Mood matters: the interplay of personality in ethical perceptions in crowdsourcing

ABSTRACT

This research delves into the intricate interplay between mood and ethical perceptions within the context of crowdsourcing online labour markets. The study employs a carefully crafted experimental design, conducted in November 2021, involving two distinct groups: the benchmark group, serving as the control, and the treatment group, subjected to mood induction through film exposure. The benchmark group is presented with a neutral placebo film, while the treatment group is treated to a comedy film. By using OLS estimation methods, the paper provides evidence of the impact of positive mood manipulation, which remarkably enhances individuals’ ethical perceptions, fosters value co-creation, and augments the quality of their relationships with the platform. The study's findings strongly indicate that the potent effect of positive mood primarily operates through the lens of the agreeableness trait. This observation sheds light on the intricate psychological mechanisms underlying mood's influence on various outcomes within the online labour market setting. Contributions to the ethical perception, mood research and online-economy literature are discussed.

KEYWORDS:

• ethical perception
• value co-creation
• mood
• personality
• online labour markets
• crowdsourcing

JEL CODES:

• D91
• D87
• D53
• D23
• D01

1. Introduction

When you’re in a good mood, the world appears to be a better place overall. Conversely, feeling disappointed or dejected can lead to added stress in both your body and mind. Seminal research by Wilhelm and Schoebi (2007) reveals that mood is not solely an emotional experience but also influences one’s perception of the world. However, does this phenomenon extend to the digital world, and if so, to what extent? In this context, the ‘digital world’ refers to the realm of digital tools used for internet communication, including digital devices, smart gadgets, and various online communities like online labour markets (OLMs), sharing economy platforms (SEPs), and social networks (Plant 2004).

The study is oriented toward online labour markets (OLMs) and how workers regulate their behaviour in relation to their ethical perceptions and co-value intention.¹ The labour procedure within this context, is called ‘crowdsourcing’ and is the act of taking a job traditionally performed by a designed agent (e.g. employee) and outsourcing it to an undefined, generally unknown large pool of people in the form of an open call. This new form of labour, as a term, is a strategic model to attract an interested, motivated crowd of individuals capable of providing solutions superior in quality and quantity to those that even traditional forms of business can and has become a new, complementary form (not necessarily substitute but in many cases supplementary) to traditional employment as companies seek to tap the latent talent of the crowds (i.e. wisdom of crowds) (Mourelatos and Tzagarakis 2018). To accommodate the increasing online labour force, numerous platforms have emerged over the years, such as Amazon Mechanical Turk, Microworkers, CrowdFlower etc (Mourelatos, Tzagarakis, and Dimara 2016).²

Although many studies have focused on several aspects of the crowdsourcing process and the need for the standardisation of labour on the part of the requesters, research with individual studies on workers has not yet received much scholarly attention (Blohm et al. 2018). Preliminary evidence on workers’ working behaviour suggests that their performance depends on demographics, human capital, income-related factors and cognitive skills (Autor and Handel 2013).³ However, behaviour also relies on psychology. For that reason, economists and psychologists are currently trying to develop a deeper understanding of how different personality traits and emotions as indicators for individual-specific soft skills may affect individuals’ preferences and perceptions within their working environment (Heckman, Jagelka, and Kautz 2021; Manganari et al. 2022). Both common sense and scientific psychology assume that mood and personality can have strong effects on thought and action. Indeed, the role of these psychological variables is well established in standard models of individual behaviour regarding task performance by adopting personality and mood inventories capturing individual-specific differences in the ways of thinking, feeling, and behaving (Mueller and Plug 2006; Müller and Schwieren 2017). Roberts (2009) describes these traits as the relatively enduring patterns of thoughts, feelings and behaviours that reflect the tendency to respond in various ways under certain circumstances. Against this backdrop, the same effort in online labour working platforms produced evidence that workers’ personality traits and current emotions correlate with their performance, working engagement, co-creation intention (Kazai, Kamps, and Milic-Frayling 2011; 2012; Mourelatos, Giannakopoulos, and Tzagarakis 2022), incentives (Mourelatos, Giannakopoulos, and Tzagarakis 2023) and hiring choices (Mourelatos 2023).

Workers on online labour markets often perceive the crowdsourcing procedure as unethical due to a range of additional reasons. These may include exploitative working conditions, lack of transparency in task requirements, ambiguous ownership of intellectual property rights, and limited opportunities for career advancement or skill development. Furthermore, concerns about data privacy and security, unequal power dynamics between workers and platform owners, and the potential for intellectual property theft may also contribute to their negative perception of the ethics involved in crowdsourcing practices (Schlagwein, Cecez-Kecmanovic, and Hanckel 2019). These ethical concerns may impact workers’ motivation and willingness to actively participate in crowdsourcing activities. To foster trust and alleviate these concerns, platforms can implement measures and mechanisms to address workers’ grievances. Α good mood can be a key factor in fostering trust among online workers in the crowdsourcing process.

In this direction, this study aims to address two key research questions: (1) What are the personality dimensions affecting workers’ ethical perceptions of OLMs? and (2) How does workers’ current mood relate to their ethical perceptions and intentions for value co-creation within OLMs?

To investigate this, several experimental sessions in Amazon Mechanical Turk (i.e. AMT) were conducted. AMT is one of the leading OLM and the most representative field for exploring issues related to the online economy (Dube et al. 2020). An induced positive mood stimulus was used, to explore workers’ behavioural change in terms of ethical perceptions and preferences (Gross and John 2003; Isen and Shalker 1982). By analyzing experimental data (N = 320) from online workers using regressions, the results confirm the essential role of a worker’s personality traits on their ethical perceptions and the underlying mechanism of co-creation intention. The study also experimentally supports that workers’ mood is sensitive to fairness perceptions.

This paper significantly contributes to the literature on OLMs, serving as one of the pioneering investigations into workers’ psychological aspects and their influence on ethical perceptions and co-creation intentions. By doing so, it adds to the expanding knowledge base in this domain. The study provides valuable insights to managers on the formation of workers’ ethical perceptions and the factors of significance, offering practical guidance to enhance engagement and value co-creation in OLMs.

2. Background

According to Horton and Chilton (2010), OLMs consist of three parts: a requester of a job, an online task or job and a pool of potential workers. An online labour market is where (1) labour is exchanged for money, (2) the outcomes of that labour are delivered ‘over a wire’ and (3) the allocation of the labour and the money is determined by a collection of requesters and workers operating within a particular price system (Horton 2010). Hence, economic activities have become increasingly digital, since hundreds of millions of internet users are using crowdsourcing platforms either to work at an online job as workers or as a model of problem-solving and production as requesters.

Primarily, online labour markets embed a business model that brings individuals together to participate and create value though a labour process (Bai and Yu 2021; Oppenlaender et al. 2020; Yin et al. 2022). Thus, OLMs can be a valuable and powerful tool for marketers within various types of firms (Whitla 2009). This necessary interaction between requesters and workers and its financial aspects (e.g. transactions, security, privacy, etc.) is being fully covered by OLMs in terms of ownership (Yuksel, Darmody, and Venkatraman 2019). Hence, the online platform mediator has a key role in enhancing the honesty, trust and ethical balance behaviour leading to exchange (Perren and Kozinets 2018). However, the participant parts (i.e. requesters, platforms, workers) of the online process don’t feel the same standards of fair labour (Schmidt 2013). When crowd work is not just an occasional pastime but becomes the reality of a daily workplace, an ethical debate about what conditions we regard as appropriate or acceptable becomes more pressing (Standing and Standing 2018). Even the best-known and historically respected firms in offline and online marketplaces have suffered from ethical lapses.⁴ This means that on the one hand, workers especially perceive participation in OLMs to be more convenient and easily accessible and understandable, but on the other hand, workers may develop unfair and untrustworthy labour strategies that might deter their participation on such platforms and decrease their willingness and intention for co-creation (Oppenlaender et al. 2020; Zheng, Li, and Hou 2011).

Research that touches upon ethics in online labour markets has revealed that crowd labour can prove to be astoundingly cheap and unfair. Workers may work without benefits and job security by being exposed to every requester’s continuous requirements throughout the labour process (Felstiner 2010).⁵ Moreover, workers in particular subsections of the paid crowdsourcing industry may be denied the protection of employment laws without much recourse to vindicate their rights (Dawson and Bynghall 2012). In the same direction, Brabham (2013), highlighted legal issues (e.g. intellectual property) and labour exploitation (e.g. low pay) in crowdsourcing, emphasising the importance of satisfied crowd workers (Chan, Li, and Zhu 2015). Last, McInnis et al. (2016), raised concerns about workers’ welfare, reputation, fairness and abuse. These concerns are exacerbated by OLMs’ hands-off approach to the labour market.⁶ For example, AMT’s participation agreement classifies Turkers as independent contractors free to accept any task they qualify for. At the same time, requesters have the right to reject a Turker’s completed work without payment, while AMT, providing only the venue for an exchange, is not involved in resolving any labour disputes. When a Turker’s work is rejected, the result is lost pay, time, and reputation, and AMT’s stance gives workers little recourse. These policies make the practice of crowd work risky (Kokkodis and Ipeirotis 2016; Thuan, Antunes, and Johnstone 2016).

