5.1 Introduction
Crowdsourcing involves a diverse group of individuals coming together to address a specific task or problem within an interactive online environment (Estellés-Arolas & González-Ladrón-de-Guevara, Reference Estellés-Arolas and González-Ladrón-de-Guevara2012). This process is enabled by a digital platform that allows participants or contributors (also known as crowd workers) to collaborate, exchange ideas, and share knowledge, drawing on their combined knowledge and expertise. The collective effort not only helps in solving the task at hand but also harnesses the power of collective intelligence, where the group’s shared insights and skills lead to more effective and innovative solutions.
5.1.1 Collective Intelligence from Crowdsourcing
Collective intelligence refers to the shared knowledge or group intelligence that emerges from the collaboration, collective efforts, and contributions of many crowd workers (Lévy, Reference Lévy1999). The phenomenon of creating collective intelligence to solve problems is not new and has been observed on social media sites. For example, on Wikipedia, crowd workers contribute by creating, editing, and updating content on the platform. These volunteers, who come from various backgrounds and locations, share their knowledge and expertise on various topics, ensuring that information is accurate, up-to-date, and accessible to everyone (Lindgren, Reference Lindgren2014). Through the collaborative process, they add new articles, enhance existing ones, and verify facts, all while adhering to Wikipedia’s guidelines and standards. This collective effort helps maintain Wikipedia as one of the most comprehensive and reliable sources of information on the internet. The crowd workers’ knowledge sharing is crucial to the platform’s success, as they continuously improve the quality, depth, and breadth of content available to users worldwide.
The collective intelligence from shared knowledge can lead to significant and influential outcomes, including solving major societal problems. Taking Wikipedia as an example again, it has become more than just an information repository; it has evolved into a dynamic platform where complex ideas are discussed, refined, and made accessible to a global audience. The collective intelligence is the aggregated contributions of these crowd workers that can lead to a deeper understanding of critical social issues and the formation of new insights that can inform policymaking, including but not limited to healthcare and education (Yang & Tanaka, Reference Yang and Tanaka2023).
Indeed, diversity among crowd workers enables crowdsourcing to bring together varied perspectives that can be harnessed to address societal challenges (Brabham, Reference Brabham2012). An example of addressing societal knowledge inequality through crowdsourcing is seen in the efforts of volunteers with diverse language skills who dedicate their time and expertise to translating online course materials on learning platforms (Lee et al., Reference Lee, Lim and Kim2024b). Here, the collective intelligence is the shared expertise of the crowd workers, which allows learners worldwide, particularly those whose native language is not English, to access quality educational resources. Another example is Ushahidi (www.ushahidi.com), where the crowdsourcing platform is used for social activism and public accountability to ensure that marginalized populations’ voices are heard and their needs are taken care of (Gutierrez, Reference Gutierrez2019). In the context of Ushahidi, collective intelligence refers to the shared knowledge and problem-solving capabilities that emerge from the collaboration and aggregation of inputs from a diverse group of individuals using the platform. Additionally, widely known crowdsourcing platforms such as Foldit (http://fold.it) and Eyewire (http://eyewire.org) utilize games to engage altruistic participants in solving complex medical and biological problems, from protein folding to neuron mapping. The data generated through these games contribute directly to collective intelligence, showcasing the power of crowdsourcing in scientific discovery. Collectively, these examples highlight the diverse nature of inputs to collective intelligence, spanning playful interactions to expert contributions. Whether through gamified platforms or specialized knowledge, participants from various backgrounds collaborate to address a wide range of complex challenges, demonstrating the versatility and power of collective problem-solving.
5.1.2 Limitations of Human Capacity during Crowdsourcing
While collective intelligence holds immense potential for generating valuable insights and solutions, effectively harnessing it is complex and challenging. Put differently, the efficiency of crowdsourcing is inherently limited by the human capacity for processing vast amounts of information, a challenge that becomes particularly evident when addressing large-scale societal issues (Vaughan, Reference Vaughan2018). Notably, the potential of crowdsourcing to address significant societal issues is limited by several factors. First, the cognitive limitations of crowd workers mean that individuals can only process a certain amount of data before becoming overwhelmed or making errors in judgment (Bell et al., Reference Bell, Pescher, Tellis and Füller2023). This is particularly problematic when dealing with complex, large-scale issues that require deep understanding and nuanced solutions. The sheer volume of data and the intricacy of the problems can lead to information overload, reducing the effectiveness of the collective effort.
Second, the reliance on voluntary, unpaid labor introduces additional challenges. While the altruistic nature of crowd workers can drive initial enthusiasm and participation, the lack of financial incentives may lead to burnout, disengagement, and a decline in sustained contributions over time (Goh et al., Reference Goh, Lee, Zhou and Guo2021). Without the motivation provided by compensation, volunteers may prioritize other commitments or lose interest, ultimately affecting the continuity and quality of the work (Lee et al., Reference Lee, Goh, Zhou, Sin and Theng2020). This can create gaps in the collaborative process, reducing the overall effectiveness and impact of the crowdsourcing initiative.
It is important to note that, while crowdsourcing can also involve paid contributors who are financially compensated for their efforts, this chapter primarily focuses on the dynamics and limitations of unpaid, volunteer-driven crowdsourcing to solve societal problems. Paid contributors can bring a different set of motivations, work ethics, and sustainability factors, which are not the focus here (Brabham, Reference Brabham2008). Instead, this chapter explores how users and motivational factors influence voluntary knowledge sharing in crowdsourcing efforts aimed at solving complex societal challenges. Through this process, factors such as the challenges of cognitive overload and reliance on unpaid labor that can hinder the long-term success and scalability of crowdsourcing efforts will also be clarified.
