious, little research explored the notion that negative communication may accumulate on enterprise social media. This study explores perceived negativity bias and its correlates in the context of ESM.

Design/methodology/approach

This study relies on survey data collected from 599 employees of a global organization. The response rate was 18.7%. Structural equation modeling was used to test the hypotheses.

Findings

The results contribute to research on ESM by demonstrating that perceived negativity bias is positively related to feelings of accountability and negatively associated with social support. Furthermore, the results indicate that unmet communication expectations on ESM can have implications for perceived social support beyond online contexts and accountability through perceived negativity bias.

Research limitations/implications

The findings demonstrate how employees' unmet expectations about ESM use increase feelings that a digital environment is disproportionately negative, which may create an “unsafe” space for employees and a fear of being held accountable for their contributions. This study highlights how the Expectation-Disconfirmation Theory provides a fruitful framework for studying enterprise social technologies.

Originality/value

This study suggests that work is not merely a rational endeavor, and that emotions and personal feelings (including negative ones) may shape workplace communication on ESM. We contribute to research on ESM use by using the Expectation-Disconfirmation Theory as a lens to study antecedents and implications of perceived negativity bias.

Citation

van Zoonen, W., van der Meer, T. and Sivunen, A. (2024), "Expectation dissonance: the role of perceived negativity bias in enterprise social media in explaining accountability and support", Information Technology & People, Vol. 37 No. 8, pp. 196-215. https://doi.org/10.1108/ITP-05-2023-0502

1. Introduction

Negativity bias is a widely studied phenomenon in communication science and typically refers to the dominant prevalence of negative communication in mass media and social media (Knobloch-Westerwick et al., 2020; Rim and Song, 2016; Soroka et al., 2018). Negativity in media content is a strong driver of individual media-related behavior and motivation, including information processing, media choice, and user experiences (van der Meer et al., 2020). Indeed, typically, negative content elicits stronger responses than positive content (Stieglitz and Dang-Xuan, 2013). Evolutionarily, it makes sense that we pay more attention to threats in our environment. In that way, negative information has an informative function (Zaller, 2003). Yet, when we perceive our media environment as disproportionately negative, this might come with harmful consequences, like media avoidance (Skovsgaard and Andersen, 2020) or experiencing issue fatigue (Gurr and Metag, 2022). Even though negative emotions and events are inherently part of working life and communication (Jalonen, 2014), little research has explored the presence, origins, and consequences of perceived negativity bias within organizational environments.

Notably, communication in organizations can be abundantly negative as employees may experience and communicate about fear, anger, disgust, distress, or shame in their workspace (Fisher, 2019) and vent frustration and negative emotions on available platforms (Chen and Wu, 2022). In addition, research has articulated negative workplace events such as workplace bullying, harassment, emotional abuse, or abusive supervision (Eissa et al., 2020; Zhou et al., 2021). Krishna and Kim (2015) discuss the “confessions of an angry employee,” demonstrating that anger and frustration with policies, personnel, and management are widely shared on social media. The accumulation of negative communication behaviors in organizations may occur from egoistic motivations, such as venting negative feelings, or altruistic motivations, such as warning others (Lee and Kim, 2020). Negative communication among employees on enterprise social media (ESM) or within organizations may be at odds with social and organizational norms (van Zoonen et al., 2023). For instance, Glikson and Erez (2013) found that display norms in virtual teams call for the suppression of negative emotions. As such, it seems important to examine how expected and perceived communication inform user experience, especially regarding employees' (negative) communication on ESM.

ESM is a class of internal web-based technologies that allow workers to create profiles, send messages, view, edit, or reply to other messages, and reveal organizational communication partners (Leonardi et al., 2013). This definition has been widely adopted in later studies on ESM (e.g. Li et al., 2021; Treem et al., 2015). ESM are important as they support open communication (Gibbs et al., 2015), afford visibility of communication and behavior (Treem and Leonardi, 2013), facilitate knowledge-sharing (Leonardi et al., 2013; van Zoonen et al., 2022a, b), and impression management (Sun et al., 2021). However, research calls for a deeper understanding of emotional expressiveness on ESM platforms, including the nature and valence of the communication that is made visible (Reychav et al., 2019). Research has suggested that ESM may be a platform for employees to vent negative emotions and frustrations (Lu et al., 2019), as colleagues' reinforcing reactions to disclosing negative emotions can increase feelings of inclusion and organizational self-esteem (Reynolds-Kueny and Shoss, 2021).

However, thus far, research has not considered that communication on ESM may be largely negative or, more importantly, that employees may perceive communication on ESM as disproportionately negative. Here, we aim to fill a literature gap by exploring whether the discrepancy between the expectations and perceptions of communication on ESM may lead to increased perceptions of negativity bias in organizational settings. Specifically, we examine how differences between expected and perceived ESM use may be associated with a perceived negativity bias, which may be associated with an “unsafe” space for employees and a fear of being held accountable for their contributions. As such, we contribute to communication literature by advancing our understanding of how perceived negativity bias may contribute to views of ESM as technologies of accountability (Treem, 2015) and decay in feelings of social support. We demonstrate that perceived negativity bias may emerge from a negative disconfirmation between expected and observed ESM use.