Although many defendants of crowdsourcing claim that existing federal laws do not apply to crowdsourcing, a further investigation of the determinants that influence workers’ ethical perceptions must take place to address the abovementioned ethical challenges and understand in depth what crowd workers think about this online cocreative work (Oppenlaender et al. 2020). There is much evidence that workers’ concern for ‘fair’ transactions influences their labour market behaviour and performance (Benjamin 2015). For example, Kahneman, Knetsch, and Thaler (1986) demonstrate that fairness explains sharing behaviours within internal labour markets. Bewley (2009) also suggests that workers’ mood and feelings about fairness could explain why firms typically lay off workers rather than reduce wages: still-employed workers would consider wage cuts unfair and become less productive.⁷

The real-world relevance of studying the role of mood in crowdsourcing cannot be overstated, as it directly impacts the dynamics of online labour markets and the individuals involved. Understanding the influence of mood on workers’ ethical perceptions and value co-creation intentions is crucial for several reasons. Firstly, worker engagement and motivation can be better supported. In the fast-paced and competitive world of crowdsourcing, worker engagement and motivation play a pivotal role in the success of online labour markets. Positive moods can act as catalysts, boosting workers’ enthusiasm and dedication to perform tasks diligently. For example, when workers encounter an enjoyable and humorous task description, their positive mood may translate into higher levels of effort and commitment to complete the task effectively. Secondly, adherence to moral principles, ethical standards, and trust is further enhanced. The perception of ethical practices within the crowdsourcing ecosystem significantly impacts workers’ trust and willingness to participate. Mood can influence how workers interpret the fairness of compensation, task requirements, and overall treatment by the platform. For instance, a worker in a positive mood may feel more inclined to trust the platform when they receive timely and transparent feedback on their work, leading to increased satisfaction and loyalty. Third, quality of outputs increases. The quality of outputs in crowdsourcing heavily relies on the mindset and mood of workers. Positive moods can lead to enhanced creativity, problem-solving skills, and innovative solutions. Conversely, negative moods, such as frustration or disappointment, might hinder productivity and negatively affect the quality of the work delivered. Lastly, platform reputation improves. Crowdsourcing platforms with a reputation for fostering positive moods and ethical practices are likely to attract a larger pool of talented and committed workers. Word-of-mouth referrals and positive online reviews from content workers contribute to a platform’s positive image, which, in turn, can lead to increased business opportunities and growth.

For that reason, the purpose of the current study is important and twofold. First, the paper adds additional evidence on the ethical debate of OLMs by exploring workers’ personality characteristics and their correlations with their ethical perceptions and co-creation intention (Agag et al. 2016; 2019) and building on existing models of fairness concerns (Charness and Rabin 2002; Fehr and Schmidt 1999), which are based on worker ethical judgments. Furthermore, to explain the dynamic behaviour of workers in terms of ethical concerns, the paper also explores how their current mood is associated with their fairness and ethical preferences.

3. Theoretical background and hypotheses development

3.1. Mood

In general, mood⁸, defined as diffuse or global feeling states, can lead someone to take self-regulatory action designed to maintain them (in case of good mood) or eliminate them (in case of bad mood), with direct results in his or her social behaviour (Morris and Reilly 1987). While many studies also frequently use the term ‘emotion’, the construct of interest in my case is an affective feeling state that may vary in intensity from mild to intense, and for that reason, this paper has chosen to make little distinction among emotions and mood because their boundaries are ‘unsharp’ (Frijda 1993). Mood has consistently emerged as two dominant and relatively independent dimensions: the positive affect dimension and the negative affect dimension (Watson, Clark, and Tellegen 1988). It is already well documented that an individual’s mood state affects his judgments and way of thinking (Forgas 1995; Lerner and Keltner 2000). Although several interpretations have been proposed, none can yet fully account for the varied empirical evidence now available (Faber and Christenson 1996; Forgas and Bower 1987). This inability is at least partly due to a failure to embed in a more general theory of social judgment that specifies how mood affects the underlying mechanism of an individual’s judgment process under different offline and online conditions.

Many studies to tackle this issue have already explored this mechanism in offline settings (i.e. survey and labouratory) by experimentally embodying mood-inducing events and monitoring participants’ behavioural responses. For example, Isen and Shalker (1982) found that induced negative mood led to lower pleasantness ratings for pleasant, ambiguous, and unpleasant slides, while induced positive mood led to higher pleasantness ratings. However, the idea of ‘managing’ negative emotions is complex. For that reason, research has mainly focused on positive mood stimuli and their effects (Lyubomirsky, King, and Diener 2005). Fredrickson (2001), in his paper, predicts that when you’re in a good mood state, your attention zooms out, ‘paying attention to the globality of concepts, situations, or objects’ and looking at things around you in a positive light results in a more creative mindset (Tugade and Fredrickson 2004). Moreover, Vanlessen et al. (2016) revealed that individuals with high levels of positive mood think imaginatively or, in other words, may simply be less able to think analytically.

The emerging field of mood regulation studies has also expanded in the field of economics and marketing and has also pointed out that mood is central to the actions of consumers and managers alike (Bagozzi, Gopinath, and Nyer 1999). Many studies conducted in traditional offline labour markets and labouratories revealed that a positive mood results in higher productivity (Oswald, Proto, and Sgroi 2015), higher consumption (Winterich and Haws 2011), higher elasticity (Di Tella, MacCulloch, and Oswald 2001) and higher abstract construal behaviour (Labroo and Patrick 2009).

In our case, online labour platforms differ significantly from conventional markets in their organisational structure and workforce composition. Unlike traditional markets, crowdsourcing taps into a vast network of remote and often anonymous contributors who work independently. In this context, mood and emotions become even more critical factors that impact both online and conventional markets. In traditional markets, mood can influence decision-making, negotiation outcomes, and customer interactions, affecting overall business success. Similarly, in online labour markets, where workers are dispersed and lack direct supervision, mood becomes a crucial determinant of task completion rates, task quality, and worker retention (Gadiraju and Demartini 2019). Maintaining a positive emotional climate is essential in both realms to foster productive and satisfied individuals, ensuring optimal performance and sustainable growth in the long run.

Regarding the investigation of ethics, theoretically, the prevalence of mood in individuals’ ethical decision process has been identified by researchers from various philosophical perspectives, but only in offline contexts and empirically through experimental designs. Practically, mood is often considered a nonessential aspect of the ethical decision process that is best ignored, if not controlled, as it interferes with a logical, rational, ethical decision process (Gaudine and Thorne 2001).

By taking into consideration all the above mentioned studies and that the role of positive affect mainly enhances the positivity of recall, judgment and social behaviour (Clark and Isen 1982; Isen 1984), this study experimentally manipulates positive mood state and explores how it influences individuals’ ethical decision process and value co-creation in relation to their personality traits (Reisenzein and Weber 2009) under real online working conditions within an OLM environment.

3.2. Ethical perceptions and value co-creation intention

In general, an ethical perception can occur when a situation is viewed as an accepted and institutionalised part of doing business. Bartels (1967) was the first to provide a conceptual framework of factors that influence ethics in decision-making marketing processes. Research has become more interested in the topic and has steadily contributed to several aspects of unethical behaviour in marketing contexts.⁹ In parallel, practitioners also had an engagement with marketing ethics, and companies started to adopt certain codes of ethics in their operations and development.

With the establishment of Web 1.0¹⁰ and its expansion to Web 2.0,¹¹ a part of the traditional offline labour market has rapidly transformed to its online form, including sharing and gig economy platforms¹² (Autor 2001). Furthermore, it is widely recognised among scholars that ethical aspects differ in offline and online environments, and platform participants’ ethical perceptions are formed in different ways in relation to traditional settings (Roman 2007). Additionally, the internet is generally a widespread environment for unethical behaviour (Freestone and Mitchell 2004; Hajli 2018).