5.1.3 Integrating AI into Crowdsourcing
The integration of artificial intelligence (AI) technologies into our everyday life lives has rapidly advanced in recent years. AI now assists humans in decision-making across various fields, such as healthcare, finance, transportation, education, and customer service (Mahotra & Majchrzak, Reference Mahotra and Majchrzak2024). The emergence of AI has the potential to alleviate the constraints of crowdsourcing and further transform its processes. AI enhances collective intelligence by boosting human capabilities, promoting collaboration, and efficiently processing large volumes of data (Vaughan, Reference Vaughan2018). AI’s pervasive influence necessitates a thorough understanding of its role in crowdsourcing. We assert that AI-supported crowdsourcing has the potential to improve knowledge sharing and enhance contributions to collective intelligence efforts. Even though there is an expanding body of research on motivations and crowdsourcing, it is still unclear how AI-supported crowdsourcing may affect crowd workers’ motives and, ultimately, their participation. Considering crowd workers’ perspectives in AI-supported crowdsourcing is crucial for multiple reasons. It enhances engagement by aligning with user motivations and preferences and ultimately improves usability and system design. Additionally, this understanding will help address ethical concerns such as privacy and consent, which can lead to the platform’s long-term sustainability and address long-term societal challenges.
5.1.4 Objectives
This chapter is, therefore, motivated by the limited understanding of AI-supported crowdsourcing. It has two main objectives. The first objective is to understand gaps and emerging trends in AI-supported crowdsourcing. We conducted a systematic review to shed light on existing studies that have employed AI during crowdsourcing and understand how AI aids in developing collective intelligence. From the review, we identified themes related to emerging trends, opportunities, and challenges surrounding the use of AI-supported crowdsourcing. Next, we collected empirical data from a real-life AI-supported crowdsourcing platform. Through the analysis of this case study, this chapter aims to shed light on users and motivational factors influencing knowledge sharing in crowdsourcing efforts. The focus of the case study discussion centered on leveraging crowd workers to address complex societal challenges and explored the potential obstacles that may impede the long-term success and scalability of these initiatives. Specifically, we investigated a global societal problem, child trafficking, to understand the user profiles and motives driving the use of AI-supported crowdsourcing to facilitate knowledge sharing.
5.2 Background on AI-supported Crowdsourcing
This chapter defines artificial intelligence (AI) as a system’s ability to correctly interpret external data and adapt the data effectively to achieve specific goals and accomplish tasks (Haenlein & Kaplan, Reference Haenlein and Kaplan2019). Two aspects of this definition need further discussion. In relation to human–AI interaction, AI makes it possible for machines to learn from experience, adjust to new inputs, and perform human-like tasks (Yang et al., Reference Yang, Steinfeld, Rosé and Zimmerman2020). In relation to data volume, AI enables computers to learn and perform specific tasks by analyzing large volumes of data and identifying patterns within it (Vaughan, Reference Vaughan2018). Thus, this definition highlights the synergy between AI and crowdsourcing, combining the strengths of both approaches. Specifically, AI’s ability to process and analyze large datasets is combined with human creativity, critical thinking, and diverse perspectives. This integrative perspective of AI and crowdsourcing not only amplifies the strengths of each but also creates new opportunities for innovation and problem-solving on societal challenges.
5.2.1 Differences between AI-supported Crowdsourcing and Traditional Crowdsourcing
Notably, AI-supported crowdsourcing differs significantly from traditional crowdsourcing. Generally, AI-supported crowdsourcing is far more scalable, as it is capable of handling complex and large-scale tasks that would be impractical with traditional crowdsourcing. These differences will be examined across key crowdsourcing components: the process, outcome, and user involvement.
In terms of the crowdsourcing process, AI algorithms can automate filtering, sorting, and categorizing large amounts of data, making the process faster and more efficient (Vaughan, Reference Vaughan2018). Traditional crowdsourcing, on the other hand, often relies on manual review and analysis of contributions from large numbers of participants, which can be time-consuming and prone to human error.
As for the crowdsourcing outcome, AI-supported crowdsourcing can help validate and cross-check crowd workers’ contributions, leading to improved accuracy (Mahotra & Majchrzak, Reference Mahotra and Majchrzak2024). Specifically, AI can detect inconsistencies or errors in the data submitted by human participants, automatically flagging or correcting anomalies based on predefined rules to ensure the integrity and reliability of the information being collected, enhancing the overall quality of crowdsourced outputs
From the users’ perspective, AI-supported crowdsourcing can tailor the crowdsourcing tasks to individual users based on their skills, preferences, and past performance, making the experience more engaging and relevant. For instance, AI can dynamically adapt to users’ progress over time. If a user consistently performs well in certain types of tasks, the system can offer more challenging assignments, fostering a sense of growth and mastery. Ultimately, AI can enhance user experience.
Collectively, prior studies suggest the potential for AI-supported crowdsourcing to be utilized in addressing complex situations and societal problems. Further research is needed to explore and realize this possibility.
5.2.2 Examples of AI-supported Crowdsourcing
There are several noteworthy examples of potential collaborations between AI and crowdsourcing. One is in the healthcare sector, where collaboration between AI and crowdsourcing offers the potential to improve personalized medicine and enhance the effectiveness of treatments. For instance, by combining AI’s ability to process complex, high-volume medical data with crowdsourced experiential data from crowdsourcing platforms (e.g., www.patientslikeme.com/) shared by patients, healthcare providers can gain a more holistic understanding of diseases and treatments.
Integrating AI also offers significant value in the broader context of social Q&A sites (e.g., Reddit, Stack Overflow), which are also considered crowdsourcing platforms. Specifically, AI-driven features like chatbots or virtual assistants can engage users more effectively, providing immediate responses and encouraging participation in the community. Furthermore, AI can analyze user interactions and patterns to provide valuable insights into user behavior, trends, and content quality, which can help refine platform strategies and improve user satisfaction. Thus, integrating AI into social Q&A sites can significantly enhance the platform’s functionality, user experience, and ability to manage and process information.