2. Theory

2.1 Discrepancies between expectations and observations

To understand the psychological implications and expectations of social media, it is important to consider the elements or affordances of social media technologies (Bayer et al., 2020; Treem and Leonardi, 2013). Within the situated context of use, expectations of technology's features, how they should be used, and what they are good for buffer our perceptions of technologies (Leonardi, 2009). Notably, these perceptions and expectations of technologies are formed during the practice of work (Fulk, 2017). Employees interact with ESM and coworkers – comparing experiences and expectations – thereby evaluating the utility of technologies for individual goal attainment (Treem et al., 2015).

Importantly, expectations of what technology should do or should be used for – cf. technological frames – are social constructions and may differ across individuals and use contexts (Treem et al., 2015). Orlikowski and Gash (1994) refer to technology frames as “the underlying assumptions, expectations, and knowledge that people have of a technology” (p. 174). In the context of social media, expectations and assumptions of usage may be challenged by the experiences users encounter on the platform. The importance of real-life experiences meeting expectations is widely studied in consumer research. For instance, the Expectation-Disconfirmation Theory (EDT) is mainly grounded in marketing and consumer behavior research and posits that a consumer's satisfaction is a function of a priori expectations and subsequent disconfirmation (Oliver, 1980). As such, EDT provides a theoretical basis for understanding the importance of expectations and disconfirmation – i.e. unmet expectations (Lankton et al., 2016). Notably, disconfirmation refers to a discrepancy between expectations and observed technology use, which is conceptually distinct from perceived negativity bias, which refers to the notion that employees feel that content on ESM is disproportionately skewed toward negative communication.

Importantly, ESM facilitates the possibility of making communication and behaviors visible to others (including third parties; Treem and Leonardi, 2013). Communication visibility theory (Treem et al., 2020) suggests that social media users can make their actions visible to bigger audiences and observe others' actions more widely than with more traditional communication technologies. Consequentially, ESM users may observe what other users are doing with the technology and evaluate the extent to which observed user patterns may differ from the idiosyncratic expectations of use. Research suggests that, over time, groups develop ideas about how technologies should be used, leading people within a particular context to operate with a shared understanding of technologies (Treem et al., 2015). Yet, the assumption of congruent frames is often flawed, as technologies can mean different things to different people at different times. Hence, individuals may develop different expectations and adopt different usage patterns.

ESM is important for knowledge sharing and learning about the experiences and expertise of other organizational members (van Zoonen et al., 2022a, b). However, online knowledge-sharing environments may also invite problem crowding as online platforms create a visible curation of a diverse set of problems (Haas et al., 2015). We focus on expectations related to the extent to which communication on ESM is expected and observed to be about opportunities versus threats, positive events versus negative events, gains versus losses, or solutions versus problems. Discrepancies between what individuals think ESM should be used for and what they observe ESM is used for may impact how individuals make sense of these technologies (Lankton et al., 2016). Before we can examine the specific implications of a discrepancy between expected and observed ESM use, it is important to determine the presence and extent of the difference between expected and observed usage. Hence, we hypothesize:

H1.

There is a discrepancy between what employees think ESM ought to be used for and how employees perceive ESM is currently used.

2.2 Discrepancies and perceptions of negativity bias

Negative communication is commonplace in organizations (Fisher, 2019) and may result from egoistic or altruistic motivations (Chen and Wu, 2022). Overall, negative information, rather than positive information, is considered more attention-grabbing and appealing as it is seen as unambiguous, consensual, unexpected, dramatic, sensational, entertaining, eye-catching, interesting, and short-dated (Lengauer et al., 2012). This attraction to negative news stems from specific psychological mechanisms. The negativity in the news taps into a built-in mechanism where individuals are wired to constantly scan their environment for risks (Lengauer et al., 2012). Alarming threats coming from one's digital news environment in the form of negative information need to be carefully processed to understand how to avert personal risks (Soroka et al., 2018). Thus, when offered a choice, individuals tend to generally be more attentive to and interested in “bad news” rather than “good news” (Knobloch-Westerwick et al., 2020; van der Meer et al., 2020), as positive information is nearly synonymous with the absence of news (Soroka et al., 2018). It can, therefore, be expected that workers perceive their online information environment, such as ESM, to be unevenly negative when their expectations of the tone and nature of communication do not match their observations.

Building on EDT, we suggest that it is neither the expectation nor the observation of language, emotionality, or negativity, per se, that influences the perception of negativity bias, but rather the discrepancy between what users expect from the technology and what they observe other users are doing with the technology that impacts perceptions of negativity. Simply put, disconfirmation refers to the discrepancy between expectations and experience. EDT suggests that satisfaction or dissatisfaction is a function of the size and direction of disconfirmation (Venkatesh and Goyal, 2010). Extant research has used an EDT lens to explain system use in organizations as a function of the disconfirmation between expectations about the technology and the actual experiences with the technology (Venkatesh and Goyal, 2010).