In the case of OLMs, several idiosyncratic features of online markets create concerns about unethical worker behaviour. OLMs differ from traditional labour firms, and that creates novel ethical challenges, specifically in terms of interaction. For example, OLMs do not offer systematic tools for dispute resolution between requesters and workers, creating a low perception of fairness (Fieseler, Bucher, and Hoffmann 2019). Thus, workers’ fairness perceptions are shaped directly by requesters’ behaviour (e.g. unjustified rejection of work, low pay, etc.). On the other hand, many researchers have investigated workers in terms of honesty, social ties to their employers and the dark triad of their personality, such as machiavellianism (Brink et al. 2019). A basic outcome is that online workers face greater distractions and often work in uncontrolled settings (Farrell, Grenier, and Leiby 2017) because OLMs (1) do not always ensure sound transactional processes, (2) do not efficiently prevent abusive behaviour, and (3) do not act as arbitrators in cases of conflict.

Thus, as research on OLMs has just begun to emerge, it is of paramount importance that we understand the link between the psychological facets and ethical perceptions of online workers to make further steps within the concept of emotional intelligence (i.e. mood, personality) on the perception of the ethical behaviour of workers in online labour environments (Fieseler, Bucher, and Hoffmann 2019; Joseph, Berry, and Deshpande 2009).

Recall that OLMs and crowdsourcing are relatively recent concepts that encompass many practices. This diversity leads to the blurring of the limits of this new online labour that may be identified virtually with any type of internet-based collabourative activity, such as co-creation (Estellés-Arolas and González-Ladrón-de-Guevara 2012). Value co-creation has recently emerged as a major strength for a business and refers to adopted strategies or initiatives that bring different parties together to produce valued outcomes.¹³ This approach is harnessed by companies not only to gain a competitive advantage but also to gain a corporate reputation and brand value (Merz, Zarantonello, and Grappi 2018; Prahalad and Ramaswamy 2004). In general, service-dominant logic (SE logic) implies that firms offer value propositions, and this value is cocreated and subjectively determined as value-in-use, while consumers are at the forefront of cocreating value with companies (Vargo and Lusch 2008). This comprehension led several firms to offer their consumers a more active role and engagements within the development phase of their services and products (Prahalad and Ramaswamy 2004).

Within the crowdsourcing context, the value created for the firm is a function of the alignment of strategic objectives, network effects (whether a growing crowd of contributors attracts additional customers), crowdsourcing-related costs, and risks associated with possible opportunistic behaviour. Meanwhile, little is known about the value created for and appropriated by the crowd workers. This is an important gap in our knowledge because, for many specialised tasks, access to high-caliber workers may be limited and competitive. OLMs that manage to build an engaged crowd are hard to imitate and enjoy a resource-based competitive advantage. Thus, a new model of value for the worker side is needed. In building such a model, we embrace a stakeholder perspective on value co-creation and explore the psychological and contextual factors determining how workers derive value from participation in OLMs.

In relation to mood, while its influence on individuals’ ethical decisions has been identified by several studies, little is known about how mood influences individuals’ ethical decision processes and co-creation intention (Gaudine and Thorne 2001), especially in online working environments (Zhuang and Gadiraju 2019). In general, behavioural models that have also been developed with psychological perspectives conclude that mood state mainly influences the individual’s propensity to identify ethical dilemmas and leads to ethical decision choices that promote an individual’s compliance with his or her prescriptive beliefs, according to his or her personality facets and cognitive moral structures (e.g. as a neurocognitive model) (Reynolds 2006). Hence, it is expected that workers experiencing high arousal from the positive affect treatment will change their ethical decision process and co-creation intention. Thus, it is hypothesised the following:

H1. Individuals experiencing positive affect treatment are more likely to select an ethical decision choice consistent with his or her prescriptive judgment. Thus, in general, we will observe behavioural changes in terms of ethical perception and co-creation intention.

More concretely, within an online context such as an OLM, it is already established that workers have mainly extrinsic motivation (i.e. salary-related incentives) to increase their working activity through co-creation intention (Fedorenko and Berthon 2017). Moreover, many aforementioned studies on mood have shown that positive feelings increase online and offline working performance (Oswald, Proto, and Sgroi 2015). Additionally, when individuals perceive themselves to be more productive, they are ‘more ethical’ (Caza, Barker, and Cameron 2004). Thus, it is expected that positive affect manipulation will increase individuals’ co-creation intention and ethical standards. Thus, it is hypothesised the following:

H2. Individuals being affected the most by positive mood affect manipulation treatment will have a less conservative, ethical decision-making behaviour but more creative cognitive style resulting in higher co-creation intention and ethical perception index values.

3.3. Relationship quality theory

By taking into consideration the aspects of the labour being conducted in OLMs, the study also embeds in the analysis indicators reflecting the relationship between the online platform and the workers. For that reason, relationship quality comes under the umbrella term of this relationship in terms of marketing. The aim of this theory relies on the formation of partnerships between the service providers (i.e. OLMs) and the participants, resulting in value co-creation for both requesters and workers (De Cannière, De Pelsmacker, and Geuens 2009; Yu et al. 2020). Potential crowd workers mainly enter into interactions with OLMs because they expect to receive positive value from their participation (Horton and Chilton 2010).

The vast majority of studies related to relationship quality have been conducted in offline contexts (Athanasopoulou 2009). In addition, Walsh et al. (2010) reveal that relationship quality is equally important in online and offline settings for the sake of retaining individuals and that the differences in the impacts of relationship quality dimensions are very context specific. In our case, in the OLM setting, workers face several issues related to their vulnerability when making contracting decisions due to the high degree of uncertainty, which arises mainly from information asymmetry. Most workers are now present online, and therefore, it is vital that we understand and study the concepts of relationship quality online, specifically in the context of online labour crowdsourcing markets.

Thus, to describe how strong the relationship quality is between OLMs and workers and the degree to which this relation meets the expectations and needs of the workers’ side, trust, satisfaction and commitment were measured, which are proven to be the most influencing factors of workers’ performance (Nadeem et al. 2020). According to Moorman, Deshpande, and Zaltman (1993) and Gustafsson, Johnson, and Roos (2005), in an exchange relationship, the levels of commitment, trust and satisfaction have key roles in keeping, maintaining and continuing the valued relationship. This eventually leads to an increasing comfort level, resulting from individuals’ increased engagement with the online working labour market (i.e. OLM) (Martin 2018; 2019). Consequently, several studies have revealed a positive relationship between consumers’ ethical perceptions, co-creation intention and commitment, trust and satisfaction in e-retail settings (Elbeltagi and Agag 2016) and online sharing economy contexts (i.e. SEP) (Nadeem et al. 2021; 2020). Hence, it is of high importance that this relationship is studied in the OLM context, in terms of whether a worker’s commitment, trust and satisfaction is fostered by his ethical perceptions and co-creation intention. Thus, it is hypothesised the following:

H3. Individuals’ trust, commitment and satisfaction levels will increase, resulting in higher engagement with the OLM, due to the positive mood-induced event.

3.4. Personality traits

Research has consistently revealed a significant link between an individual's personality and their online behaviour. Personality emerges as a pivotal factor that influences how users interact and conduct themselves on the internet (Amichai-Hamburger and Vinitzky 2010) and Amichai-Hamburger (2002). In online labour markets, where internet-based platforms serve as the foundation for work arrangements, the personality of workers becomes a decisive factor in shaping their working behaviour. The virtual nature of these platforms creates unique challenges and opportunities for workers to engage with requesters and colleagues. Requesters often seek to assess a worker's personality traits to ensure a good fit for specific tasks and projects, highlighting the vital role of personality in predicting and understanding workers’ behaviour in online labour markets (Mourelatos, Giannakopoulos, and Tzagarakis 2022).

From the abovementioned personality traits, the trait of neuroticism is defined as a lack of emotional stability due to the presence of anxiety, insecurity, risk aversion or other negative psychological state. Neuroticism has been consistently found to negatively affect workers’ job outcomes (Cubel et al. 2016), and it is highly related to deviant workplace and anti-social behaviour (Salgado 1997). Additionally, some of the mechanisms at play in labour relations, such as lack of self-confidence, are likely to operate as well in our experimental setting. In addition, it has already been established that fewer ethical beliefs and perceptions are correlated with behaviours with a high propensity for making risky decisions at work (Bratton and Strittmatter 2013). Hence, it is expected that positive mood treatment will have a significant influence on their psychological state and on their way of thinking (Rusting 1998; 1999). Thus, the hypothesis is that high levels of neuroticism should be correlated with low co-creation intention and high ethical perception and relationship quality levels before my positive mood manipulation but with a higher co-creation intention, ethical perception and relationship quality levels after the positive mood-inducing event. Thus, it is hypothesised the following:

H4. Neurotic individuals will have a more analytical decision-making behaviour, and thus neuroticism will have contradictory pre- and post-effects on ethical perception index, co-creation intention and relationship quality due to the positive mood manipulation.

