Human collective intelligence under dual exploration-exploitation dilemmas

Interaction among participants in a collective intelligence experiment: an emotional approach

Introduction

The construct of collective intelligence assumes that groups have a better capacity than individuals to deal with complex, poorly defined problems. The digital domain allows us to analyze this premise under circumstances different from those in the physical environment: we can gather an elevated number of participants and generate a large quantity of data.

Methods

This study adopted an emotional perspective to analyze the interactions among 794 adolescents dealing with a sexting case on an online interaction platform designed to generate group answers resulting from a certain degree of achieved consensus.

Results

Our results show that emotional responses evolve over time in several phases of interaction. From the onset, the emotional dimension predicts how individual responses will evolve, particularly in the final consensus phase.

Discussion

Responses gradually become more emotionally complex; participants tend to identify themselves with the victim in the test case while increasingly rejecting the aggressors.

Exploring new technologies for the future generation: exploration-exploitation trade-off in an intergenerational framework

Decision making on exploring or exploiting technology was studied by means of a laboratory experiment with a two-generation framework. In this framework, the design of a virtual tool is transmitted from the first to second generation, and hence, the former can help the latter by frequently exploring better tool designs but at the cost of reduced opportunities to exploit the existing tool to increase its own benefits. We set two experimental conditions ('repaid' and 'unrepaid') as well as a control condition (asocial), in which the second generation is absent. In the 'repaid' experimental condition, participants received an extra payment proportional to the score gained by the second generation, such that they were monetarily incentivized to help the second generation. Such an incentive was not given in the 'unrepaid' condition. An analysis of a formal model and computer simulations predicted that rational participants should increase investment in exploration only in the repaid condition when compared with the asocial control. The prediction was confirmed by the results of the experiment. These findings together suggest that humans may not have a propensity to invest in costly exploration of new technologies solely to help future generations.

Efficiency traps beyond the climate crisis: exploration-exploitation trade-offs and rebound effects

Higher levels of economic activity are often accompanied by higher energy use and consumption of natural resources. As fossil fuels still account for 80% of the global energy mix, energy consumption remains closely linked to greenhouse gas (GHG) emissions and thus to climate change. Under the assumption of sufficiently elastic demand, this reality of global economic development based on permanent growth of economic activity, brings into play the Jevons Paradox, which hypothesises that increases in the efficiency of resource use leads to increases in resource consumption. Previous research on the rebound effects has limitations, including a lack of studies on the connection between reinforcement learning and environmental consequences. This paper develops a mathematical model and computer simulator to study the effects of micro-level exploration-exploitation strategies on efficiency, consumption and sustainability, considering different levels of direct and indirect rebound effects. Our model shows how optimal exploration-exploitation strategies for increasing efficiency can lead to unsustainable development patterns if they are not accompanied by demand reduction measures, which are essential for mitigating climate change. Moreover, our paper speaks to the broader issue of efficiency traps by highlighting how indirect rebound effects not only affect primary energy (PE) consumption and GHG emissions, but also resource consumption in other domains. By linking these issues together, our study sheds light on the complexities and interdependencies involved in achieving sustainable development goals. This article is part of the theme issue 'Climate change adaptation needs a science of culture'.

Thoughts from the forest floor: a review of cognition in the slime mould Physarum polycephalum

Sensing, communication, navigation, decision-making, memory and learning are key components in a standard cognitive tool-kit that enhance an animal's ability to successfully survive and reproduce. However, these tools are not only useful for, or accessible to, animals-they evolved long ago in simpler organisms using mechanisms which may be either unique or widely conserved across diverse taxa. In this article, I review the recent research that demonstrates these key cognitive abilities in the plasmodial slime mould Physarum polycephalum, which has emerged as a model for non-animal cognition. I discuss the benefits and limitations of comparisons drawn between neural and non-neural systems, and the implications of common mechanisms across wide taxonomic divisions. I conclude by discussing future avenues of research that will draw the most benefit from a closer integration of Physarum and animal cognition research.

