Periodic Reporting for period 2 - GHOSTWORK (The ghostworker's well-being: An integrative framework)
Reporting period: 2023-03-01 to 2024-08-31
What is the issue being addressed?
Today, AI is increasingly demanding low-tech human support in coding, verification and cleaning of software material, such as training data. The so-called ghost workers perform such short-term tasks on demand, anonymously, through automated platforms, without any feedback from colleagues or managers. Despite the growth of this particular phenomenon, the influence of such an occupation on work-related well-being remains unresearched. To address this urgent gap, the EU-funded GHOSTWORK project conducts an in-depth study of ghost workers’ working conditions and how these impact their overall well-being. A new multi-methodological approach is developed, enabling integrated models, theory and empirical findings.
Why is it important for society?
AI applications are being trained and verified by using large amounts of data. To make the data suitable for using it this way, it needs to be annotated. This work is conducted by millions of (ghost) workers. It is important to acknowledge this human labor as part of the AI-value chain. Both for the quality of the AI-applications as for the working conditions of these workers. For Ethical and Trustworthy AI, the work and working conditions of (ghost)workers need to be taken into account.
What are the overall objectives?
I. to develop an integrative framework for analysing ghostworkers’ well-being;
II. to investigate short-term dynamics of ghost work for workers’ work conditions and well-being;
III. to test the effects of the advantages and challenges associated with ghost work on workers’ well- being over a long-term period for different demographic groups.
Today, AI is increasingly demanding low-tech human support in coding, verification and cleaning of software material, such as training data. The so-called ghost workers perform such short-term tasks on demand, anonymously, through automated platforms, without any feedback from colleagues or managers. Despite the growth of this particular phenomenon, the influence of such an occupation on work-related well-being remains unresearched. To address this urgent gap, the EU-funded GHOSTWORK project conducts an in-depth study of ghost workers’ working conditions and how these impact their overall well-being. A new multi-methodological approach is developed, enabling integrated models, theory and empirical findings.
Why is it important for society?
AI applications are being trained and verified by using large amounts of data. To make the data suitable for using it this way, it needs to be annotated. This work is conducted by millions of (ghost) workers. It is important to acknowledge this human labor as part of the AI-value chain. Both for the quality of the AI-applications as for the working conditions of these workers. For Ethical and Trustworthy AI, the work and working conditions of (ghost)workers need to be taken into account.
What are the overall objectives?
I. to develop an integrative framework for analysing ghostworkers’ well-being;
II. to investigate short-term dynamics of ghost work for workers’ work conditions and well-being;
III. to test the effects of the advantages and challenges associated with ghost work on workers’ well- being over a long-term period for different demographic groups.
The GHOSTWORK-project started with a survey focused on where in the EU microwork is performed and by whom (worker demographics). While the resulting distribution of workers across the EU highlighted the need for further investigation into the influence of geographical contexts on microwork the patterned diversity of workers that emerged from the worker classification was the more unexpected result. Previous research has highlighted the importance of worker heterogeneity within microwork, our classification specifically supported the value of contextual understanding of worker diversity and dependency. Together, these findings can enhance the ability of institutions to deliver stronger policy agendas prioritizing and directing efforts towards high density microworking areas, as well as those workers more dependent upon such work for their basic needs (https://doi.org/10.1177/09596801231171997).
The follow-up 137 interviews focused on investigating the relationship between microworking working conditions and workers’ well-being, taking an inductive analytical approach. Currently, our efforts are focused on drafting the first interview-based article elaborating on the emergent process of ‘negotiation’ workers demonstrated in navigating the precarity of platform work and the precariousness of the microworking experience. The findings suggest the importance of seeing workers as active agents navigating platform work conditions. We have disseminated this knowledge at two conferences: British Sociological Association Work, Employment & Society Annual Conference 'Rethinking Resistance in the Workplace' (Glasgow, Sept 13-15 2023); International Network on Digital Labor 6 'Digital Labor in the Wake of Pandemic Times' (Berlin, Oct 9-11 2023).
In the second Subproject, a comprehensive literature review was performed to conceptualize the working conditions in microwork. This review identified six novel working conditions very specific to microwork that were corroborated with the interview data from first Subproject. Nine working conditions from established models appeared prominently in the literature and were adapted for the measurement instrument to assess microworkers' work conditions. This resulted in a total of 15 definitions of novel and established working conditions.The initial pool of items and definitions were subjected to content validation procedures with academic experts and workers, rating the items regarding their relevance, representativeness, and clarity. Based on the content validation feedback, concept definitions were refined, and the initial item pool was reduced. To validate the newly developed measurement scale, a survey containing the new items was created and distributed among workers on the platforms MTurk, Microworker, Clickworker, and Picoworker (N=407). A 15- factor model showed sufficient fit. Item reduction was performed in an iterative manner. Based on these analyses the scale was shortened to comprise 61 items, four to six items per dimension. Model fit shows excellent fit of the data in a 15- factor model, establishing a psychometrically sound measurement scale. We have disseminated this knowledge at two conferences: The SISEC conference in Italy (2024) and the Junior Research Day of the Department of Public Administration and Sociology (Erasmus University Rotterdam).