Agreeableness describes a person’s ability to put other people’s needs above his or her own and includes attributes such as trust, altruism and cooperation (Graziano and Tobin 2002). For instance, individuals with high agreeableness tend to obtain pleasure from being part of a community and contributing to its further development. Agreeableness is linked to socially valued traits and prosocial motives (Graziano and Tobin 2002). Moreover, agreeable people are greatly motivated to maintain a positive state, and this motive system induces an agreeable person to generate, in general, a positive perception and attributions to several relationship contexts (Jensen-Campbell and Graziano 2001).

In the same direction, extraverted individuals, in general, enjoy being in social situations, have a high need for social desirability and have an ambitious nature (Watson and Clark 1997). Moreover, several studies have revealed that individuals with high levels of extraversion are characterised by positive affect, high energy and assertiveness (Rusting 1998). Like neuroticism, extraversion was conceptualised prior to the advent of the Big Five and thus has a history of investigation with respect to dishonesty and unethical behaviour (e.g. Giluk and Postlethwaite 2015; Rallapalli et al. 1994). The excitement-seeking facet of extraversion provides the strongest rationale for linking this trait to cheating and the tendency of extraverted individuals to have ‘more unethical’ beliefs concerning working behaviour (Oehler and Wedlich 2018).

Thus, it is expected that positive mood treatment will further boost the agreeables and extraverted nature of the individuals (Chirico, Shiota, and Gaggioli 2021). The hypothesis is that high levels of agreeableness should be correlated with high value co-creation intention, ethical perceptions and relationship quality before our positive mood manipulation. The effects will be more robust after my positive mood-inducing event. Extraversion should follow the same trend, but with small differences in T2. Thus, it is hypothesised the following:

H5a. Agreeables and extraverted individuals will have a more stable decision-making behaviour, and these personality traits will have positive effects on the ethical perception index, value co-creation intention and relationship quality indicators.

H5b. The magnitude of the abovementioned effects will be boosted further after the induced positive mood manipulation, mainly for agreeableness.

4. Methodology

4.1. Amazon Mechanical Turk

Recall that the experimental sessions were conducted in Amazon Mechanical Turk. Amazon Mechanical Turk (MTurk) is widely popular for experiments due to its vast and diverse global workforce, offering researchers access to a representative sample efficiently and at a low cost (Johnson and Ryan 2020). Its user-friendly interface enables researchers to create tasks easily, while workers can quickly participate. Additionally, MTurk provides valuable demographic and behavioural data about workers, enhancing the study's precision. Furthermore, built-in quality control measures ensure data reliability, solidifying MTurk’s reputation as one of the best online labour markets for experimental research. This OLM efficiently replicates the principles of a real labour market (i.e. offline labour markets) and where buyers contract with individual sellers (Horton 2010). Hence, the bibliography shows that Mechanical Turk, which has been operated since 2005 by Amazon, is the most well-known crowdsourcing platform and is well-spread in the research body for having the major elements required for experimental research (Horton, Rand, and Zeckhauser 2011; Paolacci, Chandler, and Ipeirotis 2010). Moreover, many studies have shown that this online labour marketplace consists of a representative large pool of workers more similar to the U.S. population than in-person convenience samples (Mason and Suri 2012). Thus, the way that Amazon Mechanical Turk is designed does shape the market dynamics as an online labour market and was very compatible with my experiment’s workflow and characteristics.

4.2. Measurements

To assess the mood state of workers, this study utilised the widely recognised scale developed by Watson, Clark, and Tellegen (1988). The inventory comprises two 10-item mood scales, designed for the easy and efficient administration of the Positive and Negative Affect Schedule (PANAS).¹⁴ Respondents are asked to read 20 words that describe a series of feelings and emotions and then indicate the extent to which they usually feel them, responding on a Likert-type scale ranging from very rarely or not at all (1) to extremely often (5). Total scores on each scale (PA and NA) are obtained by adding the scores for each item (Seib-Pfeifer et al. 2017). These scales have demonstrated strong internal consistency, minimal correlation, and high validity. Consequently, they have been widely employed in numerous experiments by psychologists and economists due to their heightened intensity, prolonged duration, and more targeted response to environmental stimuli (Ifcher and Zarghamee 2011).

To measure ethical perceptions, this study builds upon Roman’s research, and it adopts ethics factors appropriate for online settings (Roman 2007; Román and Cuestas 2008). Hence, a worker’s constructed ethical perception index based on five dimensions was used, which were further reworded to a small extent to fit the research context of OLMs (Fieseler, Bucher, and Hoffmann 2019). First, privacy¹⁵ and security¹⁶ have traditionally been the two main ethical concerns in online settings. In the context of OLMs, these issues are underexplored, and other potentially important ethical issues need to be considered (Sutherland and Jarrahi 2018). Privacy and security challenges in OLMs mainly include transactional threats, workers’ available profile information and worker privacy preservation (Schlagwein, Cecez-Kecmanovic, and Hanckel 2019). Second, concerning fulfillment/reliability, it addresses the degree to which workers believe that they are able to work in a reliable online environment that offers well-grounded working conditions in terms of wage policies, fair reputation system, etc. (Horton, Rand, and Zeckhauser 2011). Next, share value measures the extent to which workers and online service providers believe the degree to which both have common values regarding which goals, behaviours or policies are right or wrong, important or unimportant (Morgan and Hunt 1994), for example, when an OLM seeks the permission of the worker for changes in the transaction process (Lin, Liu, and Viswanathan 2018). Last, service recovery addresses the course of actions an online platform service provider takes in case of a failure. For instance, it is very common that a worker’s outcome may be rejected in the end, without any explanation, resulting in various detrimental effects on crowd workers. To deal with unfair rejections, an OLM could create a system that provides requesters’ feedback (Gadiraju and Demartini 2019). A seven-point Likert scale was adopted for each item (ranging from 1 = ‘strongly disagree’ to 7 = ‘strongly agree’).

Moreover, workers’ value co-creation intention was measured on the basis of the literature on service-dominant logic value co-creation intentions (Vargo and Lusch 2008; Vargo, Maglio, and Akaka 2008), embedding the basic sharing economy aspects (Nadeem et al. 2021; 2020) within the concept of OLMs (Fedorenko and Berthon 2017).¹⁷ A seven-point Likert scale was used (ranging from 1 = ‘strongly disagree’ to 7 = ‘strongly agree’). The measurement items were further reworded to a small extent to fit the research context of OLMs.

Regarding the components of the relationship quality theory, trust is defined as the willingness to rely on an exchange partner with whom a certain level of confidence has been built (Moorman, Deshpande, and Zaltman 1993). Satisfaction refers to the worker’s evaluation of the comprehensive performance of a product/service provider such as an OLM (Gustafsson, Johnson, and Roos 2005). Commitment refers to the notion or desire to keep and maintain the relationship (Morgan and Hunt 1994). A seven-point Likert scale was used (ranging from 1 = ‘strongly disagree’ to 7 = ‘strongly agree’). The measurement items were further reworded to a small extent to fit the research context of OLMs (Liang et al. 2011).

To capture differences in personality among subjects, the Big Five personality questionnaire of a 44-item inventory is used, which provides measures for each personality trait, i.e. openness, conscientiousness, extraversion, agreeableness, neuroticism – hereafter, OCEAN (John and Srivastava 1999; McCrae and Costa 1999). The Big Five dimensions of personality were estimated on a scale of 1–5, where 1 = disagree, 2 = slightly disagree, 3 = neutral, 4 = slightly agree and 5 = agree. Afterward, the OCEAN factors were constructed through a factor analysis to ensure that each trait was orthogonal to the rest (McCrae and Costa 1999). To allow for an easier interpretation of my estimates, the Big Five scores were standardised to have a mean of zero and a standard deviation of one in all reported specifications (Cubel et al. 2016).