In terms of the actual implementation of AI-supported crowdsourcing, one example is Zooniverse, which leverages AI to assist volunteers in analyzing large datasets for scientific research, from identifying galaxies in space to tracking wildlife patterns on Earth (Ceccaroni et al., Reference Ceccaroni, Oliver, Roger, Bibby, Flemons, Michael and Joly2023). Further, recent crowdsourcing research in organizations suggested leveraging the power of AI to more efficiently screen out bad ideas and focus on only good ideas in the crowdsourcing process within ideation (Bell et al., Reference Bell, Pescher, Tellis and Füller2023). The collective intelligence generated from contributions is enhanced by AI tools that model and process the data, leading to breakthroughs in research and problem-solving. Bjarnason et al. (Reference Bjarnason, Gambrell and Lanthier-Welch2024) also discussed using AI to enhance collective intelligence. They presented an innovative toolkit that leverages AI to enable “smart” crowdsourcing, making the problem-solving approach more scalable, effective, and efficient.
In sum, AI-supported crowdsourcing represents a transformative advancement to enhance the efficiency and accuracy of crowdsourcing processes. By integrating AI’s analytical capabilities with the collective intelligence of crowd workers, this approach can address complex challenges more effectively and open new avenues for innovation.
5.2.3 Lack of Research on AI-supported Crowdsourcing
While the above examples have underscored AI’s potential to enhance collective intelligence in crowdsourcing, the factors influencing the adoption or non-adoption of AI-supported platforms remain unclear and require further in-depth investigation. Additionally, the literature on AI-supported crowdsourcing is limited. In particular, a holistic understanding of how AI could be applied to facilitate the development of collective intelligence is still lacking, but it is necessary to guide further research on crowdsourcing. More importantly, a user’s perspective on AI-supported crowdsourcing and how it can enhance crowdsourcing initiatives for the public good is needed. Section 5.3 addresses these gaps through a systematic review.
5.3 Systematic Review of AI-supported Crowdsourcing
5.3.1 Systematic Review
A systematic review is a rigorous and methodical approach to reviewing and synthesizing existing research on a specific topic. We conducted a systematic review to help identify gaps in the literature and guide future research directions in AI-supported crowdsourcing. The PICO framework (Population, Intervention, Comparison/Control, Outcome) was applied to guide the systematic review (Lee et al., Reference Lee, Lim and Kim2024a). The population includes crowd workers, and individuals who participate in crowdsourcing tasks, and datasets contributed through crowdsourcing. Intervention refers to using artificial intelligence or algorithmic methods to facilitate or address crowdsourcing. Comparison/Control involves benchmarking the outcomes. Outcome is the resultant output behavior, perceptions, or processes after interventions. This systematic review also follows the PRISMA guidelines, which include a 27-item checklist and a flowchart to ensure a thorough and transparent review process (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow, Shamseer, Tetzlaff, Akl, Brennan, Chou, Glanville, Grimshaw, Hróbjartsson, Lalu, Li, Loder, Mayo-Wilson, McDonald and Moher2021).
The keywords for the searches were developed based on the research scope. The keywords also encompassed synonyms and alternative spellings, strung together using Boolean operators (OR/AND) to ensure comprehensive searches. The search query was as follows: “crowdsourc*” AND “collab*” AND (“artificial intelligence” OR “AI” OR “generative artificial intelligence” OR “generative AI” OR “GAI” OR “large language model” OR “LLM” OR “machine learning” OR “ML”) AND (“societal” OR “welfare”). The searches were conducted in May 2024 across six major databases (Scopus, Web of Science, ACM Digital Library, IEEE Xplore, ScienceDirect, and Google Scholar) between January 2000 and May 2024. In addition, backward and forward citation searches (Bandara et al., Reference Bandara, Miskon, Fielt, Virpi, Joe, Matti and Wael2011) were conducted in May 2024. The search results were captured and managed using a bibliographic management tool, Endnote (Bandara et al., Reference Bandara, Miskon, Fielt, Virpi, Joe, Matti and Wael2011).
The inclusion criteria were: (1) published between January 2000 and May 2024; (2) written in English; (3) published in peer-reviewed journals; (4) available in full-text; (5) related to the scope of this study; and (6) not belonging to any grey literature. The exclusion criteria were: (1) duplicated studies; (2) grey literature (including studies published in preprint or related to the conceptual, working discussion, or literature review); (3) not written in English; and (4) not related to the scope of this study. These criteria were used to screen retrieved records from the databases, as well as during the full-text assessment. Further, PICO was used as an additional selection criterion to ensure the quality of the selected studies.
The five databases’ searches produced 190 records, while the backward and forward citation searches identified 11 records (see Figure 5.1). After removing duplicated records (n = 11) and applying the filtering criteria, sixty-four records were further excluded. The remaining records (n = 115) were sought for full-text retrieval for detailed assessment. The final process yielded 6 studies for inclusion in this review after excluding 109 studies not within this review’s scope. The limited number of studies currently available indicates that the concept of AI-supported crowdsourcing requires more comprehensive research and investigation to fully understand its potential and implications.
Figure 5.1 PRISMA flow diagram for studies selection
Figure 5.1Long description
For identification of studies via databases and registers, it begins with the record identified from n equals 190. The records removed before screening have duplicate records of n equals 11. Records screening has n equals 179, which follows reports sought for retrieval and assessed for eligibility of n equals 115. It results in reports of included studies of n equals 6. For identification of studies via other methods, it starts with records identified from n equals 11. It follows record screening of n equals 11 and then reports sought for retrieval and assessed for eligibility. It by reports excluded of n equals 11, leads to reports of included studies of n equals 6.
