Challenges in Sustainable Development Goals implementation

Implementing the United Nations Sustainable Development Goals (SDGs) presents a multifaceted set of challenges. First, it requires diverse governance frameworks since no single governance style—whether hierarchical, network, or market-based—or any combination thereof, can effectively address all SDG-related challenges (Meuleman and Niestro 2015). Second, governance is no longer exclusive to governments but involves non-state actors working with or even without them (Florini and Pauli 2018).

The 2003Commission’s Report on European Governance highlighted the necessity for shared responsibility in achieving a new governance model characterised by openness, participation, accountability, effectiveness, and coherence. The outcome document of the 1992 UN Conference on Environment and Development, Sustainable Development—Agenda 21 encouraged the “active involvement of non-governmental organisations and other groups” (UN 1992, paragraph 1.3).

Shulla et al. (2020) argue, drawing on Sachs (2012), that a networked problem-solving approach involving all actors, especially young people, is necessary rather than a top-down strategy. This approach aligns with SDG 16, which focuses on “Promoting just, peaceful, and inclusive societies.”

Despite the undisputed significance of proper governance, including citizen participation, for accomplishing the SDGs, designing effective mechanisms to launch and carry out the required transformations is a complex, multi-scale, and multi-actor process with numerous knowledge gaps (Allen et al. 2023). Furthermore, many initiatives promoting citizen participation fall short, partly because of distrust and frustration arising from the lack of tangible outcomes and accountability.

In this context, along with the continuous development of communication technologies, collective intelligence seems an optimal approach to address these collective challenges. However, its potential remains untapped, particularly in environmental decision-making (Vercammen and Burgman 2019). In fact, there are hardly any references on collective intelligence methodologies to address decision-making and/or proposal development in the context of SDGs. Some cases are found, for example, in the field of sustainable innovation (Erbguth et al. 2022; Wang et al. 2018), geographic knowledge systems (Laurini 2021), or social economy (Miedes-Ugarte, Flores-Ruiz and Wanner 2020).

Objective of this paper

In this paper, we investigate whether it is possible to use collective intelligence to generate solutions to the challenges associated with the SDGs, and, simultaneously, to create or strengthen the necessary governance mechanisms. We explore the methodology used in a massive online collective intelligence experiment, which includes the use of the Thinkhub platform (https://ic.kampal.com/). We are also interested in deepening our understanding of the effectiveness of collective intelligence under specific situations.

The structure of the article is as follows: First, we provide an introduction to the foundations of collective intelligence, as well as an explanation of its potential to address SDGs and governance issues; second, we describe the experiment methodology; next, we detail the experiment results; and finally, we present the discussion, conclusions, and future work.

Collective intelligence foundations

The concept of collective intelligence has been a polysemic expression since it was coined by Kropotkin in Mutual Support: A Factor in Evolution, a work that introduced in biology the importance of cooperation as a factor in survival (Kropotkin 2016). Currently, the notion is also used in other fields such as anthropology, social psychology, literature, management, political studies, and the relationship of human beings with technology. The concept is unstable and polysemic; some examples of its diverse meanings are the following: the capacity for decision-making in democratic spaces (Atlee 2002), a new form of production of knowledge and its value embedded in cyberculture (Lévy 2002; Mazzone 2019), and the properties of highly organised gregarious species to form superorganisms, from bees to human beings (Bert and Wilson 2014; Toca Torres 2014; Wheeler 1926). These meanings have in common that they always slide around the ideas of group, coordination and self-organisation, and a higher cognitive capacity as emergent results of this interaction.

In a similar vein to what Bloom (2000) called group IQ, researchers at Carnegie Mellon University and the Massachusetts Institute of Technology (MIT) operationalized in 2010 the concept of collective intelligence as the better or worse performance of a group of people interacting with each other for the resolution of different tasks (Woolley, Aggarwal, and Malone 2015). This work opened a field to study the capacity that human groups or sociotechnical systems have to act more intelligently than the most intelligent individual in the group would. In an experimental investigation, they assigned the groups they worked with, which had to solve a particular set of tasks, a metric called the c-factor. This metric represents the ability with which they executed different challenges (Woolley et al. 2010). Similar to the g factor of individual IQ (Brand 1996), the c-factor allows the intelligence of a group of people to be measured beyond the specific task assigned and can be used to estimate the ability of that group to solve future tasks.

In the pursuit of improving and understanding how collective intelligence works, two streams of analysis have emerged (Woolley and Aggarwal 2020). The first one focuses on the individual abilities of participants and their influence on the resulting collective outcomes. The second stream examines how the interaction processes taking place within the activity itself influence the resulting collective outcomes. In these individual problem-solving interactions, the specific traits of the social situation created impact the generation of various responses.