Subsequently, in Subproject 2, we address the short-term dynamics of microwork for workers' work conditions and well-being by investigating intraindividual processes and the importance of interindividual differences. We measured stable individual differences in motivation for platform work, daily indicators of well-being, daily activities (platform, job, household, social, physical, and low-effort activities), and playful work design (i.e. proactive play during activities). 122 individuals filled out 457 days and 3582 activities. We found that the intrinsic and extrinsic motivation to enter microwork has important implications for well-being in terms of exhaustion, work engagement, and detachment, as well as how the daily hours expended impact their well-being. In addition, we found that designing fun and designing competition primarily attenuate how extrinsically motivated activities are experienced. In contrast, playful work design mainly had main effects during intrinsically motivated activities. Taken together, the findings elucidate how inter- and intraindividual differences determine well-being of microworkers. This study enables us to discern the unique effects of platform activities and time spent on these activities on well-being while correcting for the effects of other activities and time spent on them. The study represents a significant methodological development in the field of platform work. We have disseminated this knowledge at two conferences: The SISEC conference in Italy (2024) and the Work Design Conference in Australia (2024).
The follow-up 137 interviews focused on investigating the relationship between microworking working conditions and workers’ well-being, taking an inductive analytical approach. Currently, our efforts are focused on drafting the first interview-based article elaborating on the emergent process of ‘negotiation’ workers demonstrated in navigating the precarity of platform work and the precariousness of the microworking experience. The findings suggest the importance of seeing workers as active agents navigating platform work conditions. We have disseminated this knowledge at two conferences: British Sociological Association Work, Employment & Society Annual Conference 'Rethinking Resistance in the Workplace' (Glasgow, Sept 13-15 2023); International Network on Digital Labor 6 'Digital Labor in the Wake of Pandemic Times' (Berlin, Oct 9-11 2023).
In the second Subproject, a comprehensive literature review was performed to conceptualize the working conditions in microwork. This review identified six novel working conditions very specific to microwork that were corroborated with the interview data from first Subproject. Nine working conditions from established models appeared prominently in the literature and were adapted for the measurement instrument to assess microworkers' work conditions. This resulted in a total of 15 definitions of novel and established working conditions.The initial pool of items and definitions were subjected to content validation procedures with academic experts and workers, rating the items regarding their relevance, representativeness, and clarity. Based on the content validation feedback, concept definitions were refined, and the initial item pool was reduced. To validate the newly developed measurement scale, a survey containing the new items was created and distributed among workers on the platforms MTurk, Microworker, Clickworker, and Picoworker (N=407). A 15- factor model showed sufficient fit. Item reduction was performed in an iterative manner. Based on these analyses the scale was shortened to comprise 61 items, four to six items per dimension. Model fit shows excellent fit of the data in a 15- factor model, establishing a psychometrically sound measurement scale. We have disseminated this knowledge at two conferences: The SISEC conference in Italy (2024) and the Junior Research Day of the Department of Public Administration and Sociology (Erasmus University Rotterdam).
Subsequently, in Subproject 2, we address the short-term dynamics of microwork for workers' work conditions and well-being by investigating intraindividual processes and the importance of interindividual differences. We measured stable individual differences in motivation for platform work, daily indicators of well-being, daily activities (platform, job, household, social, physical, and low-effort activities), and playful work design (i.e. proactive play during activities). 122 individuals filled out 457 days and 3582 activities. We found that the intrinsic and extrinsic motivation to enter microwork has important implications for well-being in terms of exhaustion, work engagement, and detachment, as well as how the daily hours expended impact their well-being. In addition, we found that designing fun and designing competition primarily attenuate how extrinsically motivated activities are experienced. In contrast, playful work design mainly had main effects during intrinsically motivated activities. Taken together, the findings elucidate how inter- and intraindividual differences determine well-being of microworkers. This study enables us to discern the unique effects of platform activities and time spent on these activities on well-being while correcting for the effects of other activities and time spent on them. The study represents a significant methodological development in the field of platform work. We have disseminated this knowledge at two conferences: The SISEC conference in Italy (2024) and the Work Design Conference in Australia (2024).
The survey in Subproject 1 on EU microworkers focused on where microwork is performed and by whom. Using Latent Class Analysis on worker demographics, we distinguished four classes of workers. We label these classes Explorers, Enthusiasts, Supplementers, and Dependents. The identification of these emergent classes and varied prevalence of microworking across the EU, suggest the importance of heterogeneity to both the future study and regulation of microwork. The survey also included an invitation for an in-person or online interview. In the end, 137 EU workers were interviewed, providing, among others, important insights for Subproject 2 about microworkers' work conditions.