The included studies were analyzed, and data were extracted using six pre-defined categories. These include the year of publication, study context, sample characteristics, terminology related to the scope of this study, contributions or outcomes, and limitations. Extracted data were then clustered according to common themes for narrative synthesis. Table 5.1 shows a summary of the studies included in this review. Table 5.2 shows a detailed list of the reviewed studies and the terms associated with societal challenges and problem-solving tasks.
| Author / Year | Sample | How does AI support crowdsourcing? |
|---|---|---|
| Gimpel et al. (Reference Gimpel, Graf-Seyfried, Laubacher and Meindl2023) | 311 workers | Macro-task crowdsourcing is labor-intensive and complex to facilitate (e.g., sustainable development goals defined by the United Nations), and technology such as AI might overcome these limits. This study evaluates AI affordances and how they could facilitate macro-task crowdsourcing and advance the discourse on facilitation. |
| Haider et al. (Reference Haider, Clifton and Yin2024) | 238 participants from Prolific platform | The increasing prevalence of automatic decision-making systems has raised concerns regarding the fairness of these systems, and the crowdsourced approach to solicit fairness definition are highly context-dependent. This study investigates the hypothesis toward people’s fairness perceptions in three societal contexts, each differ on the expected level of risk associated with different types of decision mistakes. |
| Köhl et al. (Reference Köhl, Fuger, Lang, Füller and Stuchtey2019) | Data from openIDEO platform | Crowdsourcing is increasingly being used to find solutions for pressing wicked problems (e.g., environmental pollution). However, most solutions focus on a few ideas and ignore the large volume of content created by the community. This study applies an automated text-mining technique to analyze the ideas contributed by the community. |
| Vinella et al. (Reference Vinella, Hu, Lykourentzou and Masthoff2022a) | Data for modeling and simulation | Crowdsourcing offers potential solutions to solve complex tasks (e.g., pandemic prevention and emergency response) that require teamwork and collective labor. However, forming project teams from the vast scale of the crowd is a difficult problem. This chapter investigates different ways that crowd teams can be formed from the algorithmic simulation of three team formation models. |
| Vinella et al. (Reference Vinella, Odo, Lykourentzou and Masthoff2022b) | 120 participants from Amazon Mechanical Turk | Intense and high-pressure tasks (e.g., environmental disasters) are often solved by teams that are cohesive, adaptable, and prepared, but little is known about how teams of crowdsourced strangers would cooperate and contribute to the teamwork. This study explores which factors matter for team success and perceptions of collaboration quality, as well as how it could support the future work on AI-supported crowdsourcing of remote emergency response teams. |
| Yu et al. (Reference Yu, Liu, Wei, Zheng, Chen, Yang and Peng2019) | Data from the prototype system | Businesses face large fluctuations in manpower demand and require efficient ways to meet the demands. This study proposes an AI-empowered crowdsourcing platform to perform efficient explainable task-worker matching, which aims to address societal challenges (e.g., unfair treatment of workers). |
| Author | Societal Impact | Description |
|---|---|---|
| Gimpel et al. (Reference Gimpel, Graf-Seyfried, Laubacher and Meindl2023) | Macro-task related to sustainable development goals | Refers to tasks (e.g., sustainable development goals) that are difficult or at times not possible to break down into smaller sub-tasks (Robert, Reference Robert, Khan, Papangelis, Lykourentzou and Markopoulos2019). |
| Haider et al. (Reference Haider, Clifton and Yin2024) | Events related to societal and cultural context | Refers to social or cultural factors that influence decision-making associated with the events (e.g., perceived risks in requirement prediction). |
| Köhl et al. (Reference Köhl, Fuger, Lang, Füller and Stuchtey2019) | Wicked problems | Refers to problems (e.g., environmental pollution) which resulted from innumerable causes which are difficult to describe their complete magnitude, and there is no easy way out (Rittel & Webber, Reference Rittel and Webber1973; Villarrubia-Gómez et al., Reference Villarrubia-Gómez, Cornell and Fabres2017). |
| Vinella et al. (Reference Vinella, Hu, Lykourentzou and Masthoff2022a; Reference Vinella, Odo, Lykourentzou and Masthoff2022b) | Complex task | A collective references to macro-task or wicked problems, as well as tasks that are critical, time-bounded, and high pressure (e.g., environmental disaster). |
| Yu et al. (Reference Yu, Liu, Wei, Zheng, Chen, Yang and Peng2019) | Societal challenge with a focus on user profiles | Refers to the problems such as those posed by demographic changes, job securities and wellbeing that require concerted efforts to address it (e.g., ad-hoc work arrangements through untrustworthy agents exposes workers to high risk of exploitation and swindling). |
| Kokshagina (Reference Kokshagina2022) | Grand challenges related to global problems | Refers as global problems (e.g., COVID-19), characterized as complex, uncertain, and evaluative (Ferraro et al., Reference Ferraro, Etzion and Gehman2015), that are only effectively addressed through a coordinated and collaborative effort (George et al., Reference George, Howard-Grenville, Joshi and Tihanyi2016) which require technical and societal components to deal with them (Eisenhardt et al., Reference Eisenhardt, Graebner and Sonenshein2016). |
5.3.2 Themes from the Systematic Review
The review identified three key themes that highlight both the trends and gaps in AI-supported crowdsourcing: solving complex and challenging issues, lacking practical research, and addressing ethical considerations. By understanding and addressing these themes, researchers and practitioners can enhance crowdsourcing initiatives’ effectiveness, efficiency, and ethical standards to solve societal tasks.