4.3. Experimental framework

The study used a quasi-experimental method with a mood-induced intervention proposed by Oswald, Proto, and Sgroi (2015). To avoid self-selection biases, the offered wage is in line with the price policy of Amazon Mechanical Turk and was set to $0.80 (Banfi and Villena-Roldan 2019). The experiment consisted of two rounds, with an obligatory break in the middle. In each round, my required task was a fulfillment of a questionnaire that contained both the questions of interest (i.e. ethical perceptions and co-creation intention) and several other irrelative questions with various contents. To avoid bias from the Hawthorne effect,¹⁸ my questions of interest were allocated randomly, rephrased in the second round (Adair 1984) and part of a broader online survey job. The task was programmed using zTree (Fischbacher 2007). All participant workers were randomly assigned (using the uniform distribution algorithm) to one of two different groups in each session. The first group serves as the ‘the benchmark group’ and the second as the ‘treatment group’, in which mood induction took place. During the experimental sessions, all workers were not aware of the randomisation process, which led them to a mandatory break in which the ‘benchmark group’ was exposed to a neutral placebo film while the ‘treatment group’ was exposed to a comedy film. By following Oswald, Proto, and Sgroi’s (2015) experimental design, the study used as a ‘placebo’ film a moderately interesting but not intrinsically happy clip that depicts patterns of coloured sticks that appear and disappear randomly on the screen. The film is considered ‘neutral’ by social psychologists.¹⁹ By setting the process to repeat, it was possible to play this clip for the appropriate length of time (i.e. 2 minutes). On the other hand, a similarly induced positive mood was used, with a ‘comedy’ film consisting of a 2-minute composition of well-known U.S.A. comedians.²⁰ There are many ways to regulate emotions, but considerable research attention has been given to cognitive reappraisal.²¹ The following technique involves reframing thoughts or situations to decrease their emotional impact through exogenous stimuli (McRae, Ciesielski, and Gross 2012). At the end of the obligatory break, workers received a brief reminder of the task and conditions of the experiment ahead (i.e. like Phase 1), and they proceeded to the second phase, in which they answer a number of questions, including the questions of interest (i.e. ethical perception, value co-creation and relationship quality scales), with different orders and slightly altered. Each session started with a survey on basic demographics (i.e. age, gender, marital status), social economic attributes (labour status, income), cognitive skills (i.e. educational level, crowdsourcing experience) and the Big Five personality inventory. To confirm the treatment efficiency, the workers reported their mood with the PANAS inventory at both the beginning and the end of the experimental process (Watson, Clark, and Tellegen 1988).

5. Empirical analysis

5.1. Model

By following Cubel’s econometric model on individual personality differences, the study estimates the following specification by ordinary least squares (OLS) to investigate whether the relationships between the psychological indicators and ethical perceptions and co-creation intention are heterogeneous across individual characteristics (Cubel et al. 2016): (1) $Y_{\mathrm{it}}=\alpha +T_i+\sum _{\kappa =1}^5{\beta }^{\kappa }{P}_i^{\kappa }+\gamma {\mathrm{M}}_{\mathrm{it}}+\delta X_i+u_i$(1) where Y is the dependent outcome (i.e. ith individual’s degree of ethical perception, co-creation intention and the relationship quality variables) by worker and by experimental phase t = 1,2. Τ is a dummy variable indicating the treatment group specific effect (=1), and Post is the time trend common to the control and treatment groups (i.e. a dummy variable indicating pre (t = 0) and post (t = 1) treatment). Μ is a vector containing the mood’s positive and negative affect levels of the ith worker. P refers to the personality traits, where k = 1 … , 5 are each of the noncognitive Big Five personality traits (openness to experience, conscientiousness, extraversion, agreeableness and neuroticism, or the OCEAN variables), and X are individual characteristics (i.e. demographics, social economic attributes and cognitive skills). Finally, u is the idiosyncratic error term.

5.2. Data analysis and estimation results

The online survey link was open for respondents for one day. The desired number of responses was obtained within the given timeframe, and no reminders were sent to the respondents, which means that the data were obtained from one group within a certain timeframe. Consequently, nonresponse bias, which refers to comparing early and late responses, is a nonissue in the current study. However, the problem of common method bias can occur when the data are collected from the same population at the same time and might influence the validity of the study (Podsakoff et al. 2003). To address the issue of common method bias, an algorithm was used, and worker IDs that had already participated in the experiment and worker IDs that had aberrant behaviour,²² were excluded. Additionally, by default, a psychological separation was used, when measuring my independent (predictor) and dependent (criterion) variables (Podsakoff et al. 2003).

After excluding individuals with sloppy behaviour, three hundred and twenty (N = 320) individuals participated in the study²³ (Gadiraju et al. 2015). A sample profile and characteristics are shown in Table 1. High social economic index includes individuals having a FAS index (i.e. The Family Affluence Scale)²⁴ over the Q3 quantile. This index was used to capture behavioural differences which may correlate with the socioeconomic status of the participants (Thebault-Spieker, Terveen, and Hecht 2015). Moreover, it is widely observed that married couples are more likely to have children compared to single workers, and this factor influences their time allocation choices. By incorporating marital status as a control variable, the study sought to capture the differences in time use allocation between the market and the household (Jiao, Li, and Liu 2021). Lastly, in the regression analysis we used the continuous forms of age (M = 40.07 and SD = 12.71) and experience on AMT (M = 3.75 and SD = 1.51).

Table 1. Sample characteristics.

	%
Gender
Male	44.06 Boyce
Female	55.94
Age
GenZ (Less than 19 years)	1.25
GenY (19–37 years)	49.69
GenX (37–54 years)	49.06
Whites	73.75
Married	60.63
At least college education	65.94
High social economic index	43.75
For how long have you been using the AMT (experience)?
1–2 years	0.94
3–4 years	70.94
Over 5 years	28.13

Source: Author’s calculations. Data drawn from Amazon Mechanical Turk.

Notes: N = 320.

The Stata 17 software package was employed to analyze the data. The reliability and validity of the constructs were examined first. The psychometric properties of each construct were assessed, and each measurement scale was assessed as reliable. The Cronbach’s alphas ranged higher than the 0.70 threshold suggested by Nunnally (1978). The lowest Cronbach’s alpha value in our study for the construct was 0.761; thus, there were no issues in meeting the reliability criterion. All the retained items and constructs showed good internal consistency. Table 2 includes the Cronbach’s alpha, mean and standard deviation values of each constructed dependent variable.

Table 2. Measurement items.

Scales	T1			T2
A. Ethical perceptions	CA	Mean	SD	CA	Mean	SD
1. Privacy	0.867	22.330	4.224	0.883	22.830	4.609
A1. Without the consent of workers, Amazon Mechanical Turk will not use personal information for purposes other than for the original transaction.
A2. Amazon Mechanical Turk guarantees that personal information of workers will be handled in accordance with a third party’s privacy-protection regulations and has acquired authentication knowledge.
A3. Amazon Mechanical Turk will not apply special technology to collect and analyze the internet behaviour and working habits of workers without their consent.
A4. Amazon Mechanical Turk clearly explains how information provided by workers is used.
2. Security	0.842	17.228	3.334	0.848	16.953	3.569
A5. Amazon Mechanical Turk guarantees that the transmission of transactional data will be protected without any unauthorised modification or sabotage.
A6. Amazon Mechanical Turk has a transactional security policy that workers can understand easily.
A7. Amazon Mechanical Turk guides workers to correct and safe transaction steps.
3. Fulfillment/reliability	0.828	21.230	4.223	0.800	21.398	4.863
A8. Workers receive the correct ratings as part of Amazon Mechanical Turk's reputation system.
A9. Workers browse online tasks that are matching the description of general task characteristics being adopted by Amazon Mechanical Turk.
A10. Amazon Mechanical Turk guarantees that posted tasks are authentic online jobs.
A11. The earnings on Amazon Mechanical Turk are based on and consistent with a legal minimum hourly rate.
4. Shared value	0.877	15.955	3.932	0.894	16.006	4.094
A12. Amazon Mechanical Turk respects my business values.
A13. Amazon Mechanical Turk and I have a mutual understanding of each other’s business values.
A14. Amazon Mechanical Turk sticks to the highest level of business ethics in all its transactions.
5. Service recovery	0.788	16.228	3.379	0.833	16.231	3.685
A15. Amazon Mechanical Turk responds to workers’ complaints promptly.
A16. Amazon Mechanical Turk could create a system that provides feedback on why a worker’s outcome was rejected.
A17. A failure is not neglected by Amazon Mechanical Turk, and it is promptly dealt with via a reasonable recovery strategy.
B. Value co-creation	0.802	16.868	3.669	0.817	17.106	3.697
B1. I am willing to provide my experiences and suggestions when my friends on my favourite online market want my advice on how to participate with Amazon Mechanical Turk.
B2. I am willing to participate in an online market recommended by my friends through Amazon Mechanical Turk.
B3. I consider the experiences of my friends through Amazon Mechanical Turk when I also want to work in an online task of Amazon Mechanical Turk.
C. Relationship quality
1. Trust	0.772	11.874	1.960	0.869	12.326	2.239
C1. I would characterise Amazon Mechanical Turk as honest.
C2. Amazon Mechanical Turk is trustworthy.
2. Satisfaction	0.761	11.923	1.858	0.840	12.365	2.117
C3. Overall, I am satisfied with Amazon Mechanical Turk.
C4. The last time I used Amazon Mechanical Turk, it fulfilled my expectation.
3. Commitment	0.848	17.078	3.847	0.841	17.318	3.797
C5. I feel a sense of belonging with Amazon Mechanical Turk.
C6. I am a loyal patron of Amazon Mechanical Turk.