Disconfirmation, as the extent to which technology performs better or worse than expected, has been linked with trust in technology and user intentions (Lankton et al., 2014, 2016). In addition, EDT research demonstrated that disconfirmation in cases where observed usage falls short of expected usage of information technologies leads to dissatisfaction among users (Fan and Suh, 2014). In addition, in consumer research, negative expectation disconfirmation leads to customer distrust and bias in subsequent evaluations (Zhang et al., 2021). These findings are relevant in the context of employees' ESM use because EDT assumes that users of technology, whether consumers or employees, will evaluate their experiences based on a calculus of expected and experienced use. When differences between expectations and experiences increase or fall outside a zone of tolerance, this may lead to dissatisfaction or negative evaluations, or even worse, spawn avoidance, workarounds, and dysfunctional behavior. Brown et al. (2012) further noted that compared to positive disconfirmation, there is “a disproportionately higher negative effect of negative disconfirmation.” (p. 477).

Cognizant of the notion that negative stimuli are weighted more heavily than positive stimuli (Brown et al., 2012; Peeters and Czapinski, 1990), and in line with EDT, we argue that perceived negativity bias is manifested when negative disconfirmation occurs. This is particularly interesting because (enterprise) social media technologies are typically cast as social lubricants and echo chambers – assuming confirmation rather than disconfirmation (Leonardi et al., 2013). However, negative disconfirmation may occur when observations of problematic or negative information exceed the expected amount of such information. In line with previous findings that suggest negative disconfirmation may have negative consequences (Lankton et al., 2014), result in dissatisfaction (Fan and Suh, 2014), and disproportionately higher adverse effects (Brown et al., 2012), we hypothesize that such disconfirmation may trigger perceptions of negativity bias. Hence:

H2.

Negative disconfirmation between user expectations and perceived ESM content will increase perceived negativity bias.

2.3 The mediating role of perceived negativity bias

Finally, this study explores the implications of perceived negativity bias in organizations beyond ESM use. The notion that individuals' perceptions drive decisions and risk assessments relates to the literature on equivalence framing. Since logically equivalent information is perceived differently based on how it is framed, subjective perceptions can be the driving force in decision-making and anticipated behavior outcomes (Kahneman and Tversky, 1984). In particular, the viability of an option may depend on whether a negative outcome is perceived as a cost or an unrecoverable loss during evaluation. In a well-known demonstration of equivalence framing, Tversky and Kahneman (1981) showed that when logically equivalent information is presented in terms of gains, it tends to garner more support for risk-aversive programs than risk-seeking ones. Conversely, when the same information emphasizes losses, the effect is reversed. Here, we conclude that people are not idealized decision-makers who perfectly predict future experiences and therefore presume that subjective perceptions of negativity bias can be associated with variation in ESM-related outcomes, here perceived accountability (Treem, 2015) and social support (Li et al., 2021).

Accountability is a fundamental aspect of organizing within society and the organizations that inhabit it (Hall et al., 2017; Karunakaran et al., 2022). Although accountability is defined in numerous ways in different disciplines, the concept, in general, refers to the “perceived expectation that one's decisions or actions will be evaluated by a salient audience and that rewards or sanctions are believed to be contingent on this evaluation” (Hall and Ferris, 2011, p. 134). While accountability literature covers levels of analysis and spans a wide array of disciplines from political science to social psychology (Frink et al., 2008; Power, 1996; Tetlock et al., 2013), we focus on individuals' perceived accountability in organizational settings. Central to the notion of accountability is the possibility of an evaluation occurring; in other words, an individual could believe that an account-giving might be required based on their action (Hall et al., 2017).

Perceived accountability is particularly salient in the context of ESM. Indeed, research often casts ESM technologies as technologies of accountability because the affordances of these technologies, when appropriated, provide an account of the user's activity, making them accountable to each communicative act (Treem, 2015). In this sense, perceived accountability in the context of ESM is rooted in visibility, as ESM enables a visible record of communication that individuals are then accountable for in assessments by others (Treem et al., 2020). Suchman (1995) suggests that systems aimed at the documentation of actions – such as social media – create accountability akin to a bookkeeper's ledger, where records of accounts paid and outstanding are documented and visible. We argue that perceptions of negativity bias will increase perceptions of accountability and decrease social support by highlighting a social context conducive to potentially negative evaluations of accountable acts.

While ESM affordances may provide new ways of making communication and behavior visible long after its initial communication, the appropriation of this ESM affordance may generate a heightened sense of the possibility of being held accountable for communication and behavior on ESM (Treem, 2015; van Zoonen et al., 2022a, b). Razmerita et al. (2016) argued that a fear of accountability might reduce knowledge sharing because others may criticize, misinterpret, or misuse contributions. We argue that perceptions of accountability may increase when communication is perceived as disproportionately negative, infused by negative disconfirmation. Perceived negativity bias may make the anticipated consequences of being criticized or misinterpreted more profound for users than neutral, positive, or supportive responses. This aligns with reasoning suggesting that negativity biases create accountability pressures for public service managers (Oh et al., 2022). Indeed, managers in public sectors are found to respond asymmetrically to negative performance information (i.e. negativity bias), leading them to engage in blame avoidance strategies (Hong et al., 2020).