A central theme that emerged across the six studies examining integrating AI into crowdsourcing was its potential to tackle complex and challenging societal problems. Although varying in nature, these problems were all characterized by their complexity and the significant obstacles they present, requiring innovative and collaborative approaches to find solutions. These issues ranged from global ones, such as sustainable developmental goals and the COVID-19 pandemic, to localized challenges like demographic changes and emergency responses. Put simply, AI-supported crowdsourcing can enhance collaborative human efforts and solve difficult and complex societal challenges. The review also revealed that successfully leveraging AI requires careful consideration of how to align the AI technology with the specific needs and motivations of the crowd, ensuring that the resulting solutions are not only innovative but also socially and ethically responsible. Therefore, a thorough examination of the motives of crowd workers to understand their needs will be a crucial first step.
Next, although technological advancement has accelerated tremendously in recent years, particularly for AI, the review shows that the integration of AI technology into crowdsourcing platforms has not fully reached its potential to overcome existing limitations, such as task scalability, quality control, and the efficient use of collective intelligence. Indeed, there are challenges related to the ethical use of AI, ensuring transparency, and maintaining trust among users – these are barriers to the implementation of AI-supported crowdsourcing. This suggests that, while AI-supported crowdsourcing holds great promise, much work remains to fully leverage AI’s capabilities in solving complex societal problems. Specifically, more practical research on utilizing existing platforms is needed to close this gap and harness the full potential of AI in enhancing crowdsourcing efforts for societal good.
Finally, this review examined two ethical considerations about using AI in crowdsourcing, reflecting the complexity and dual nature of the impact of technology on society. Haider et al. (Reference Haider, Clifton and Yin2024) highlighted the fairness of automatic decision-making systems, as the perception of fairness is highly context-dependent. On the other hand, Yu et al. (Reference Yu, Liu, Wei, Zheng, Chen, Yang and Peng2019) suggested exploring how artificial intelligence could be applied to ensure fairness, transparency, and accountability in facilitating crowdsourcing tasks. From a broader perspective, these concerns are interrelated with the transparency and accountability of AI systems. This means that the complexity of AI algorithms can make it difficult for participants to understand how their contributions are being used, raising questions about consent, data privacy, and trust. These ethical concerns may influence the motives of crowd workers, and they need to be addressed.
5.3.3 Summary from the Systematic Review
In summary, integrating AI into crowdsourcing can significantly alter the dynamics of the factors affecting usage. AI introduces new elements into the process that may influence individuals’ willingness to share knowledge and contribute to the development of collective intelligence. However, the impact of these AI-supported crowdsourcing elements on participants’ profiles and motives has not been fully understood, representing a gap in the literature.
To address this gap, we employed a case study that examined the factors underlying the use of an AI-supported crowdsourcing platform to address a significant societal issue – child trafficking. This is presented in Section 5.4.
5.4 Case Study
Child trafficking is a critical issue that affects societies worldwide. It is a social justice issue that significantly affects human autonomy (Flores, Reference Flores2009). In the United Kingdom, an estimated 112,853 children are reported missing every year (Global Missing Children’s Network, 2023). The number of declared missing children was 359,094 in the United States and more than 25,000 in Africa in 2022 (AfricaNews, 2022; Roush, Reference Roush2023). Due to the large scale of child trafficking victims, combating it is challenging as limited human capacity and resources are available for tracing suspected trafficked children and providing timely rescue.
To address this issue, crowdsourcing is a feasible option, given its ability to amass a community of users to collectively address a designated task (Assis Neto & Santos, Reference Assis Neto and Santos2018). By connecting willing crowd workers (volunteers) and relevant stakeholders focused on trafficked children, crowdsourcing offers a platform for collaborative rescue efforts (Goh et al., Reference Goh, Lee and Guo2018). Nonetheless, it is still hard for human beings to process large amounts of information regarding victims, and it is even harder for them to recognize trafficked children as their appearance changes with age. Due to the limits of human information processing capacity, we examined an AI-supported crowdsourcing platform to combat child trafficking. The study aimed to address one question – will incorporating AI technology affect crowd workers and their motivations to participate?
5.4.1 Motivations of AI-supported Crowdsourcing
Understanding how user motivations interact with AI-supported crowdsourcing platforms can contribute to effective designs by tapping into crowd workers’ altruistic desires to help others with their practical, outcome-driven considerations. More broadly, studying motives is essential for the successful implementation of AI-related platforms to solve societal problems. By understanding what drives participation, developers of AI tools can design AI-integrated platforms that appeal to users, ensuring that the platform resonates with users on a deeper level, hence fostering a sense of engagement, ownership, trust, and long-term commitment to the cause.
Further, understanding the motives of potential users helps in designing AI platforms that align with their values and needs. For example, if a platform is intended to address a societal problem, knowing that users are motivated by altruism (wanting to help others) can lead to features that emphasize the social impact of their contributions. On the other hand, if users are driven by utilitarian motives, such as skill development or career advancement, the platform can include elements like certifications, badges, or networking opportunities (Goh et al., Reference Goh, Pe-Than and Lee2017).
5.4.2 Introducing Zhongxun
This section introduces Zhongxun, an AI-supported crowdsourcing platform. Zhongxun’s AI has three functions: (1) accounting for the effect of aging on children, (2) calculating the similarities between the uploaded children’s photos with trafficked children’s photos stored in the database, and (3) learning from crowdsourcing users’ feedback to improve facial recognition results (Goh et al., Reference Goh, Lee, Zhou and Guo2021). Figure 5.2 illustrates the collaboration between users and Zhongxun’s AI in finding missing children. When encountering suspected trafficked children, users can take and upload photos to Zhongxun (see Figure 5.2a). Next, Zhongxun uses its facial recognition algorithm to compare the uploaded photos with the stored information of known trafficked children. After that, the matched children and the similarity rates are posted on Zhongxun’s official Weibo account of Zhongxun (see Figure 5.2b). Knowledge shared by the user is used to improve the matching results (see Figure 5.2c). Once the matching process is accomplished, notification to relevant parties will be done (e.g., authority and parents of the reported missing children).
