The World Wide Web, and online social networks in particular, have increased connectivity between people such that information can spread to millions of people in a matter of minutes. This form of online collective contagion has provided many benefits to society, such as providing reassurance and emergency management in the immediate aftermath of natural disasters. However, it also poses a potential risk to vulnerable Web users who receive this information and could subsequently come to harm. One example of this would be the spread of suicidal ideation in online social networks, about which concerns have been raised. In this paper we report the results of a number of machine classifiers built with the aim of classifying text relating to suicide on Twitter. The classifier distinguishes between the more worrying content, such as suicidal ideation, and other suicide-related topics such as reporting of a suicide, memorial, campaigning and support. It also aims to identify flippant references to suicide. We built a set of baseline classifiers using lexical, structural, emotive and psychological features extracted from Twitter posts. We then improved on the baseline classifiers by building an ensemble classifier using the Rotation Forest algorithm and a Maximum Probability voting classification decision method, based on the outcome of base classifiers. This achieved an F-measure of 0.728 overall (for 7 classes, including suicidal ideation) and 0.69 for the suicidal ideation class. We summarise the results by reflecting on the most significant predictive principle components of the suicidal ideation class to provide insight into the language used on Twitter to express suicidal ideation. Finally, we perform a 12-month case study of suicide-related posts where we further evaluate the classification approach - showing a sustained classification performance and providing anonymous insights into the trends and demographic profile of Twitter users posting content of this type.

HighlightsWe investigate the characteristics of the authors of Tweets containing suicidal intent or thinking, through the analysis of their online social network relationships and interactions.Results show a high degree of reciprocal connectivity between the authors of suicidal content when compared to other studies of Twitter users, suggesting a tightly-coupled virtual community.Analysis of the retweet graph identified bridge nodes and hub nodes connecting users posting suicidal ideation with users who were not, suggesting a potential for information cascade and risk of possible ‘contagion’.Retweet graphs of suicidal content exhibit an average shortest path similar to that of a large comparison network, demonstrating large scale information propagation in small-scale networks.

Cooperation is a fundamental human trait but our understanding of how it functions remains incomplete. Indirect reciprocity is a particular case in point, where one-shot donations are made to unrelated beneficiaries without any guarantee of payback. Existing insights are largely from two independent perspectives: i) individual-level cognitive behaviour in decision making, and ii) identification of conditions that favour evolution of cooperation. We identify a fundamental connection between these two areas by examining social comparison as a means through which indirect reciprocity can evolve. Social comparison is well established as an inherent human disposition through which humans navigate the social world by self-referential evaluation of others. Donating to those that are at least as reputable as oneself emerges as a dominant heuristic, which represents aspirational homophily. This heuristic is found to be implicitly present in the current knowledge of conditions that favour indirect reciprocity. The effective social norms for updating reputation are also observed to support this heuristic. We hypothesise that the cognitive challenge associated with social comparison has contributed to cerebral expansion and the disproportionate human brain size, consistent with the social complexity hypothesis. The findings have relevance for the evolution of autonomous systems that are characterised by one-shot interactions.

Prejudicial attitudes are widely seen between human groups, with significant consequences. Actions taken in light of prejudice result in discrimination, and can contribute to societal division and hostile behaviours. We define a new class of group, the prejudicial group, with membership based on a common prejudicial attitude towards the out-group. It is assumed that prejudice acts as a phenotypic tag, enabling groups to form and identify themselves on this basis. Using computational simulation, we study the evolution of prejudicial groups, where members interact through indirect reciprocity. We observe how cooperation and prejudice coevolve, with cooperation being directed in-group. We also consider the co-evolution of these variables when out-group interaction and global learning are immutable, emulating the possible pluralism of a society. Diversity through three factors is found to be influential, namely out-group interaction, out-group learning and number of sub-populations. Additionally populations with greater in-group interaction promote both cooperation and prejudice, while global rather than local learning promotes cooperation and reduces prejudice. The results also demonstrate that prejudice is not dependent on sophisticated human cognition and is easily manifested in simple agents with limited intelligence, having potential implications for future autonomous systems and human-machine interaction.

We address the problem of cooperation in decentralized systems, specifically looking at interactions between independent pairs of peers where mutual exchange of resources (e.g., updating or sharing content) is required. In the absence of any enforcement mechanism or protocol, there is no incentive for one party to directly reciprocate during a transaction with another. Consequently, for such decentralized systems to function, protocols for self-organization need to explicitly promote cooperation in a manner where adherence to the protocol is incentivized.In this article we introduce a new generic model to achieve this. The model is based on peers repeatedly interacting to build up and maintain a dynamic social network of others that they can trust based on similarity of cooperation. This mechanism effectively incentivizes unselfish behavior, where peers with higher levels of cooperation gain higher payoff. We examine the model's behavior and robustness in detail. This includes the effect of peers self-adapting their cooperation level in response to maximizing their payoff, representing a Nash-equilibrium of the system. The study shows that the formation of a social network based on reflexive cooperation levels can be a highly effective and robust incentive mechanism for autonomous decentralized systems.