Along similar lines, social support is important in ESM use as social connections in individuals' ESM networks are typically used to request or offer social support (Maier et al., 2015). When perceptions of communication (on ESM) are predominantly negative, this may reflect an unsupportive and demoralizing communication environment. Bowling et al. (2005) concluded that “negative communications about work should not be regarded as a form of emotional support in future research” (p. 486). Similarly, Stieglitz and Dang-Xuan (2013) indicate that while positive affect in messages may reinforce a sense of community, negative affect in messages may discourage continued participation or result in hostile or insulting interactions. Importantly, research on negative interactions in the workplace suggests that negative events may undermine social support, partly because individuals are prone to infer malicious intentions from others' negative behavior or communication (Duffy et al., 2002). Social support in organizations refers to the idea that an employee believes they are valued, informed, communicated with, emotionally cared for, and part of a social group (Foy et al., 2019). Reynolds-Kueny and Shoss (2021) found that colleagues' negative reactions to employees' disclosures reduced social inclusion and organization-based self-esteem. Lee and Kim (2020) suggested that negative employee communication on social media may have detrimental effects on organizations, labeling such communication behaviors “destructive voice.”

Moreover, negative communication may convey information about the sender's emotional state and judgments of a person or situation (Stieglitz and Dang-Xuan, 2013) and may be viewed as inappropriate in a professional communication space. Notably, negative reactions on social media are often viewed as inappropriate and found to decrease the willingness to provide support, mainly because it is considered a norm violation (Ziegele and Reinecke, 2017). Because perceived negativity bias reflects a sensitivity to negative or unfavorable stimuli, the communication environment may be perceived as less safe, highlighting members' shortcomings, errors, or missteps while discounting or overlooking successes, solutions, or positive events (Hood et al., 2016). Hence, employees who perceive that communication on ESM is disproportionately geared toward negative content may feel that such an environment is less conducive to providing social support and more likely to hold individuals accountable for their contributions. Ensuing the reasoning above, we hypothesize:

H3.

Perceived negativity bias will be positively related to a) perceptions of accountability and negatively related to b) perceptions of social support.

H4.

Negative disconfirmation between user expectations and perceptions of ESM use will be related to a) perceptions of accountability and b) perceptions of reduced social support through perceived negativity bias.

3. Method

3.1 Research site and sample

This study was conducted among office employees of an industrial company headquartered in the Nordics. The organization facilitates using an ESM modeled after popular organizational social media such as Yammer by Microsoft and Workplace by Facebook. In total, the company employed 4,872 employees in 2021, and in total, 3,199 office employees were eligible for participation. Only the office employees relied on ESM for communication, had uninterrupted access to ESM, and were provided resources (e.g. laptop, smartphone) to access the platform. Hence, only office employees received an invitation to participate voluntarily in the survey. Participants were informed about the study and individually provided consent for their participation and processing of their data. A total of 599 employees returned a completed questionnaire, yielding a response rate of 18.7%. Respondents were spread across global locations where the organization had offices. However, about 60% of the respondents were from Finland, where the company is headquartered and has several other locations. Other respondents indicated working in various European countries, the USA, Australia, and Singapore. On average, respondents were 42.32 years old (SD = 10.76), and 60.4% of the respondents were male. Most respondents had obtained a university degree (57.6%) or an applied science degree (27%). Respondents reported an average organizational tenure of 9.63 years (SD = 11.26) and worked 41.12 h per week (SD = 8.37).

3.2 Measures

Table 1 reports the factor loadings and standard errors for all measurement items in the hypothesized model. All items were measured on a seven-point Likert scale unless indicated otherwise.

Negative disconfirmation. Negative disconfirmation refers to the perceived difference between expected and perceived ESM use. We developed four semantic differentials based on risk perception measures by Sitkin and Weingart (1995). Risk perceptions typically focus on perceived opportunity and threat, potential losses and gains, and overall positive or negative assessments of a situation (Sitkin and Weingart, 1995). When applied to communication on ESM, we believe these aspects allow us to gauge perceived negative disconfirmation in ESM communication expectations.

Respondents were asked to indicate what they felt ESM ought to be used: “To what extent do you think [Platform name] should mainly be used to talk about …” We used a seven-point semantic differential scale ranging between (1) threats and opportunities, (2) losses and gains, (3) problems and solutions, and (4) positive and negative events. In addition, we asked respondents to indicate how they felt ESM was currently used: “At this point, is [platform name] mainly used to talk about ….” To calculate the negative disconfirmation, we used the difference between each concurrent set of items between the expected and perceived use of ESM. The four resultant change scores – (1) threats and opportunities, (2) losses and gains, (3) problems and solutions, and (4) positive and negative events – were used in a CFA and regression analysis. Higher scores indicate that the actual use of ESM is more strongly geared toward threats, losses, problems, and negative events, as opposed to opportunities, gains, solutions, and positive events, relative to how individuals feel the ESM ought to be used.

Perceived negativity bias is measured using three items based on van der Meer et al. (2020) and previously reported by van Zoonen et al. (2023). Negativity bias refers to the idea that media content is skewed toward negative information. The aim was not to measure how much of the content is negative but to determine the extent to which these platforms are perceived to be disproportionally flooded with negativity. Sample items include “I think communication on [platform name] is overflowed with obstacles.”

Accountability is measured using four items based on Treem (2015) and van Zoonen et al. (2022a, b). Perceptions of accountability refer to the idea that social media users may feel that they may be held accountable for the information they communicate in online environments. Statements include: “I am worried that I will be held accountable for something I communicated online.”

Social Support was measured using five items by Hammer et al. (2005) and Karatepe (2012). Social support refers to the extent to which respondents feel they receive support from their coworkers. Items include “My coworkers back me up when I need it.”