Figure 5.2 The process of finding trafficked children
Figure 5.2Long description
The screenshots present A. Submission of a Zhong Xun. A matching photo is displayed on Zhong Xun's Weibo account. C. User provides feedback on the similarity between photos. The photo of a young boy sent in A is compared with another photo in B, which has feedback in C.
5.4.3 Research Participants
An online survey with snowball sampling was conducted to reach Weibo and WeChat users who were interested in participating in this study. In the survey, an introduction to the study was presented, followed by an introduction video about Zhongxun. Next, participants were asked to answer questions assessing (1) their perception of the features of Zhongxun, (2), their intention to share knowledge and contribute to Zhongxun, and (3) their motives for doing so. A total of 407 participants participated in the study. Around 35 percent of the participants were males (N = 142), while the rest were females (N = 265). Most of the participants (79%) were between 23 and 30 years old.
5.4.4 Construct Operationalization
The scale for measuring voluntary motives, as adapted from Clary et al. (Reference Clary, Snyder, Ridge, Copeland, Stukas, Haugen and Miene1998), is relevant to the current chapter. It is a comprehensive tool designed to assess the various reasons individuals may have for engaging in volunteer activities. The scale comprises thirty items from six subdimensions, including (1) values function (desire to express values), (2) protective function (protect oneself from negative feelings such as guilt), (3) enhancement function (enhancing personal growth), (4) social function (desire to engage with others), (5) career function (benefits that support personal development), and (6) understanding function (desire to acquire knowledge). Additionally, altruistic and utilitarian motives are incorporated into the motives measurement. Altruistic motives (5 items) are rooted in a desire to help others, contribute to the greater good, or support a cause without expecting personal gain (Bucher et al., Reference Bucher, Fieseler and Lutz2016) while Utilitarian motives (5 items) are based on practical considerations and the pursuit of specific benefits or outcomes (Bucher et al., Reference Bucher, Fieseler and Lutz2016). Additionally, demographic information such as age, gender, nationality, country of residence, and experience with crowdsourcing was collected from the participants.
5.4.5 Analysis and Results
Hierarchical regression analysis was conducted to investigate the factors that influenced the use of AI-supported crowdsourcing platforms. The first block comprised demographic variables and prior crowdsourcing experience, followed by altruistic and utilitarian motives in the second block and voluntary motives in the third block.
In terms of demographic variables and prior experience, results suggested that local crowd workers (who lived in China) were more likely to share knowledge on Zhongxun crowdsourcing platform (β = 0.15, p = 0.003 < 0.01) (see Table 5.3). As for the motives (second block), results showed that altruistic motives (β = 0.34, p < 0.001) had statistically significant and positive relationships with the intention to use AI-supported crowdsourcing to share knowledge. In contrast, utilitarian motives had no significant effects. As for the voluntary motives, only the values function (β = 0.22, p < 0.001) was statistically significant and positively related to the intention to use AI-supported crowdsourcing to share knowledge.
Note: * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 5.3Long description
The table has three columns for the independent variable, standardized beta, and t value. It has three sections titled model 1, model 2, and model 3.
For model 1.
Row 1 reads. Nationality Influence. Minus 0.10. Minus 1.54.
Row 2 reads. Geographical Influence. 0.17. 2.72, double asterisk.
Row 3 reads. Age. Minus 0.02. Minus 0.43.
Row 4 reads. Gender. 0.00. 0.04.
Row 5 reads. Prior Experience. 0.12. 2.43, asterisk.
Row 6 reads. R-squared. 0.029. Blank.
Row 7 reads. Increase Adjusted R 2. 0.029, asterisk. Blank.
For model 2.
Row 1 reads. Nationality Influence. Minus 0.09. Minus 1.63.
Row 2 reads. Geographical Influence. 0.17. 3.48, double asterisk.
Row 3 reads. Age. Minus 0.02. Minus 0.34.
Row 4 reads. Gender. 0.07. 1.81.
Row 5 reads. Prior Experience. 0.06. 1.58.
Row 6 reads. Utilitarian Motives. 0.06. 1.28.
Row 7 reads. Altruistic Motives. 0.58. 13.86, triple asterisk.
Row 8 reads. R-squared. 0.383. Blank.
Row 9 reads. Increase Adjusted R-squared. 0.354, triple asterisk. Blank.
For model 3.
Row 1 reads. Nationality. Minus 0.04. Minus 0.80.
Row 2 reads. Geographical Influence. 0.15. 3.02, double asterisk.
Row 3 reads. Age. Minus 0.03. Minus 0.63.
Row 4 reads. Gender. 0.07. 1.75.
Row 5 reads. Prior Experience. 0.05. 1.37.
Row 6 reads. Utilitarian Motives. 0.00. Minus 0.07.
Row 7 reads. Altruistic Motives. 0.34. 5.41, triple asterisk.
Row 8 reads. Protective Function. Minus 0.06. Minus 1.03.
Row 9 reads. Values Function. 0.22. Minus 3.82, triple asterisk.
Row 10 reads. Career Function. 0.09. 1.20.
Row 11 reads. Social Function. 0.00. Minus 0.08.
Row 12 reads. Understanding Function. 0.10. 1.51.
Row 13 reads. Enhancement Function. 0.07. 0.97.
Row 14 reads. R-squared. 0.430. Blank.
Row 15 reads. Increase Adjusted R-squared. 0.047, triple asterisk. Blank.