Notes: Scales adapted from the mentioned authors in the text and altered in the context of online labor markets.

SD = Standard deviation; CA = Cronbach’s alpha; N = 320.

In the total sample, the mean PANAS-positive affect was 32.265 (SD = 9.217), and the mean PANAS-negative affect was 18.940 (SD = 10.624). Additionally, manipulation checks revealed that participants’ positive affect was increased by the intervention, by 1.350 points, only for the treated individuals (Table A1 Appendix). According to the observed variation in personality traits, we see that the mean score for openness is 3.708, for conscientiousness is 3.829, for extraversion is 3.106, for agreeableness is 3, and for neuroticism is 2.746. A representation of the distribution of the two mood indicators (positive and negative affect) is provided by Figures 1 and 2 and gives us the ability to further investigate their observed variation across the personality traits.

Figure 1. Relationship between positive affect and Big Five personality.

Figure 2. Relationship between negative affect and Big Five personality.

Concerning the treatment, to investigate the impact of positive mood administration, a paired t test was conducted between the treated and untreated individuals in the T₁ and T₂ experimental phases. Paired t tests did not reveal statistically significant differences within the values of the dependent variables before the treatment took place (T₁) between the control and treated groups. On the other hand, the dependent variables were significantly increased after positive mood administration for the experimental group in relation to the control group of individuals (T₂) (Table 3). Hence, individuals being manipulated with the positive event seem to have higher levels of ethical perceptions, higher intention for co-creation and higher quality engagement with the OLM, mainly in terms of trust and satisfaction. These results are consistent with H1, H2 and H3.

Table 3. Paired sample t tests.

Participants	Paired variables		Difference [1]–[2]	t test |t|
	No treated group	Treated group
	[1]	[2]	[3]	[4]
T1
Privacy	22.390	22.269	−0.121	1.140
Security	17.421	17.035	−0.386	0.041
Fulfillment/reliability	21.280	21.179	−0.101	0.911
Shared value	15.713	16.197	0.484	1.458*
Service recovery	15.842	16.034	0.192	1.110
Value co-creation	16.567	17.158	0.619	1.411
Trust	11.682	12.065	0.383	1.490
Satisfaction	11.756	12.090	0.334	1.003
Commitment	17.692	17.542	−0.150	0.328
T2
Privacy	22.390	23.269	0.879	1.711**
Security	16.701	17.205	0.504	1.564*
Fulfillment/reliability	20.975	21.821	0.844	1.600**
Shared value	15.589	16.340	0.651	1.623**
Service recovery	15.756	16.731	0.975	2.382***
Value co-creation	16.756	17.475	0.718	1.743**
Trust	12.012	12.639	0.627	1.625**
Satisfaction	12.097	12.633	0.536	1.579*
Commitment	17.323	17.635	0.312	0.712
Observations	164	156

Source: Author’s calculations (mean scores). Data drawn from Amazon Mechanical Turk.

Notes: Statistical significance: *** p < 0.01, ** p < 0.05, *p < 0.10.

Now, going in depth in the analysis, Tables 4 and 5 present the estimates of the effects of mood (i.e. positive and negative affects) and personality traits (i.e. openness, conscientiousness, extraversion, agreeableness, and neuroticism) on ethical perception indices (i.e. privacy, security, fulfillment/reliability, shared value, service recovery), on value co-creation and on relationship quality indicators (i.e. trust, satisfaction, commitment) before and after the treatment. Additionally, the tables include the effects of gender by focusing on the female participants.

Table 4. Determinants of dependent variables before treatment (T₁).

Dependent variables	Ethical perceptions					Co-creation	Relationship quality
	Privacy [1]	Security [2]	Reliability [3]	Shared value [4]	Service recovery [5]	Value [6]	Trust [7]	Satisfaction [8]	Commitment [9]
Female	0.823** (0.309)	0.530 (0.314)	0.338 (0.357)	0.214 (0.341)	0.123 (0.315)	0.183 (0.359)	−0.011 (0.148)	0.347* (0.185)	0.468 (0.290)
Mood
Positive affect	0.129*** (0.026)	0.092*** (0.019)	0.155*** (0.041)	0.131*** (0.037)	0.146*** (0.028)	0.168*** (0.023)	0.081*** (0.018)	0.066*** (0.015)	0.147*** (0.029)
Negative affect	0.049 (0.034)	0.051* (0.021)	0.126** (0.037)	0.106** (0.036)	0.075** (0.024)	0.032 (0.020)	−0.002 (0.010)	0.002 (0.013)	0.057 (0.028)
Personality traits
Openness	0.763 (0.335)	0.211 (0.180)	−0.247 (0.345)	−0.275 (0.346)	−0.137 (0.370)	−0.019 (0.320)	0.189 (0.165)	0.031 (0.189)	−0.140 (0.408)
Conscientiousness	0.813 (0.633)	0.591 (0.497)	0.454 (0.390)	0.502 (0.507)	0.106 (0.308)	– 0.664* (0.300)	0.336* (0.132)	0.042 (0.178)	0.624 (0.488)
Extraversion	−0.143 (0.355)	0.184 (0.390)	0.363 (0.383)	0.328 (0.346)	−0.210 (0.232)	0.512*** (0.232)	0.098 (0.128)	0.129 (0.142)	0.478 (0.309)
Agreeableness	1.085*** (0.459)	1.114*** (0.390)	0.698* (0.374)	0.994** (0.398)	1.023*** (0.309)	0.721** (0.332)	0.398** (0.204)	0.491** (0.197)	0.621* (0.386)
Neuroticism	0.519 (0.459)	0.322 (0.320)	0.178 (0.354)	0.015 (0.405)	0.389 (0.294)	– 0.334** (0.284)	0.371** (0.167)	0.268* (0.170)	0.306 (0.308)
Constant	5.046 (2.576)	9.509 (3.581)	3.947 (3.262)	5.222 (2.758)	6.989 (2.944)	9.221 (2.926)	3.933 (1.499)	6.475** (1.542)	3.818 (2.889)
R-squared	0.244	0.305	0.305	0.298	0.269	0.336	0.249	0.185	0.284
F-Stat	7.33	28.09	25.66	27.53	12.12	17.04	11.13	6.09	23.32
Demographic controls	Yes
Social economic controls	Yes
Cognitive skills controls	Yes

Source: Author’s calculations. Data drawn from Amazon Mechanical Turk.

Notes: The number of participants is N = 320. The specifications control for personal characteristics, cognitive skills, socioeconomic background and individuals’ origins, i.e. regions of U.S.A. fixed effects. Robust standard errors in parentheses. Statistical significance: *** p < 0.01, ** p < 0.05, *p < 0.10.

Table 5. Determinants of dependent variables after treatment (T₂).

Dependent variables	Ethical perceptions					Co-creation	Relationship quality
	Privacy [1]	Security [2]	Reliability [3]	Shared value [4]	Service recovery [5]	Value [6]	Trust [7]	Satisfaction [8]	Commitment [9]
Female	1.039*** (0.378)	0.640* (0.352)	0.182 (0.467)	0.293 (0.305)	0.313 (0.329)	0.645** (0.333)	0.299* (0.163)	0.369** (0.197)	0.643 (0.420)
Mood manipulation
Treated	1.071*** (0.371)	0.612* (0.389)	1.124** (0.483)	0.811** (0.403)	1.124*** (0.378)	0.763** (0.377)	0.473** (0.230)	0.375* (0.247)	0.589* (0.340)
Positive affect	0.169*** (0.028)	0.114*** (0.021)	0.169*** (0.036)	0.132*** (0.029)	0.169*** (0.031)	0.150*** (0.028)	0.081*** (0.013)	0.056*** (0.014)	0.122*** (0.029)
Negative affect	0.067* (0.032)	0.075** (0.026)	0.142** (0.047)	0.112** (0.036)	0.053 (0.031)	0.038 (0.026)	0.010 (0.013)	0.011 (0.012)	0.057 (0.033)
Personality traits
Openness	0.819 (0.385)	0.251 (0.263)	−0.209 (0.420)	0.085 (0.416)	0.044 (0.381)	0.146 (0.314)	0.251 (0.224)	0.212 (0.201)	0.244 (0.380)
Conscientiousness	0.816 (0.555)	0.433 (0.392)	0.497 (0.509)	0.750 (0.442)	−0.132 (0.351)	−0.185 (0.472)	0.257 (0.213)	0.234 (0.208)	0.628 (0.524)
Extraversion	−0.281 (0.365)	−0.024 (0.330)	0.314 (0.396)	0.361 (0.332)	−0.062 (0.245)	0.721*** (0.227)	0.116 (0.187)	0.099 (0.162)	0.594* (0.333)
Agreeableness	1.264*** (0.377)	1.311*** (0.311)	1.074*** (0.391)	1.042*** (0.376)	1.139*** (0.344)	0.855*** (0.280)	0.703*** (0.223)	0.748*** (0.223)	0.707** (0.360)
Neuroticism	0.805* (0.429)	0.568* (0.329)	0.508 (0.366)	0.226 (0.335)	0.535* (0.319)	0.019 (0.358)	0.595*** (0.173)	0.500*** (0.174)	0.501* (0.323)
Constant	2.610 (3.657)	2.531 (3.118)	3.681 (3.490)	−0.191 (2.521)	3.381 (2.768)	4.944 (3.719)	1.035 (1.642)	2.816* (1.446)	−0.003 (2.963)
R-squared	0.273	0.256	0.282	0.295	0.298	0.332	0.264	0.193	0.277
F-Stat	6.98	19.18	29.67	18.20	27.53	16.85	14.70	9.56	26.10
Demographic controls	Yes
Social economic controls	Yes
Cognitive skills controls	Yes