3.3 Control variables

While not the focal interest of this study, we investigate alternative explanations by studying the impact of several control variables. Before testing the hypothesized effects, we seek to rule out the possibility that the observed relationships are due to variables extraneous to the theorized relationships (Spector and Brannick, 2011). Specifically, we investigate whether the relationship between expectations about content and negativity perceptions may be due to social media use and whether the consequences of negativity perceptions are explained by an individual's tolerance for negative content.

ESM use was measured using four items related to typical social media behaviors. Respondents were asked to indicate the frequency with which they used the ESM. Statements were: “How often do you use [ESM name] to do the following …” (1) post to [ESM name], (2) “read other people's posts”, (3) “comments to other's content,” or (4) “like a post.” The answers ranged between never (1) and multiple times per hour (7).

Tolerance for negativity was measured using two items: “I do not mind if communication on [platform name] is negative” and “Negative issues on [platform name] help us to alert important risks to the company.”

3.4 Data analysis

Before testing our hypotheses, we estimated a measurement model to assess the validity and reliability of the instruments. Model fit was evaluated using the chi-square/df ratio and incremental and absolute fit indices. A χ²/df ratio below three indicates a good model fit. We further report the Adjusted Goodness of Fit (AGFI), the Tucker-Lewis Index (TLI), and the Comparative Fit Index (CFI), for which values above 0.90 indicate good model fit. Finally, a Root Mean Square Error of Approximation (RMSEA) and Standardized Root Mean Square Residual (SRMR) below 0.08 are desired (Hu and Bentler, 1999). These model fit indices are examined for the measurement and structural models. Models were estimated using maximum likelihood, and bias-corrected standard errors for parameters were obtained by extracting 5,000 bootstrap resamples from the data. Furthermore, since our analysis relies on cross-sectional and single-source data, we examined common method variance by conducting Harman's single-factor test. If one factor accounts for more than 50% of the variance in measurement items, common method variance is an issue.

4. Findings

4.1 Measurement model

The confirmatory factor analysis consisting of four latent constructs – i.e. negative disconfirmation, negativity bias, accountability, and social support – demonstrates adequate model fit: χ² (98) = 295.81; χ²/df = 3.02; AGFI = 0.92; CFI = 0.96; TLI = 0.95; SRMR = 0.035; and RMSEA = 0.058 (CI: 0.051, 0.066). We examined the omega (ω) reliability coefficients and the maximum reliability (H). Omega reliability (ω) is preferable to alpha (α) as it tends to produce a more accurate estimation of reliability, for instance, by better accommodating the non-uniformity of factor loadings (for a more thorough discussion, see Hayes and Coutts, 2020). As the values all exceed 0.70, the measurement items are considered to capture our latent constructs reliably. As indicated in Table 2, the ω reliabilities range between 0.70 and 0.91, and Maximum reliability (H) values range from 0.71 to 0.92. Validity was established by examining discriminant and convergent validity. Examination of the measurement model did not indicate substantial cross-loadings of measurement items. Furthermore, the square root of the average variance extracted is greater than the correlation among constructs. Hence, discriminant validity was established. Convergent validity is generally established when the average variance extracted (AVE) is above 0.50. For negativity bias (0.62), accountability (0.65), and social support (0.68), this threshold is met. However, the average variance extracted for negative disconfirmation is 0.30. Notably, the AVE is a conservative measure, and since the ω reliability and the maximum reliability (H) coefficients exceed 0.70, convergent validity can still be assumed (Malhotra and Dash, 2011). Finally, we examined the severity of common method bias. Harman's single-factor test indicated that common method variance was not an issue in the data (27.36%). Hence, no corrective action was taken before estimating the structural model.

4.2 Hypotheses testing

The first hypothesis refers to the discrepancy between what employees think ESM ought to be used for and what they perceive the ESM is currently used for. Respondents were asked to respond based on four semantic differentials. Using paired samples T-tests, we found that employees perceived that ESM was less often used to talk about opportunities (M = 4.71 SD = 1.21) than they felt it should (M = 5.12 SD = 1.17, ΔM = 0.409, t = 6.944, p < 0.001). In addition, the results indicate employees felt the platform was less often used to talk about gains (M = 4.88 SD = 1.24) and solutions (M = 4.49 SD = 1.25) than they felt it should (M_gains = 5.06 SD = 1.13, ΔM = 0.179, t = 3.192, p < 0.001; M_solutions = 5.11 SD = 1.25, ΔM = 0.621, t = 9.531, p < 0.001). There was no significant difference in the extent to which workers felt ESM was used to talk about positive events (M = 4.81 SD = 1.13) versus how often they felt it should be used for that type of communication (M = 4.81 SD = 1.22, ΔM = −0.005, t = −0.084, p = 0.933). Overall, these findings provide partial support for hypothesis 1.