In these situations, striking a balance between the tendency for herd mentality and individual isolation is crucial (Toyokawa et al. 2019). Too much of the former can hinder creativity, while excessive dispersion can limit interaction and prevent reaching a consensus. Channels that enhance social influence among participants raise the average quality of responses but may suppress the most successful possibilities. Bernstein et al. (2018) suggest that intermittent influence is the most effective approach. In general, various behaviours such as creativity, leadership, followership, cooperation, coordination, discussion, agreement-seeking, or overcoming challenges are observed. These behaviours largely depend on the rules and participation methods established by the researchers.

Collective intelligence with large online groups is still a relatively new methodology, with few studies available. However, some researchers have highlighted the significance of group size (Toyokawa et al. 2019). Experiments involving face-to-face interactions typically focus on groups of two to five people, without leveraging the online potential for multi-group coordination (Engel et al. 2014).

In the only large-scale online study we found, conducted by Toyokawa et al. (similar to the experiment we present), there is no moderator or facilitator, and participant interaction is limited. Toyokawa and colleagues distinguish between two behavioural approaches: the herd effect, characterised by a high level of replication and suboptimal response effectiveness, and a phenomenon representing crowd intelligence, an optimal response rate that surpasses the simple sum of unrelated individuals.

Collective intelligence and citizen science to address Sustainable Development Goals and governance issues

Technological and sociopolitical strategies need to be developed to promote interrelation, interdependence, and communication in collective action; examples of such strategies are found in citizen science, such as various platforms and networks, shared resources, citizen laboratories, etc. (Pelacho et al. 2021). Collective intelligence, as an emergent result of human interaction, is closely linked to certain methodologies of citizen science (Lukyanenko et al. 2020). This connection is evident in research areas such as astronomy (Cedazo et al. 2020), rare diseases (Radu et al. 2021), and “wicked problems” (Schoder et al. 2014), among others. In citizen science practices, participants can act as information gatherers, analysts, knowledge creators, or decision-makers (Shirk et al. 2012). Consequently, citizen science is recognized as a valuable approach for addressing sustainability challenges and contributing to the implementation of the SDGs (e.g., Fraisl et al. 2020; Shulla et al. 2020; Turbé et al. 2019). Specifically, in terms of governance, Turbé et al. (2019) demonstrate, through the analysis of 503 European environmental projects, how citizen science can contribute to each step of the policy process.

Collective intelligence, through its openness to decisions traditionally reserved for experts and its interaction with science, shows promise for addressing “grand challenges” like those identified in the SDGs (Elia and Margherita 2018), particularly those related to sustainability, which require transformations in production, relationships, and consumption (Atlee 2017). Engaging in collective debate can help translate abstract concepts such as climate change into practical agreements and transformations of everyday actions, integrating learnings and involving the wider population (Piccolo et al. 2018). Moreover, online interconnectivity allows collective intelligence to expand, connecting hundreds or thousands of individuals in what has been termed “the wisdom of the crowd” (Bigham et al. 2018; Lévy 2002; Surowiecki 2005).

Collective intelligence can be understood as a set of practices for collaborative decision-making or proposal development, as well as a product of group decision-making. These practices can serve as a methodology within citizen science and participatory processes, as they enhance the information utilised in decision-making and situate it within a context and relationship with specific knowledge. By defining collective intelligence as a group’s ability to address various challenges, it becomes a measure of a community’s capacity to implement and address the SDGs (Woolley et al. 2010). Achieving the right balance between group cohesiveness and individual interests, as well as the process’s ability to generate original responses, is crucial for implementing localised and concrete solutions related to the SDGs.

Engagement and execution

The experiment began on December 4, 2019, with the interactive phases occurring on December 10, 2019. It involved 1,253 students aged 15 to 17 from 68 secondary schools across 13 Spanish autonomous regions. A group of students from Madrid participated in the experiment in person at the COP25 venue.

To prepare for the experiment and to reach potential participants, Ibercivis issued an open call. Additionally, secondary schools from all Spanish autonomous regions were contacted, leveraging existing relationships from previous projects with Ibercivis. Interested teachers could register their classrooms using a distributed form, which collected their personal information for communication purposes during the experiment’s organisation. The form contained all relevant information about the current legislation in force in Europe, which is Regulation (EU) 2016/679 (General Data Protection Regulation), and the associated rights. Teachers were required to provide explicit informed consent before submitting their personal information through the form.

Each participating school managed the internal logistics required for the experiment. Schools incorporated the experiment into their activities by obtaining necessary permissions from parents or legal guardians. Participation was voluntary and entirely anonymous, with no collection of students’ personal data.