Insights about the common generative rule underlying an information foraging task can be facilitated via collective search

Social learning is beneficial for efficient information search in unfamiliar environments ("within-task" learning). In the real world, however, possible search spaces are often so large that decision makers are incapable of covering all options, even if they pool their information collectively. One strategy to handle such overload is developing generalizable knowledge that extends to multiple related environments ("across-task" learning). However, it is unknown whether and how social information may facilitate such across-task learning. Here, we investigated participants' social learning processes across multiple laboratory foraging sessions in spatially correlated reward landscapes that were generated according to a common rule. The results showed that paired participants were able to improve efficiency in information search across sessions more than solo participants. Computational analysis of participants' choice-behaviors revealed that such improvement across sessions was related to better understanding of the common generative rule. Rule understanding was correlated within a pair, suggesting that social interaction is a key to the improvement of across-task learning.

Conformist social learning leads to self-organised prevention against adverse bias in risky decision making

Given the ubiquity of potentially adverse behavioural bias owing to myopic trial-and-error learning, it seems paradoxical that improvements in decision-making performance through conformist social learning, a process widely considered to be bias amplification, still prevail in animal collective behaviour. Here we show, through model analyses and large-scale interactive behavioural experiments with 585 human subjects, that conformist influence can indeed promote favourable risk taking in repeated experience-based decision making, even though many individuals are systematically biased towards adverse risk aversion. Although strong positive feedback conferred by copying the majority's behaviour could result in unfavourable informational cascades, our differential equation model of collective behavioural dynamics identified a key role for increasing exploration by negative feedback arising when a weak minority influence undermines the inherent behavioural bias. This 'collective behavioural rescue', emerging through coordination of positive and negative feedback, highlights a benefit of collective learning in a broader range of environmental conditions than previously assumed and resolves the ostensible paradox of adaptive collective behavioural flexibility under conformist influences.

Meta-control of social learning strategies

Social learning, copying other’s behavior without actual experience, offers a cost-effective means of knowledge acquisition. However, it raises the fundamental question of which individuals have reliable information: successful individuals versus the majority. The former and the latter are known respectively as success-based and conformist social learning strategies. We show here that while the success-based strategy fully exploits the benign environment of low uncertainly, it fails in uncertain environments. On the other hand, the conformist strategy can effectively mitigate this adverse effect. Based on these findings, we hypothesized that meta-control of individual and social learning strategies provides effective and sample-efficient learning in volatile and uncertain environments. Simulations on a set of environments with various levels of volatility and uncertainty confirmed our hypothesis. The results imply that meta-control of social learning affords agents the leverage to resolve environmental uncertainty with minimal exploration cost, by exploiting others’ learning as an external knowledge base.

Which individuals have reliable information: successful individuals or the majority? Seeking a suitable compromise between individual and social learning is crucial for optimum learning in a population. Motivated by the recent findings in neuroscience showing that the brain can arbitrate between different learning strategies, termed meta-control, we propose a meta-control approach in social learning context. First, we show that environmental uncertainty is a crucial predictor of the performance of the individual and two social learning strategies: success-based and conformist. Our meta social learning model uses environmental uncertainty to find a compromise between these two strategies. In simulations on a set of environments with various levels of volatility and uncertainty, we demonstrate that our model outperforms other meta-social learning approaches. In the subsequent evolutionary analysis, we show that our model dominated others in survival rate. Our work provides a new account of the trade-offs between individual, success-based, and conformist learning strategies in multi-agent settings. Critically, our work unveils an optimal social learning strategy to resolve environmental uncertainty with minimal exploration cost.