Source: Author’s calculations. Data drawn from Amazon Mechanical Turk.

Notes: The number of participants is N = 320. The specifications control for personal characteristics, cognitive skills, socioeconomic background and individuals’ origins, i.e. regions of U.S.A. fixed effects. Robust standard errors in parentheses. Statistical significance: *** p < 0.01, ** p < 0.05, *p < 0.10.

Concerning T_1, as in the previous personality literature and in line with H4 and H5a, the results show that more agreeable individuals have significantly higher levels of ethical perceptions, value co-creation and relationship quality with OLMs. The effects are higher and statistically significant at 1%, mainly in the cases of security, privacy and service recovery. Additionally, extraverted individuals have significantly higher levels of value co-creation. More concretely, an increase of a standard deviation in the level of extraversion is associated with an increase in value co-creation of approximately 0.721. We fail to confirm the positive effects of extraversion on ethical perceptions. Regarding neuroticism, the results show that neurotics are negatively correlated with low value co-creation (Column 6) and positively correlated with the indicators of trust and satisfaction (i.e. relationship quality). No statistically significant effects on ethical perceptions were found. As expected, the positive affect level of individuals has a positive and statistically significant effect at the 1% level on all the dependent variables under investigation. Last, the study also reveals that, on average, females have higher satisfaction and privacy than males. Taking advantage of the setup of the experiment, the study also explores whether the relationship between positive mood, personality traits and performance changes with treatment (i.e. T₂). Table 5 suggests that the randomised induction of positive mood increased all aspects of ethical perception, value co-creation and relationship quality indicators for the treated individuals in relation to the control group. Initially, as expected, the effects of positive affect remained almost stable after the mood manipulation. Recall that positive affect was measured only at the beginning of the experiment to have some proxy indicators for the levels of positive and negative emotions that the individuals felt in general. Interestingly, the results revealed that mainly the effect of positive mood treatment on the individuals’ outcomes operates through the trait of agreeableness, confirming the H5b hypothesis. More specifically, in all the cases of ethics, value co-creation and relationship quality dependent variables, the treatment further boosted the magnitude and robustness of the effects of an individual’s agreeableness level.

Moreover, the treatment further boosted the effect of extraversion in the case of individuals’ intention for value co-creation, while several weak positive effects of neuroticism appeared in the cases of commitment, privacy, security and service recovery. The negative relationship between neuroticism and value co-creation disappears in T₂. Last, females seem to be affected the most by positive mood manipulation, resulting in higher intention levels of value co-creation, trust, satisfaction, privacy and security.

6. Discussion and implications

6.1. Discussion and limitations

The goal of this paper was to create a framework to investigate workers’ determinants of mood and personality that affect their underlying mechanism of ethical decision choice and co-creation value. The study managed to answer the following research questions: (1) What are the personality dimensions of workers’ ethical perceptions, co-creation value and relationship quality of OLMs? and (2) What is the role of mood on ethical perceptions, value co-creation intentions and relationship quality on OLMs? A research model was developed by combining several aspects of theoretical insights from the literature on marketing ethics (Agag 2019; Vargo and Lusch 2008; Vargo, Maglio, and Akaka 2008) and mood theory. Thus, the study has derived the following main insights from the empirical analysis.

The mood of individuals working on online platforms is a crucial factor that determines their level of ethical perceptions, value co-creation and the quality of their relationship with the platform. More specifically, the positive side of individuals’ mood further boosts all the above mentioned levels, operating mainly through the personality trait of agreeableness.

Females seem to be affected the most by the positive mood manipulation, following the increased trend of the outcomes. Although that trend was not part of the main hypotheses of the paper, the research includes these findings to trigger future experimental research oriented mainly on the demographic characteristics that may result in higher levels of ethical perceptions, co-value intention and relationship quality levels with OLMs when exogenous mood treatment is taking place (McCabe, Ingram, and Dato-On 2006).

Despite the contributions of this study, some limitations need to be acknowledged. First, while experiments in OLMs and especially AMT have high internal and external validity, we cannot extend our insights to the real labour market. Although, the study followed typical recruitment methods and platform preferences, our data were from US citizens with a particular cognitive profile (Paolacci, Chandler, and Ipeirotis 2010). This means that the study offers quite a narrow perspective on the global phenomenon of OLMs in relation to the real conventional market, as it lacks, for instance, international, cross-cultural, and global viewpoints. All of these limit the global generalisation of the results. To improve the generalisability and validity of these findings, future research could follow more diverse sampling strategies by conducting experiments in several OLMs and having a broader representation of workers from various backgrounds, demographics, and geographic regions. Conducting similar experiments on multiple crowdsourcing platforms also can provide insights into the stability and consistency of the observed effects. Moreover, future studies could employ randomised controlled trials (RCTs) to help minimise biases and improve the internal validity of these studies.

A second limitation pertains to the fact that, while the study focused on exploring the impact of a positive mood manipulation strategy in an online working environment, such environments are inherently more complex and encompass a dynamic interplay of both positive and negative effects. Hence, a further examination of the role of mood could include replicating the current study in the near future and focusing on examining whether workers’ perceptions differ due to negative affect manipulation. Last, although it is already known that the Big Five personality variables are stable for adolescents and working-age adults and can be used to explain economic behaviour, they do not provide compelling causal explanations for human behaviour. Hence, the Big Five may be viewed as an important model in personality studies but not the integrative model of personality.

6.2. Theoretical implications

While studies examining OLMs are becoming more prevalent, there are still gaps in the research that this study helps to address. Prior papers on this online and innovative economy have tended to focus on the determinants of workers’ productivity level and their behavioural response under various exogenous stimuli (Arechar, Gächter, and Molleman 2018; Mason and Suri 2012). However, workers’ psychological profiles should not be underestimated. While crowd workers as a target group and relevant population are less investigated, in general psychology crowdsourced data – as participant recruitment pools – are well established and common in those domains (Paolacci, Chandler, and Ipeirotis 2010; 2014). Therefore, the research should investigate OLMs as a new form of work and its psychological impact on the people performing it.

Hence, the article contributes to the literature on OLMs in several ways. First, the models of Hajli (2014) and Nadeem et al. (2021; 2020), were empirically confirmed, which already gave a profound understanding of the ethical perceptions in online sharing economy environments, and we expand these findings on crowdsourcing OLMs. Second, the paper is the first to create an empirically validated framework that helps explain the effects of workers’ personality traits on their judgments regarding their ethical perceptions and their intention to cocreate value. Although there are many well-established psychological models and theories that could be applied to OLMs to explain underlying mechanisms that are not sufficiently answered today, we still have to demonstrate whether these models are suitable for this new form of work (Brawley 2017). This attempt was based on the trait theory of personality, which is a generally admitted approach to the study of human personality and has already been applied to offline and online labour market studies (Haylock and Kampkötter 2019; Mourelatos, Giannakopoulos, and Tzagarakis 2022). Hence, this research provides insights into the personality drivers of perceptions on the ethical framework of OLMs and workers’ co-creation intention by also taking into consideration terms of the relationship quality theory (i.e. trust, commitment and satisfaction).