Next, we examined the hypothesized relationships between negative disconfirmation, perceptions of negativity bias, accountability, and social support, as well as the influence of control variables ESM use and tolerance for negative content. The findings indicate that ESM use is negatively associated with negativity bias (B = −0.612 CI95% [−0.899; −0.382], p = 0.002). In addition, ESM use is positively associated with accountability (B = 0.341 CI95% [0.129; 0.718], p = 0.007) but is not significantly related to social support (B = −0.112 CI95% [−0.369; 0.083], p = 0.180). Importantly, none of the hypothesized relationships were affected by the inclusion of ESM use. Furthermore, the results indicate that tolerance for negative content was not significantly related to negativity bias (B = −0.090 CI95% [−0.184; 0.044], p = 0.197), perceptions of accountability (B = 0.009 CI95% [−0.070; 0.077], p = 0.547), or social support (B = −0.010 CI95% [−0.059; 0.051], p = 0.582). Again, none of the hypothesized relationships were affected by the inclusion of tolerance. Hence, the results below are based on the model without these control variables. The model without controls is preferred because there are no strong theoretical arguments to retain the controls, and the model without the controls is more parsimonious.

The retained model demonstrates excellent model fit: χ² (100) = 298.12; χ²/df = 2.98; CFI = 0.96; TLI = 0.95; SRMR = 0.038; and RMSEA = 0.058 (CI: 0.050, 0.065). Figure 1 represents the nomological network with standardized coefficients. Below, the unstandardized solution is provided. Hypothesis 2 reflects the assumption that the negative disconfirmation between user expectations of what ESM ought to be used for and perceptions of what ESM is used for triggers negativity bias. The results indicate that negative disconfirmation is positively associated with negativity bias (B = 0.253 CI 95% [0.156; 0.367], p = 0.001). Hence, hypothesis 2 is supported.

Hypothesis 3 posits that negativity bias is positively associated with accountability and negatively associated with social support. The findings demonstrate a positive association between negative bias and accountability (B = 0.350 CI 95% [0.240; 0.479], p = 0.001). This suggests that employees who perceive communication on ESM to be more geared toward negative content are also more likely to perceive ESM as a technology of accountability. In addition, the findings demonstrate a negative regression coefficient between negativity bias and social support (B = −0.144 CI 95% [−0.232; −0.063], p = 0.003). These findings support hypotheses 3a and 3b.

Finally, hypothesis 4 reflects the indirect effects implied by the preceding relationships. Specifically, hypothesis 4a suggests that a greater negative disconfirmation between expectations and usage perceptions is positively related to accountability through negativity bias. The results demonstrate a significant positive indirect relationship between negative disconfirmation and accountability through negativity bias (B = 0.088 CI 95% [0.052; 0.144], p < 0.001). In addition, hypothesis 4b suggests that negative disconfirmation is negatively associated with social support through negativity bias. The results support this assumption as the indirect relationship is negative and significant (B = −0.036 CI 95% [−0.075; −0.016], p = 0.001). Notably, there is no significant direct relationship between negative disconfirmation and accountability (B = −0.044 CI 95% [−0.167; 0.076], p = 0.461) and social support (B = 0.068 CI 95% [−0.024; 0.168], p = 0.161). These results support the reasoning reflected in hypotheses 4a and 4b.

4.3 Post hoc power analysis

At the time of writing, we are unaware of existing research on negative disconfirmation and perceived negativity bias on ESM in organizational settings. As such, following recommendations by Cohen (2013), we conducted a post hoc power analysis focusing on the explained variances and standardized direct effects ranging between −0.160 to 0.316. A post hoc power analysis revealed that the observed statistical power to detect effects was substantial (0.98). Hence, we conclude that the study was sufficiently powered.

5. Discussion

The findings of this study elucidate some of the antecedents and consequences of negativity bias in the context of ESM. The findings demonstrate that the negative disconfirmation between expectations of ESM use and perceptions of actual ESM use in organizations is an important predictor of perceived negativity bias on the platform. Furthermore, the results show that negativity bias in organizational contexts may have far-reaching implications for perceptions of ESM and coworker relationships. Specifically, we demonstrate that perceived negativity bias is positively related to perceptions of ESM as a technology of accountability and negatively associated with social support.

5.1 Theoretical implications

This study contributes to the theoretical discussion of negativity bias in various ways. First, we extend the scope of research on negativity bias from public social media and mass media research (Knobloch-Westerwick et al., 2020; Rim and Song, 2016; Soroka et al., 2018; van der Meer et al., 2020) and demonstrate that perceived negativity bias is a phenomenon that occurs in organizations and on ESM. We draw on the Disconfirmation Theory (Oliver, 1980) to demonstrate that the sources of perceived negativity bias in employee communication may be rooted in unmet a priori expectations about the nature of communication on an ESM platform. While EDT has primarily gained traction in consumer research to explain consumers' product evaluations (Darke et al., 2010; Venkatesh and Goyal, 2010), the findings of this study demonstrate that disconfirmation may lead to perceived negativity bias. Specifically, the findings demonstrate that the extent of negative disconfirmation – i.e. a negative discrepancy between expected and observed ESM use – is positively associated with a perceived negativity bias. Hence, when employees observe that ESM content is more about threats, losses, problems, and negative events than they expect it to be, this may relate to greater perceived negativity bias. Notably, we propose that it is not the expectation of what the platform should be used for, nor the observed use of the platform, but rather the discrepancy between the two that determines employees' evaluations of ESM content. Hence, EDT is a valuable theoretical framework for understanding the sources of negativity bias in organizations and employee communication.