A Social Interpolation Model of Group Problem-Solving

How do people use information from others to solve complex problems? Prior work has addressed this question by placing people in social learning situations where the problems they were asked to solve required varying degrees of exploration. This past work uncovered important interactions between groups' connectivity and the problem's complexity: the advantage of less connected networks over more connected networks increased as exploration was increasingly required for optimally solving the problem at hand. We propose the Social Interpolation Model (SIM), an agent-based model to explore the cognitive mechanisms that can underlie exploratory behavior in groups. Through results from simulation experiments, we conclude that "exploration" may not be a single cognitive property, but rather the emergent result of three distinct behavioral and cognitive mechanisms, namely, (a) breadth of generalization, (b) quality of prior expectation, and (c) relative valuation of self-obtained information. We formalize these mechanisms in the SIM, and explore their effects on group dynamics and success at solving different kinds of problems. Our main finding is that broad generalization and high quality of prior expectation facilitate successful search in environments where exploration is important, and hinder successful search in environments where exploitation alone is sufficient.

Characterization and Comparison of the Utilization of Facebook Groups Between Public Medical Professionals and Technical Communities to Facilitate Idea Sharing and Crowdsourcing During the COVID-19 Pandemic: Cross-sectional Observational Study

BACKGROUND

Strict social distancing measures owing to the COVID-19 pandemic have led people to rely more heavily on social media, such as Facebook groups, as a means of communication and information sharing. Multiple Facebook groups have been formed by medical professionals, laypeople, and engineering or technical groups to discuss current issues and possible solutions to the current medical crisis.

OBJECTIVE

This study aimed to characterize Facebook groups formed by laypersons, medical professionals, and technical professionals, with specific focus on information dissemination and requests for crowdsourcing.

METHODS

Facebook was queried for user-created groups with the keywords "COVID," "Coronavirus," and "SARS-CoV-2" at a single time point on March 31, 2020. The characteristics of each group were recorded, including language, privacy settings, security requirements to attain membership, and membership type. For each membership type, the group with the greatest number of members was selected, and in each of these groups, the top 100 posts were identified using Facebook's algorithm. Each post was categorized and characterized (evidence-based, crowd-sourced, and whether the poster self-identified). STATA (version 13 SE, Stata Corp) was used for statistical analysis.

RESULTS

Our search yielded 257 COVID-19-related Facebook groups. Majority of the groups (n=229, 89%) were for laypersons, 26 (10%) were for medical professionals, and only 2 (1%) were for technical professionals. The number of members was significantly greater in medical groups (21,215, SD 35,040) than in layperson groups (7623, SD 19,480) (P<.01). Medical groups were significantly more likely to require security checks to attain membership (81% vs 43%; P<.001) and less likely to be public (3 vs 123; P<.001) than layperson groups. Medical groups had the highest user engagement, averaging 502 (SD 633) reactions (P<.01) and 224 (SD 311) comments (P<.01) per post. Medical professionals were more likely to use the Facebook groups for education and information sharing, including academic posts (P<.001), idea sharing (P=.003), resource sharing (P=.02) and professional opinions (P<.001), and requesting for crowdsourcing (P=.003). Layperson groups were more likely to share news (P<.001), humor and motivation (P<.001), and layperson opinions (P<.001). There was no significant difference in the number of evidence-based posts among the groups (P=.10).

CONCLUSIONS

Medical professionals utilize Facebook groups as a forum to facilitate collective intelligence (CI) and are more likely to use Facebook groups for education and information sharing, including academic posts, idea sharing, resource sharing, and professional opinions, which highlights the power of social media to facilitate CI across geographic distances. Layperson groups were more likely to share news, humor, and motivation, which suggests the utilization of Facebook groups to provide comedic relief as a coping mechanism. Further investigations are necessary to study Facebook groups' roles in facilitating CI, crowdsourcing, education, and community-building.