Row 16 reads. Total Adjusted R-squared. 0.430. Blank.
The note below reads. Asterisk p less than 0.05. Double-asterisk p less than 0.01. Triple asterisk p less than 0.001.
5.4.6 Lessons Learnt from the Case Study
The factors and motivations driving knowledge sharing in AI-supported crowdsourcing were uncovered, revealing several key insights and important lessons.
First, the intention to use AI-supported crowdsourcing to share knowledge is influenced by geography. A possible explanation is that, although child trafficking is a global issue, users tend to be more engaged in volunteerism when the issue is geographically closer to them (Mohan et al., Reference Mohan, Twigg, Jones, Barnard, Milligan and Conradson2006). This suggests that contributions from local crowd workers to collective intelligence may be more valuable than those from nonlocal crowd workers, due to their familiarity with the region and its unique context. However, prior research has shown that geography does not always influence participation. One study found that users involved in a mapping project were often more active in tasks located far from their homes, indicating that distance may not be a limiting factor in some crowdsourcing efforts (Baruch et al., Reference Baruch, May and Yu2016). So, the geographical influence on AI-supported crowdsourcing participation may vary depending on the nature of the task, with certain tasks being more accessible or appealing to participants from specific regions due to factors such as cultural context, available resources, technological infrastructure, or local expertise.
Second, altruistic motives positively influence the intention to share knowledge on AI-supported crowdsourcing platforms, as users are often driven by a sense of social responsibility and a desire to contribute to the greater good. In contrast, utilitarian motives – such as personal gain or efficiency – do not appear to significantly affect this intention, suggesting that the willingness to share knowledge in this context is more strongly associated with intrinsic, rather than extrinsic rewards (e.g., financial compensations). More importantly, AI does not alter deeply ingrained altruistic motives. While AI can facilitate opportunities for altruistic actions (e.g., by enabling easier and more efficient ways to contribute to causes), the underlying motivations for altruism remain rooted in human psychology and social dynamics.
Third, the strong influence of the values function suggests that individuals are more likely to engage in crowdsourcing efforts when they perceive that their participation aligns with their core beliefs and ethical principles. This alignment not only motivates initial involvement but also sustains long-term commitment to the cause, making the values function a critical factor in the success of AI-supported crowdsourcing projects that rely on volunteer contributions. Specifically, the features of Zhongxun (e.g., facial aging and matching algorithms) extend individuals’ cognitive capabilities, allowing them to contribute and enhance the collaborative effort. Empirical data from this case study indicates that participants expressed appreciation for AI’s cognitive support, provided that the process aligns with their personal values. The alignment of individuals’ personal values with the goals of the crowdsourcing initiative is an important factor. Specifically, individuals are driven to contribute based on a sense of moral obligation, social responsibility, or a desire to make a positive impact.
5.4.7 Summary of the Case Study
The empirical analysis of Zhongxun as a case study demonstrates that AI-supported crowdsourcing is a viable approach to addressing societal challenges. Key factors, such as geographical influences, along with motivations like altruism and value functions, are critical to its implementation.
5.5 Recommendations
This chapter examines the use of AI-supported crowdsourcing to solve societal challenges. As technology continues to advance, investigating the role of AI in collaboration with humans becomes essential to effectively integrate this technology into our everyday lives and work (Wang et al., Reference Wang, Churchill, Maes, Fan, Shneiderman, Shi and Wang2020). The findings from the systematic review and case study discussion have implications for practice and research.
5.5.1 Recommendations for Practitioners
This section offers practical contributions to the use of AI-supported crowdsourcing, which are summarized next.
First, practitioners should take into account users’ motivations in AI designs. Given that value motivation plays a pivotal role in users’ acceptance intentions, designers could use AI to offer timely updates and feedback related to the process of rescuing trafficked children. Highlighting the successful completion of each crowdsourcing task and how the participation contributes to the collective intelligence can help sustain users’ engagement and sense of contributing to others’ well-being.
Second, while AI development is promising, exciting, and attracting attention, local crowd workers play a crucial role in addressing societal issues within their communities. Put simply, local crowd workers are vital in addressing societal issues within their communities through AI-supported crowdsourcing. The knowledge of local culture, social dynamics, language, and unique challenges enables them to contribute insights and solutions that are deeply relevant to the issues faced by their region. These, in turn, help enhance the effectiveness of the AI models being used. Consequently, effort should be made to match crowd workers with preferred locales. AI should be seen as a complement to these workers, and not a substitute.
Third, designing altruistic features in AI platforms can significantly enhance their appeal to users by aligning with people’s innate desire to contribute positively to society. Several approaches are viable. AI can assist users in identifying opportunities where their skills can make a significant impact. Specifically, AI can match users with crowdsourced projects or social causes where their unique expertise or experience can be best applied. Another approach is to utilize AI to recommend causes or tasks that align with the user’s preferences, past contributions, or areas of interest. Personalizing these recommendations can increase engagement by ensuring users feel connected to the causes they support.
Fourth, the case study discussion on Zhongxun underscores the relevance of incentivizing the social sharing of good deeds through AI-supported crowdsourcing. Thus, platforms should deploy features that encourage crowd workers to share their altruistic acts and inspire others to join them and contribute, creating a network effect that amplifies social good. Alternatively, stories of individuals who have benefited from altruistic actions or testimonials from those who benefited from the platforms’ efforts should be made available. In some cases, altruism may need to be balanced with other forms of motivation, such as recognition. AI tools should be designed to accommodate various motivations for different profiles of crowd workers while ensuring that altruistic contributors are valued and their contributions are integrated effectively into collective intelligence systems.