Second, we explore the potential consequences of perceived negativity bias. Previous studies on ESM have suggested that these technologies can be conceptualized as “technologies of accountability” (Treem, 2015). The underlying argument for this is rooted in the notion that these technologies afford visibility of communication to others. For instance, ESM may “create a visible record of use by individuals, and produce communication that individuals are then accountable for in assessments by others” (Treem et al., 2020, p. 52). We go beyond the notion that visibility creates accountability (Quan-Haase et al., 2005) and suggest we must consider what becomes visible. The findings of this study demonstrate that a perceived negativity bias is associated with the idea of being held accountable and is negatively associated with social support. This suggests that visibility may be particularly problematic when leading to a perceived negativity bias. As such, making negative comments visible may make accountability particularly salient for communicators compared to making positive comments visible through increased negativity bias. As such, we suggest that making anger, frustration, or negative comments about colleagues, policies, or processes visible may contribute to perceptions of ESM as an unsafe space to participate and engage, which might be counter-productive for the success of such platforms. While this falls beyond the scope of the current study, it could be argued that positive disconfirmation and the absence of a negativity bias may not have such detrimental consequences.

Finally, the findings extend research that has suggested that ESM may not support managerial intentions, are met with resistance (Kim, 2018), lead to paradoxical consequences (van Zoonen et al., 2022a, b), and host divergent usage patterns resulting in conflict (Pitafi et al., 2020). Indeed, ESM can be considered a contested domain, and the organizational outcomes need to be more thoroughly understood. This study presents a novel approach to measuring and studying the dissonance between expected and observed uses of ESM platforms and their implications in organizations (e.g. perceived negativity bias). In doing so, this study contributes to advancing empirical and theoretical inquiry into the role of ESM technologies in organizations. For instance, the findings highlight that a negative disconfirmation of expectations about ESM use may have implications beyond technology use as it is positively related to feelings of accountability and negatively associated with social support through negativity bias.

5.2 Practical implications

The findings of this study provide valuable guidelines for practitioners. Many organizations face challenges regarding the under-utilization of ESM platforms (Sun et al., 2020). This study demonstrates that negative disconfirmation between expected use and observed use of ESM may create an environment less conducive to contributions and information sharing. As a result of negative disconfirmation, employees may view communication environments as arenas of accountability without social support. Therefore, employees and managers need to manage user expectations. This may involve greater transparency about the nature of the information shared on ESM, as well as active content moderation and communication of the purpose of ESM platforms. The notion that information remains visible long after their initial communication and beyond intended audiences warrants careful content curation. In that regard, managers may introduce content management functions to ESM, such as showing summaries of threads, pinning or highlighting solutions, or promoting new initiatives (Chen and Wei, 2019).

More broadly, this study sheds light on the presence and implications of perceived negativity bias. For organizations and managers, it is important to recognize that ESM platforms may not be used in the ways initially intended. Unmet expectations, divergent uses, or emotional communication may lead to feelings of increased accountability and low social support. Ultimately, this may obstruct information sharing and aggravate knowledge disparity or knowledge hoarding. As such, it may be helpful to examine usage patterns and provide feedback while also establishing clear policies and guidelines that help align usage patterns with organizational values and manage the expectations of users.

5.3 Limitations and future research

Several limitations of the study need to be acknowledged. First, the study has limitations inherent to any cross-sectional analysis. For instance, it is impossible to draw conclusions about the directionality or causality of the relationships under study. In addition, we relied on self-reported data, increasing the possibility of the data being subjected to bias. Future studies could draw on longitudinal study designs and rely on multi-sourced data to address these caveats. For instance, rather than examining perceived negativity bias, content and sentiment analyses of ESM data could help to reduce self-report biases and examine ESM usage patterns more directly.

Second, the generalizability of the findings should be explored in other organizations. The data underlying this study was collected at one organization that operated globally. Future studies may investigate the pattern of results in a diverse set of organizations, including those operating locally or with workforces that may be less dispersed. It might be possible that employees who work in closer physical or cultural proximity may experience fewer discrepancies or are better equipped to resolve such discrepancies and prevent adverse outcomes. Further research is needed to examine the robustness and generalizability of our findings to other organizations and under various conditions.

Third, we report a response rate of 18.7%. While we have taken measures to reduce nonresponse, such as advance survey notice, personalization, reminders, and organizational endorsement, this response rate is on the lower end of response rates commonly reported in organizational studies (Anseel et al., 2010). However, we note that organization studies frequently report response rates as low as 5% (e.g. Kulik and Perry, 2008; Sikora et al., 2015), the survey distribution in this study was considered a good reflection of the population, and the responses to the statements seem to range widely across answer anchors, minimizing concerns over sampling bias (Blair and Zinkhan, 2006). Yet, future studies may consider sampling procedures that have proven to yield higher response rates than online surveys, such as in-person and on-site data collection.

Fourth, this study reveals that discrepancy is correlated with perceived negativity bias, which is, in turn, positively related to accountability and negatively related to social support in organizational contexts. However, the findings of this study provide little insight into the contextual factors under which these relationships may be more or less profound. Future studies could examine how individuals' regulatory focus and digital or professional proficiency may impact the relationships central to the present study.

Finally, we explored the correlates of a perceived negativity bias, measured as the perception that content on ESM is disproportionately geared toward negative communication. Future studies may also directly explore negativity bias–i.e. negative communication attracts more attention than positive communication (Rozin and Royzman, 2001). For instance, future studies may investigate how negativity bias is associated with content selection, avoidance (van der Meer et al., 2020), and processing (Rozin and Royzman, 2001) on ESM.