Bayesian collective learning emerges from heuristic social learning

Researchers across cognitive science, economics, and evolutionary biology have studied the ubiquitous phenomenon of social learning-the use of information about other people's decisions to make your own. Decision-making with the benefit of the accumulated knowledge of a community can result in superior decisions compared to what people can achieve alone. However, groups of people face two coupled challenges in accumulating knowledge to make good decisions: (1) aggregating information and (2) addressing an informational public goods problem known as the exploration-exploitation dilemma. Here, we show how a Bayesian social sampling model can in principle simultaneously optimally aggregate information and nearly optimally solve the exploration-exploitation dilemma. The key idea we explore is that Bayesian rationality at the level of a population can be implemented through a more simplistic heuristic social learning mechanism at the individual level. This simple individual-level behavioral rule in the context of a group of decision-makers functions as a distributed algorithm that tracks a Bayesian posterior in population-level statistics. We test this model using a large-scale dataset from an online financial trading platform.

Haptic communication optimises joint decisions and affords implicit confidence sharing

Group decisions can outperform the choices of the best individual group members. Previous research suggested that optimal group decisions require individuals to communicate explicitly (e.g., verbally) their confidence levels. Our study addresses the untested hypothesis that implicit communication using a sensorimotor channel—haptic coupling—may afford optimal group decisions, too. We report that haptically coupled dyads solve a perceptual discrimination task more accurately than their best individual members; and five times faster than dyads using explicit communication. Furthermore, our computational analyses indicate that the haptic channel affords implicit confidence sharing. We found that dyads take leadership over the choice and communicate their confidence in it by modulating both the timing and the force of their movements. Our findings may pave the way to negotiation technologies using fast sensorimotor communication to solve problems in groups.

Collective intelligence in medical decision-making: a systematic scoping review

BACKGROUND

Collective intelligence, facilitated by information technology or manual techniques, refers to the collective insight of groups working on a task and has the potential to generate more accurate information or decisions than individuals can make alone. This concept is gaining traction in healthcare and has potential in enhancing diagnostic accuracy. We aim to characterize the current state of research with respect to collective intelligence in medical decision-making and describe a framework for diverse studies in this topic.

METHODS

For this systematic scoping review, we conducted a systematic search for published literature using PubMed, Embase, Web of Science, and CINAHL on August 8, 2017. We included studies that combined the insights of two or more medical experts to make decisions related to patient care. Studies that examined medical decisions such as diagnosis, treatment, and management in the context of an actual or theoretical patient case were included. We include studies of complex medical decision-making rather than identification of a visual finding, as in radiology or pathology. We differentiate between medical decisions, in which synthesis of multiple types of information is required over time, and studies of radiological scans or pathological specimens, in which objective identification of a visual finding is performed. Two reviewers performed article screening, data extraction, and final inclusion for analysis.

RESULTS

Of 3303 original articles, 15 were included. Each study examined the medical decisions of two or more individuals; however, studies were heterogeneous in their methods and outcomes. We present a framework to characterize these diverse studies, and future investigations, based on how they operationalize collective intelligence for medical decision-making: 1) how the initial decision task was completed (group vs. individual), 2) how opinions were synthesized (information technology vs. manual vs. in-person), and 3) the availability of collective intelligence to participants.

DISCUSSION

Collective intelligence in medical decision-making is gaining popularity to advance medical decision-making and holds promise to improve patient outcomes. However, heterogeneous methods and outcomes make it difficult to assess the utility of collective intelligence approaches across settings and studies. A better understanding of collective intelligence and its applications to medicine may improve medical decision-making.