Finally, the present article also addresses the moderating influence of mood states on constructive biases in the context of ethical judgments by delineating and empirically testing that people in good mood states are more prone to constructive processes than people in bad or depressed moods. Specifically, from a theoretical standpoint, our findings suggest that a worker’s ethical sensitivity is being expressed through higher conciliatory ethical judgments as his positive mood increases, resulting in a higher co-creation intention (Abele and Gendolla 1999). This is an important distinction revealing that while ethical propensities are generally considered to be stable over time, this research suggests that ethical decisions and the factors that influence those decisions can vary across mood states (Curtis 2006). Unlocking the power of mood will help us to gain a better understanding of a worker’s mechanism on ethical perceptions related to OLMs, and throughout our experimental design with mood-inducing events, our knowledge can be advanced. For that reason, several aspects of mood deserve further inquiry.

6.3. Practical implications

Considering the growing significance of OLMs, with an estimated 36% of U.S. workers presently engaged in them²⁵, coupled with the ongoing economic challenges precipitated by the pandemic, it becomes evident that this innovative labor paradigm is experiencing a notable upward trajectory. Although OLMs offer excellent earning opportunities, with a decreasing need for physical presence in the workplace, making it easier than ever to work for multiple employers simultaneously, being a worker still carries risks. Thus, as this market matures, several concurrent forces jointly reduce the effect of moral hazard in tandem (Pavlou and Gefen 2005). For that reason, the present study may operate as an initial step in OLM ethics management. It becomes paramount and crucial for both OLM providers and requesters to understand workers’ ethical perceptions, how they are defined and how they influence their value co-creation intentions in online jobs. A job pursuit intention, with ethical climate engagement, will have a high possibility for high-quality working outcomes, which is ultimately the goal of OLMs and crowdsourcing procedures (Wong et al. 2021).

Findings reveal that workers’ personality traits and current mood play a key role in explaining the formation of their ethical perceptions of the platform, their co-creation intentions on online jobs and their relationship quality with the OLM. In other words, workers’ personalities and moods are ubiquitous throughout the managerial marketing strategies that OLMs must design and identify (Zahay, Hajli, and Sihi 2018). These include primarily the improvement of the already existing ethical aspects, such as the reliability of the online working environment, as well as the workers’ perception that the OLM providers’ values are in line with their own values. For example, researchers identify the biggest issues of OLMs in the communication between workers and the platform, the platform architecture and legal questions, such as whether to impose a tax on online tasks.

Results also help OLM providers better understand the role and importance of the multidimensionality of workers’ ethical perceptions in relation to their psychological profiles. This, eventually, further will enable OLMs to have as a priority a more personalised-oriented and friendly framework of their crowdsourcing flow in terms of aesthetics, colours, navigation consistency, textual content and alignment (Jiang et al. 2016; Moshagen and Thielsch 2013). A website with a good atmosphere provides workers with a feeling of control, which enhances their enjoyment and extends their intentions to engage in co-creation on the platform.

6.4. Conclusions

Traditionally, the research on ethical perceptions has been quite distinct from the study of mood. Psychology has tacitly treated perception and mood as separable phenomena to be studied in isolation. However, the revolutionary psychological field of mood science revealed that relevant areas of the brain and the processes they support are highly interactive. Thus, by keeping in mind that mood influences information processing, mediates responses to persuasive appeals, measures the effects of stimuli, initiates goal setting, enacts goal-directed behaviours, serves as ends and measures workers’ working preferences and sense of working environment, This study tries to illustrate evidence that mood is clearly connected with ethical perceptions and co-creation intention in relation to workers’ individual personality characteristics.

The experimental findings have developed several arguments on how mood interacts with workers’ online behaviour (Loewenstein 2000; 2001). However, we are only beginning to understand the role of psychology in OLMs. It is widely recognised among researchers that ethical aspects differ in offline and online environments, workers’ ethical evaluations are formed in different ways on online platforms and in offline settings (Roman 2007), and in general, the internet is often seen as an environment for unethical behaviour (Freestone and Mitchell 2004; Hajli 2018). Hence, this trend has implications for future research within the context of online working environments and trying to investigate and develop the ability to foster a co-creation environment and build communities of enthusiastic contributors, which will play a greater role in the future of organisational employment relationships.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 OLMs and SEPs encompass a wide range of activities and business models and share a large number of common characteristics. For that reason, in some instances a virtual platform can be categorized as both a sharing and online-crowdsourcing labor platform (Taeihagh 2017).

2 The use of these OLMs has increased by approximately 20%, especially over the last five years, and the estimated total market size is approximately $25 billion, with over 48 million registered online workers (Kässi and Lehdonvirta 2018).

3 The empirical investigation of individual performance relies on data drawn from several sources, i.e. surveys, administrative datasets and lab experiments.

4 Facebook in terms of data protection and privacy by harvesting data from millions of users in 2013 – without their explicit consent, Best Buy with data breaches, Uber with alleged cheating of its drivers by rounding fees to the nearest dollar in favor of the company, Equifax with a major security breach that it said affected approximately 145 million of its US consumers, etc.

5 In October 2012, an online worker named Christopher Otey filed a lawsuit against CrowdFlower, claiming that companies are failing to pay the federal minimum wage ($7.25 an hour at the time of the lawsuit) under the Fair Labor Standards Act (Schmidt 2013). CrowdFlower’s defense said workers are working voluntarily and are free contractors instead of employees. CrowdFlower settled in court on September 2, 2014, paying a gross settlement of $585,507.00 (2014).

6 Most online jobs involve little or no personnel administration costs because an employer does not need to hire managers to supervise the crowd and can avoid turnover and recruitment expenses.

7 Fehr, Goette, and Zehnder (2009) review these and other empirical findings and make the case that fairness concerns play an important role in labor market outcomes.

8 Many emotion theorists seem to find no special purpose for the term mood, using it interchangeably with other labels, such as affect or emotion (Bower 1981; Solomon 1980; Watson and Tellegen 1985).

9 Such as deceptive advertising, dangerous products, and misleading prices.

10 Web 1.0 refers to the first stage of the World Wide Web evolution, which mainly included static pages without a client–server interaction.

11 Web 2.0 refers to the updated World Wide Web, which highlights user-generated content, usability and interoperability for end users. Web 2.0 is also called the participative social web, because it includes a number of online tools and platforms where people can share their perspectives, opinions, etc.

12 This includes crowdfunding and crowdsourcing online labor platforms.

13 In other words, value cocreation refers to a collaborative effort in which different actors jointly and reciprocally participate in creating value.

14 Positive affect refers to the propensity to experience positive emotions and interact with others positively, even though the challenges of life. Negative affect, on the other hand, involves experiencing the world in a more negative way.

15 Privacy deals with uncertainty linked to personal information that is provided on online platforms, and the risk of such information being exposed to unintended individuals or parties (Bart et al. 2005).

16 Security pertains to the notion of uncertainty regarding online platforms that could lead to incurring monetary losses during interaction on those platforms (Roman 2007).

17 In essence, OLMs, while they offer the opportunity for interactions between requesters and workers, they do not empower many tools for appropriate communication between workers and themselves, undermining a trustworthy behavior that may lead to cocreation (Horton 2010; Perren and Kozinets 2018).

18 The Hawthorne effect refers to a type of reactivity in which individuals modify an aspect of their behavior in response to their awareness of being observed.

19 The film clip was ‘Computer Graphic’ on James Gross’s resources.

20 The research team conducted a two-step research in order to decide which comedians and which videos to embed into the ‘comedy’ film. First, we searched in Google, Quora and Reddit for the best USA comedians, and then we took into consideration their metrics in YouTube, Facebook and Instagram.

21 For example, Alpert and Alpert (1990) used background music in order to stimulate consumer response.

22 For example, a concentration of answers, or a fast pace of answering.

23 Ten participants were excluded from the experiment because they answered questions quickly and with the same pattern (response bias).

24 FAS index consists of 6 items measuring social economic status of individuals. The responses to the items are given as specific values and calculated as an aggregated FAS index ranging from 0 to 13. (Boyce et al. 2006)

25 Only 28% of independent workers belong to Generation X (people aged 40 to 55).

Appendices

Appendix A. The experimental flowchart:

Appendix B

Table A1. Manipulation checks (t tests).

	Start (T1)	End (T2)	Difference	|t|
	[1]	[2]	[3]	[4]
Pooled
Positive affect	32.265	33.615	1.350**	1.972
Control group (placebo film)
Positive affect	32.141	31.044	−1.097	1.058
Treated group (comedy film)
Positive affect	32.384	36.060	3.676***	3.796

Source: Dataset with results drawn from the experiment. Author’s calculations.

Notes: Difference is [2]–[1].

Statistical significance: *** 1%, ** 5% and * 10%.