6. Conclusion

In conclusion, this study explores the antecedents and consequences of perceived negativity bias on ESM. The study demonstrates that the discrepancy between expectations and perceptions of actual ESM use is an important predictor of perceived negativity bias on the platform. This advances the Expectation-Disconfirmation Theory in the context of ESM technologies, which facilitate greater visibility of communication and behavior. We demonstrate that negative disconfirmation may trigger perceived negativity bias in organizations and highlight that perceived negativity bias may have far-reaching implications for perceptions of ESM and coworker relationships. The findings provide valuable guidelines for practitioners, and it is important for employees and managers to manage user expectations, introduce content management functions to ESM, and be aware of the presence and implications of perceived negativity bias.

Figures

Figure 1

Nomological Network with standardized coefficients

Table 1

Measurement Items and factor loadings

Measurement items	Mean (S.E.)	R2	St. Factor loading	Unst. Factor loading	Se
Negative disconfirmation
To what extent do you think [ESM] should mainly be used to talk about …. a	–	–	–	–	–
The next questions are about how you think [ESM] is currently being used. At this point, is [ESM] mainly used to talk about … a	–	–	–	–	–
Opportunities – Threats	0.41 (0.06)	0.43	0.659	1.000b
Gains – Losses	0.18 (0.06)	0.38	0.619	0.893	0.09
Solutions – Problems	0.62 (0.07)	0.37	0.611	1.026	0.10
Positive events – Negative events	−0.01 (0.06)	0.29	0.541	0.826	0.09
Negativity Bias
I think communication on [ESM] is generally skewed toward the negative rather than the positive	2.76 (0.05)	0.63	0.795	1.000b
In my opinion, communication on [ESM] is disproportionally about problems within [company]	2.71 (0.05)	0.71	0.843	1.045	0.06
I think communication on [ESM] is overflowed with obstacles	3.17 (0.06)	0.51	0.712	1.077	0.07
Accountability
I am worried that online, I appear less professional than offline	2.86 (0.05)	0.56	0.745	1.000b
I fear that other organizational members might misinterpret my online communication	3.31 (0.07)	0.60	0.775	1.164	0.06
I am anxious about not knowing who may be viewing my online communication	2.84 (0.06)	0.76	0.869	1.253	0.06
I am worried that I will be held accountable for something I communicated online	2.97 (0.07)	0.70	0.834	1.259	0.06
Social Support
I receive help and support from my coworkers	5.82 (0.04)	0.63	0.795	1.000b
I feel I am accepted in my workgroup	5.87 (0.04)	0.75	0.868	1.089	0.05
My coworkers are understanding if I have a bad day	5.31 (0.05)	0.51	0.712	0.997	0.05
My coworkers back me up when I need it	5.23 (0.05)	0.73	0.856	1.206	0.05
I feel comfortable with my coworkers	5.85 (0.04)	0.75	0.867	1.072	0.05

Note(s): All factor loadings are significant at p < 0.05. ^aThese statements represent the prompts to inquire about expectations and usage perceptions. The factor loadings are based on the difference scores taken from the semantic differentials listed here. ^bUnit loading indicator constrained to 1

Source(s): Authors’ own work

Table 2

Reliability and validity statistics

	Mean (Sd)	MaxR (H)	ω	AVE	MSVa	1	2	3	4	5	6	7	8	9
1. Negative disconfirmation	0.30 (1.06)	0.71	0.70	0.30	0.06	0.79
2. Negativity bias	2.88 (1.10)	0.84	0.82	0.62	0.10	0.253*	0.55
3. Accountability	2.99 (1.32)	0.89	0.88	0.65	0.10	0.038	0.317*	0.81
4. Social Support	5.67 (0.96)	0.92	0.90	0.68	0.05	0.035	−0.163*	−0.228*	0.82
5. ESM use	1.60 (0.52)	–	–	–	–	−0.050	−0.211*	0.043	0.017	–
6. Tolerance	4.17 (0.85)	–	–	–	–	0.112*	−0.312*	−0.125*	0.065	0.086*	–
7. Genderb	0.40 (0.49)	–	–	–	–	−0.003	−0.169*	−0.047	−0.026	0.098*	0.085*	–
8. Age	42.32 (10.76)	–	–	–	–	0.136*	0.082*	−0.059	−0.004	−0.007	−0.073	−0.131	–
9. Work hours	41.12 (8.37)	–	–	–	–	0.003	−0.019	0.034	0.008	0.066	−0.060	−0.116*	0.020	–
10. Tenure	9.63 (11.26)	–	–	–	–	0.139*	0.114*	−0.049	0.017	−0.029	−0.071	−0.105*	0.589*	−0.045

Note(s): Ω = Omega Reliability; MaxR(H) = Maximum Reliability H; AVE = Average Variance Extracted; MSV = Maximum Shared Variance; the Square Root of the AVE is reported on the diagonal. Reliability and validity statistics are only provided for latent constructs retained in the final measurement and structural model. Significant correlations are flagged *. ^bGender was coded 0 = male, 1 = female

Source(s): Authors’ own work

We have no known conflict of interest to disclose.