The impact of the 'open' workspace on human collaboration

Organizations’ pursuit of increased workplace collaboration has led managers to transform traditional office spaces into ‘open’, transparency-enhancing architectures with fewer walls, doors and other spatial boundaries, yet there is scant direct empirical research on how human interaction patterns change as a result of these architectural changes. In two intervention-based field studies of corporate headquarters transitioning to more open office spaces, we empirically examined—using digital data from advanced wearable devices and from electronic communication servers—the effect of open office architectures on employees' face-to-face, email and instant messaging (IM) interaction patterns. Contrary to common belief, the volume of face-to-face interaction decreased significantly (approx. 70%) in both cases, with an associated increase in electronic interaction. In short, rather than prompting increasingly vibrant face-to-face collaboration, open architecture appeared to trigger a natural human response to socially withdraw from officemates and interact instead over email and IM. This is the first study to empirically measure both face-to-face and electronic interaction before and after the adoption of open office architecture. The results inform our understanding of the impact on human behaviour of workspaces that trend towards fewer spatial boundaries.

This article is part of the theme issue ‘Interdisciplinary approaches for uncovering the impacts of architecture on collective behaviour’.

The Expression and Transfer of Valence Associated with Social Conformity

Consensus seeking – abandoning one’s own judgment to align with a group majority – is a fundamental feature of human social interaction. Notably, such striving for majority affiliation often occurs in the absence of any apparent economic or social gain, suggesting that achieving consensus might have intrinsic value. Here, using a simple gambling task, in which the decisions of ostensible previous gamblers were indicated below available options on each trial, we examined the affective properties of agreeing with a group majority by assessing the trade-off between social and non-social currencies, and the transfer of social valence to concomitant stimuli. In spite of demonstrating near-perfect knowledge of objective reward probabilities, participant’s choices and evaluative judgments reflected a reliable preference for conformity, consistent with the hypothesized value of social alignment.

Social learning strategies regulate the wisdom and madness of interactive crowds

Why groups of individuals sometimes exhibit collective 'wisdom' and other times maladaptive 'herding' is an enduring conundrum. Here we show that this apparent conflict is regulated by the social learning strategies deployed. We examined the patterns of human social learning through an interactive online experiment with 699 participants, varying both task uncertainty and group size, then used hierarchical Bayesian model fitting to identify the individual learning strategies exhibited by participants. Challenging tasks elicit greater conformity among individuals, with rates of copying increasing with group size, leading to high probabilities of herding among large groups confronted with uncertainty. Conversely, the reduced social learning of small groups, and the greater probability that social information would be accurate for less-challenging tasks, generated 'wisdom of the crowd' effects in other circumstances. Our model-based approach provides evidence that the likelihood of collective intelligence versus herding can be predicted, resolving a long-standing puzzle in the literature.

Adolescents show collective intelligence which can be driven by a geometric mean rule of thumb

How effective groups are in making decisions is a long-standing question in studying human and animal behaviour. Despite the limited social and cognitive abilities of younger people, skills which are often required for collective intelligence, studies of group performance have been limited to adults. Using a simple task of estimating the number of sweets in jars, we show in two experiments that adolescents at least as young as 11 years old improve their estimation accuracy after a period of group discussion, demonstrating collective intelligence. Although this effect was robust to the overall distribution of initial estimates, when the task generated positively skewed estimates, the geometric mean of initial estimates gave the best fit to the data compared to other tested aggregation rules. A geometric mean heuristic in consensus decision making is also likely to apply to adults, as it provides a robust and well-performing rule for aggregating different opinions. The geometric mean rule is likely to be based on an intuitive logarithmic-like number representation, and our study suggests that this mental number scaling may be beneficial in collective decisions.

Decision-making without a brain: how an amoeboid organism solves the two-armed bandit

Several recent studies hint at shared patterns in decision-making between taxonomically distant organisms, yet few studies demonstrate and dissect mechanisms of decision-making in simpler organisms. We examine decision-making in the unicellular slime mould Physarum polycephalum using a classical decision problem adapted from human and animal decision-making studies: the two-armed bandit problem. This problem has previously only been used to study organisms with brains, yet here we demonstrate that a brainless unicellular organism compares the relative qualities of multiple options, integrates over repeated samplings to perform well in random environments, and combines information on reward frequency and magnitude in order to make correct and adaptive decisions. We extend our inquiry by using Bayesian model selection to determine the most likely algorithm used by the cell when making decisions. We deduce that this algorithm centres around a tendency to exploit environments in proportion to their reward experienced through past sampling. The algorithm is intermediate in computational complexity between simple, reactionary heuristics and calculation-intensive optimal performance algorithms, yet it has very good relative performance. Our study provides insight into ancestral mechanisms of decision-making and suggests that fundamental principles of decision-making, information processing and even cognition are shared among diverse biological systems.