In Subproject 2 we developed a comprehensive measurement instrument to assess microworkers' work conditions. We integrated both established measures and theories as well as recently discovered working conditions in microwork (during the interviews of the first Subproject). This is the first measurement instrument to assess microworkers' work conditions, which is helpful for the subsequent studies in this project and for future research on microwork in general. Valuable established insights into working conditions in microwork and platform work rely on existing theoretical frameworks such as the job characteristics model (Deng et al. 2016; Orhan et al. 2022) or Marx’s theory of alienation (Bucher et al. 2021), or the decent work criteria (Wood et al. 2019). Other studies focus on describing working conditions that are unique to this working context (Deng et al. 2016; Howson et al. 2023; Irani 2015; McInnis et al. 2016; Strunk et al. 2022; Wood 2021). To our knowledge there are no studies on microwork combining both existing theories of work with the working conditions specific to microwork. Providing a validated measurement scale, will facilitate more comprehensive and nuanced research unravelling the effects of microwork on well-being and other related work outcomes.
Previous literature addresses the importance of utilizing methodologies that adequately examine microdynamics implied in microwork. GHOSTWORK is the first to achieve this goal to this degree. By focusing on the activity level and correcting for lagged effects, we truly focus on “micro” intraindividual changes in well-being throughout the day. This stands in stark contrast with approaches that take a general approach. Second, previous literature has suggested that microwork may benefit some individuals. We investigated this by hypothesizing that extrinsic and intrinsic motivation amplify or change the association between time spent on microwork and energy levels, respectively. While our research is not conclusive, it contributes to this discourse by showing that microwork is an activity that drains individuals’ energy overall, which can be buffered by intrinsic motivation. Nonetheless, we found no evidence that intrinsically motivated microworkers are happier when they spend more time on microwork. This finding goes against the “positive” narrative that digital labor platforms may provide meaning for some.
We have also prepared and received ethical approval for Subproject 3, investigating the effects of the advantages and challenges associated with microwork on workers’ well- being over a long-term period. We will collect data from various online platforms throughout Europe and assess the implications of novel microworking conditions, using the measurement scale developed in Subproject 2. To our assumptions, we conduct a four-wave longitudinal study with a time lag of 2 months. We analyze the results using the recently introduced Bayesian Multilevel Latent CLPM (cross-lagged panel model).
In Subproject 2 we developed a comprehensive measurement instrument to assess microworkers' work conditions. We integrated both established measures and theories as well as recently discovered working conditions in microwork (during the interviews of the first Subproject). This is the first measurement instrument to assess microworkers' work conditions, which is helpful for the subsequent studies in this project and for future research on microwork in general. Valuable established insights into working conditions in microwork and platform work rely on existing theoretical frameworks such as the job characteristics model (Deng et al. 2016; Orhan et al. 2022) or Marx’s theory of alienation (Bucher et al. 2021), or the decent work criteria (Wood et al. 2019). Other studies focus on describing working conditions that are unique to this working context (Deng et al. 2016; Howson et al. 2023; Irani 2015; McInnis et al. 2016; Strunk et al. 2022; Wood 2021). To our knowledge there are no studies on microwork combining both existing theories of work with the working conditions specific to microwork. Providing a validated measurement scale, will facilitate more comprehensive and nuanced research unravelling the effects of microwork on well-being and other related work outcomes.
Previous literature addresses the importance of utilizing methodologies that adequately examine microdynamics implied in microwork. GHOSTWORK is the first to achieve this goal to this degree. By focusing on the activity level and correcting for lagged effects, we truly focus on “micro” intraindividual changes in well-being throughout the day. This stands in stark contrast with approaches that take a general approach. Second, previous literature has suggested that microwork may benefit some individuals. We investigated this by hypothesizing that extrinsic and intrinsic motivation amplify or change the association between time spent on microwork and energy levels, respectively. While our research is not conclusive, it contributes to this discourse by showing that microwork is an activity that drains individuals’ energy overall, which can be buffered by intrinsic motivation. Nonetheless, we found no evidence that intrinsically motivated microworkers are happier when they spend more time on microwork. This finding goes against the “positive” narrative that digital labor platforms may provide meaning for some.
We have also prepared and received ethical approval for Subproject 3, investigating the effects of the advantages and challenges associated with microwork on workers’ well- being over a long-term period. We will collect data from various online platforms throughout Europe and assess the implications of novel microworking conditions, using the measurement scale developed in Subproject 2. To our assumptions, we conduct a four-wave longitudinal study with a time lag of 2 months. We analyze the results using the recently introduced Bayesian Multilevel Latent CLPM (cross-lagged panel model).