Experiment description

The experiment begins by outlining a challenge that needs to be collectively resolved. The challenge is presented through the following text:

Table 1 shows the consecutive phases along the experiment. In Phase 1, which began on December 4, students individually brainstormed and presented their five proposals for each SDG. By Phase 2, on December 10, they could view each other’s responses. During Phases 3 and 4, students were able to copy one or all of their peers’ proposals, modify their own, or keep them unchanged. Consequently, the level of interaction within the network increased over time, as transformed proposals emerged from observing previously inaccessible contributions. In Phase 5, the platform identified the 10 most copied proposals (“TOP 10”). Participants then had to choose between these proposals or maintain or modify one of their own. In Phase 6, they could no longer edit their submissions but could only replace them with one from the TOP 10. Lastly, Phase 7 revealed the five most copied entries, that is, the “best solutions,” for each challenge. As seen in these final three phases, a concentration process unfolds, driving the selection of the best proposals.

How results are evaluated

To evaluate the collective intelligence generated in the process, we used three indicators, which can be seen in Table 2.

In this paper, we examine the effectiveness and outcomes of collective intelligence, through the analysis of interaction processes and the study of the balance between group cohesion and individual interest. Our experimental design enables us to assess three key dimensions of collective intelligence within a group in a given context.

First, we analyse the balance between individuality and the herd effect to evaluate the potential for generating innovative solutions to the SDGs. The herd effect is characterised by a high degree of participants following and accepting others’ responses, while strong individuality suggests that participants are unwilling to accept the ideas of others. Second, we investigate the origin of the responses, determining whether they stem from advanced interactions between participants or individual thought processes. Lastly, we explore the extent to which the online method enables the generation of responses that exhibit high levels of agreement and broad dissemination within the network.

Observed interaction processes

Our experiment demonstrates the viability and usefulness of collective intelligence and open decision-making processes in tackling “grand challenges” (Elia and Margherita 2018). Utilising the ThinkHub platform, we observed extensive online collaboration involving numerous participants, which generated collective intelligence dynamics and led to the successful creation and prioritisation of solutions.

However, our results reveal that the interaction among participants was not as high as initially anticipated. There was relative isolation, low sharing frequency, and a reluctance to alter or transform initial responses. Participants’ opinions remained mostly unchanged throughout the experiment’s phases. Additionally, in the case of SDG 13, the development of the final responses did not achieve the desired level of elaboration.

A deeper investigation is required to understand the questions arising during the solution-generation process, and how they are influenced by the specific configuration of the tool and methodology. Additionally, we believe it would be useful to examine the interactions between participants in a collective intelligence process, and identify variables that impact the process and the outcome in terms of individualization, herd effect, the level of collective work in creating final answers, and the degree of agreement.

Observed balance between group cohesion and individual interest

The experimental design itself presents a conflict between enhancing individual contributions and disseminating collective responses. When a participant alters their response to improve it, they create a new idea that must start from scratch and compete with already established proposals. Interestingly, most final answers were generated in the initial phases, showing a certain stubbornness to modify and improve one’s own responses. This outcome allows us to pinpoint two distinct variables necessary for collective intelligence among large groups, further elaborating on Toyokawa et al.’s (2019) distinction between herd and isolated behaviour. These variables include the generation of collective responses through interaction and the diffusion of these responses within groups, thus creating subsets of participants supporting a specific idea (Figure 4).

We can regard these two approaches to idea generation as orthogonal. For instance, when individually conceived responses are spread across the network or when each individual generates their response from interaction, leading to fewer instances of copying but responses created in later phases. However, these variables are interconnected and positively impact one another.

Proposed process improvements

In order to address the challenges faced in the experiment and enhance the collective intelligence methodology for tackling SDG-related issues, we propose the following measures. First, proposals could be streamlined by enforcing copying, periodically removing low-diffusion proposals, and strengthening the idea selection system. Providing participants with information on the number of copies each item has may further incentivize choosing popular ones, promoting centralization. Gamification techniques could boost participation. Additionally, controlling or eliminating “trolls” or users who submit unrelated proposals can prevent distortion of the process.

Second, it would be worth exploring other features to increase interaction during the experiment. Identifying questions and issues that more directly affect participants and even generating positions that force a choice between alternatives could be beneficial.

Improving the process may be achieved by separating the construction of collective solutions from their dissemination and discussion. For instance, an initial stage could involve producing the main answers (e.g., reaching best 5 answers through various dynamics) where answers must be mutually exclusive or at least require preference. In a subsequent stage, participants could choose an answer and provide a short supporting argument, which could then be copied, transformed, or maintained. This approach enables participants to change their position based on arguments, which may or may not spread across the network, fostering genuine discursive poles (Conde 2009). Interactions can be further developed between arguments by refining, merging, or selecting those that fail to convince. Thus, the dynamics would revolve around both answers and arguments, fostering interaction and collective intelligence.

Furthermore, broadening participation offers diverse decision-making processes and facilitates the transfer and approximation of abstract concepts (Piccolo et al. 2018), particularly those related to environmental and climate change issues. Problems and solutions can be made more concrete and connected to everyday actions.