Group decision-making from an evolutionary/adaptationist perspective

Over the 20 years that Group Processes & Intergroup Relations has been in existence, evolutionary theory has begun to play a larger role in our understanding of human social behavior. Theory and research on group decision-making is no exception and the present paper attempts to briefly highlight how an evolutionary/adaptationist perspective has informed our understanding of how groups reach consensus and make collective choices. In addition, we attempt to show that humans are not the only species that use group processes to make important choices. Looking for similarities and continuities among research domains with different species should lead to a more unified and informed understanding of group decision-making processes and outcomes.

Nash Equilibrium of Social-Learning Agents in a Restless Multiarmed Bandit Game

We study a simple model for social-learning agents in a restless multiarmed bandit (rMAB). The bandit has one good arm that changes to a bad one with a certain probability. Each agent stochastically selects one of the two methods, random search (individual learning) or copying information from other agents (social learning), using which he/she seeks the good arm. Fitness of an agent is the probability to know the good arm in the steady state of the agent system. In this model, we explicitly construct the unique Nash equilibrium state and show that the corresponding strategy for each agent is an evolutionarily stable strategy (ESS) in the sense of Thomas. It is shown that the fitness of an agent with ESS is superior to that of an asocial learner when the success probability of social learning is greater than a threshold determined from the probability of success of individual learning, the probability of change of state of the rMAB, and the number of agents. The ESS Nash equilibrium is a solution to Rogers’ paradox.

Phase transition of social learning collectives and the echo chamber

We study a simple model for social learning agents in a restless multiarmed bandit. There are N agents, and the bandit has M good arms that change to bad with the probability q_{c}/N. If the agents do not know a good arm, they look for it by a random search (with the success probability q_{I}) or copy the information of other agents' good arms (with the success probability q_{O}) with probabilities 1-p or p, respectively. The distribution of the agents in M good arms obeys the Yule distribution with the power-law exponent 1+γ in the limit N,M→∞, and γ=1+(1-p)q_{I}/pq_{O}. The system shows a phase transition at p_{c}=q_{I}/q_{I}+q_{o}. For p<p_{c}(>p_{c}), the variance of N_{1} per agent is finite (diverges as ∝N^{2-γ} with N). There is a threshold value N_{s} for the system size that scales as lnN_{s}∝1/(γ-1). The expected value of the number of the agents with a good arm N_{1} increases with p for N>N_{s}. For p>p_{c} and N<N_{s}, all agents tend to share only one good arm. If the shared arm changes to be bad, it takes a long time for the agents to find another good one. E(N_{1}) decreases to zero as p→1, which is referred to as the "echo chamber."

A model for gossip-mediated evolution of altruism with various types of false information by speakers and assessment by listeners

Indirect reciprocity is considered to be important for explaining altruism among humans. The evolution of altruism has been modeled using several types of reputational scores, most of which were assumed to be updated immediately after each game session. In this study, we introduce gossip sessions held between game sessions to capture the spread of reputation and examine the effects of false information intentionally introduced by some players. Analytical and individual-based simulation results indicated that the frequent exchange of gossip favored the evolution of altruism when no players started false information. In contrast, intermediate repetitions of gossip sessions were favored when the population included liars or biased gossipers. In addition, we found that a gossip listener's strategy of incorporating any gossip regardless of speakers usually worked better than an alternative strategy of not believing gossip from untrustworthy players.