Finally, it is important to establish clear ethical guidelines that ensure AI-supported crowdsourcing platforms operate in a way that upholds social good. Users will feel more inclined to engage in crowdsourcing activities if they know their chosen platform aligns with their values and follows ethical practices. While altruism and alignment of values enhance collective intelligence, there is a risk that platforms could exploit altruistic participants by over-relying on volunteer crowd workers. Practitioners must ensure that AI platforms uphold ethical standards by providing transparency in algorithms, safeguarding data privacy and security, and offering appropriate recognition for contributions, while also valuing and respecting the efforts of crowd workers.
5.5.2 Recommendations for Researchers
Here are key recommendations for researchers to advance the field and contribute to impactful research.
First, future research can explore the notion of localized AI-supported crowdsourcing. Local expertise allows AI systems to be customized and adapted to different regions, ensuring that solutions are not one-size-fits-all but tailored to each community’s specific needs. This is especially important in domains like healthcare, education, and public services, where regional differences can lead to varying outcomes. By integrating local knowledge, AI systems can be designed to respect cultural norms, legal frameworks, and societal expectations, thereby increasing their acceptance and effectiveness. In particular, local crowd workers can identify and prioritize issues that are of particular importance to their community, such as environmental hazards, public health crises, or infrastructure needs. More importantly, by incorporating local expertise, AI models can be enhanced to recognize specific local conditions, resulting in better predictions and recommendations.
Second, an important area of research is examining crowd workers’ commitment to long-term collaboration. A potential area here is developing and building emotional connections among crowd workers, and between them and the platform. This emotional connection fosters a deeper commitment, leading to more sustained involvement in the long run. Participants are more likely to remain engaged if they are emotionally connected and believe their ongoing contributions are crucial to achieving broader societal goals, making long-term research projects more viable and successful. Findings indicate that appropriate incentive structures for voluntary crowdsourcing will be critical.
For instance, AI platforms aimed at crowdsourcing solutions to societal problems might offer recognition or opportunities for personal growth. By understanding user motives, researchers can address potential concerns about trust, privacy, data use, and fairness, which are particularly important in AI systems.
Third, another promising extension of this research involves enhancing AI tools in crowdsourcing to better humanize participation. This can be achieved in several ways. Integrating AI to create feedback loops will be an effective way to continuously improve the platform’s ability to promote collective effort. For instance, AI can analyze user engagement patterns, identify what drives successful altruistic projects, and optimize task recommendations and platform features accordingly. Next, research should be conducted on AI tools that accurately measure and quantify the societal impact of altruistic contributions. This may involve creating AI systems that assess the real-world outcomes of crowdsourced tasks, providing users and researchers with data-driven insights into the effectiveness of their efforts.
Fourth, AI-supported crowdsourcing platforms that highlight societal concerns and impact can inspire cross-disciplinary collaborations as researchers from various fields come together with a shared goal. The focus on social outcomes encourages participants from different disciplines – such as data science, biology, public policy, and engineering – to contribute their expertise toward solving complex, multifaceted problems. Specifically, AI-supported platforms can aggregate and analyze large volumes of data contributed by interdisciplinary teams. For example, in a project exploring urban sustainability, an AI system could combine insights from urban planners, economists, environmental scientists, and data scientists to create comprehensive models of urban ecosystems. By synthesizing this data, AI platforms help researchers make sense of complex, multifaceted problems that require input from multiple perspectives. Additionally, AI-supported crowdsourcing platforms can incorporate predictive modeling and simulation tools to help interdisciplinary teams test hypotheses and simulate the impact of various interventions. This is particularly useful across various fields (e.g., medicine, environmental sustainability, urban planning) where AI can model complex systems and provide researchers from different disciplines with actionable insights.
Finally, more user-oriented research is needed to understand crowd workers’ segmentations and profiles across different societal challenges. Different users have different reasons for engaging in altruistic behavior (e.g., personal satisfaction, social recognition, or belief in the cause) (Synder et al., Reference Snyder, Clary, Stukas, Maio and Olson2000). Research into personalizing AI platform interfaces, which aims to enhance user engagement by promoting participation across diverse user groups, will be necessary. By customizing the interface to align with individual profiles, motivations, cultural values, and unique user needs, AI-supported platforms can better cater to contributors’ goals.
5.5.3 Summary of the Chapter
This chapter examines the viability of AI-supported crowdsourcing in addressing societal challenges through a systematic review and a case study of a real-world AI-supported crowdsourcing platform. The systematic review highlights the potential of integrating AI to tackle complex societal issues, though research in this area remains limited. The case study further emphasizes that AI can serve as a cognitive extension, complementing human intelligence rather than replacing crowd workers in solving societal problems.
5.6 Conclusion
In conclusion, AI-supported crowdsourcing promotes knowledge sharing among crowd workers to address societal challenges and leveraging AI to facilitate this process can be viewed as a potential solution to prevent the “Tragedy of the Commons.” The “Tragedy of the Commons,” a concept echoed by economist Hardin (Reference Hardin1968), describes a situation where individuals, acting solely in their self-interest without regard for others or the community, ultimately deplete shared resources, negatively affecting everyone in the community (Xia, Reference Xia2024). Societal challenges, such as climate change, traffic congestion, sustainability, and public health concerns, are often linked to the “Tragedy of the Commons” because they involve the overuse or degradation of shared resources (Spiliakos, Reference Spiliakos2019). In addressing societal challenges, AI-supported crowdsourcing offers the potential to overcome the “Tragedy of the Commons” by fostering voluntary participation, enhancing coordination, increasing transparency, and aligning individual actions with collective goals for sustainable resource management. Put simply, by harnessing collective intelligence, AI-supported crowdsourcing platforms (together with appropriate governing frameworks) will play important roles in managing shared resources and fostering collaboration to tackle complex societal issues.
