In the early 2000s, colleagues told me about the real-life friendships they’d developed in online games. They told me about the openness and honesty that greeted them when they logged into Everquest and Asheron’s Call, the emotional power of achieving long-term goals with people thousands of miles away, and the bizarre group ‘dinner’ parties they described, separate but together, behind their home computer screens. I dismissed their stories; these were places populated by orcs and warlocks, I reasoned. Impossible fantasies of people who needed to meet more real people. But around that time I also read an article by economist Edward Castronova who had used the economic barter and trade systems in Everquest to assess the GDP of its fictional world. Taking the average amount of time players spent ‘levelling up’, or advancing through new areas of the story by engaging in repetitive tasks, and the amount which items and extremely virtually valuable character accounts were sold for on eBay’s Category 1654 (Computer Games), Castronova placed its Gross Domestic Product between Bulgaria’s and Russia’s. His paper, On Virtual Economies, and his subsequent analysis of the relative and socially-determined monetary ‘value’ of male and female avatars in massively multiplayer online games (women sell for $47 less, based only on gender assignment), inspired a flood of scientists into online spaces like EverQuest searching for similar clues to offline social processes, including me.

VIRTUAL GROUP PSYCHOLOGIES
In my search for virtual group psychology, I discovered that the internet as a platform for social phenomena was realised early; asynchronous chat – including email, listservs and forums – populated the web from as early as 1985, with the first popular synchronous chatting between people in real time emerged in 1988. Anthropologists, already actively documenting the online bulletin board communities, descended onto these chat spaces, producing accounts of web-citizens and their practices in much the same way as they produced these accounts for populations offline. This body of research identified the ways in which interpersonal rules were made, what happened when they were broken, and how formal and informal groups converged and instituted these social norms.

A crux issue in accepting the relationships between online and offline social behaviour was the medium’s apparent lack of consequence; users appeared to have complete anonymity. As in offline environments, anonymity is frequently identified as the culprit in online anti-social behaviour, the transience of relationships, and – because of the deindividuation experienced behind the internet mask – social conformity .

THE ONLINE TIES THAT BIND AND GAG
But as these worlds have matured, it has become apparent that they are not entirely anonymous or without consequence. When bad behaviour results in dismissal or social exclusion, it may seem that an offender can simply return under another pseudonym; in fact, community members are often able to identify patterns of behaviour that make reprobates visible behind their new avatars. And although an online participant is free to create an identity which is apparently ‘anyone’, the real benefit only comes from the persistent surrogate identities developed over time within the social dynamic upon which reputations are built. Just as in the real world, people in these online environments exhibit a desire to fit into groups, from mirroring body movements with avatars to conceding in replications of Asch’s conformity tasks, which helps them to develop trust and social capital between anonymous strangers. Being part of a group, online or off, means that the members are bound by its social norms.

Virtual worlds have matured and grown since the mid-1990s from their text-based origins with populations in their tens of thousands into visually-rich, multimedia spaces with international populations in their millions. The abundance of new media, community-driven applications like video sharing sites YouTube, 3D environments like Second Life, the network of blogs and other social Web 2.0 applications, means that the internet now caters for collective action for a wider potential population across a multitude of communities of practice. These spaces are widely used as forums to collaborate, and are increasingly embedded in our media experiences. As a result, the concept of close relationships – once proximity bound – is changing. There’s evidence that it will continue to evolve: a 2008 survey by kids’ virtual world Habbo Hotel showed that 64% of 6,000 British children who go online report they have friends they have never met offline. Their offline reputations have become part of their online lives, and vice versa.

REAL BEHAVIOUR, ONLINE
As a result, the Internet now offers a unique digital place for real-world social scientists to gather data on human behaviours that have been previously difficult to capture using lab-based empirical methods. There has been a raft of research aimed at identifying the similarities between online and offline, from experiments which replicate the chameleon effect and conformity processes in group situations using precise computer-captured location and orientation data, and innovation adoption and trust behaviours collected by ‘scraping’ the server logs of online community developers. Participants in web-based research use the same cues as they do in offline experiments to determine reference groups, for example, and the effects on decision-making, influence and norm development are apparently identical.

Further, because the field of study is digital, it is possible to test models using live participants in relevant contexts that would otherwise be impossible in the real world. Databases power the virtual environments, collecting everything that happens in an online community, including what participants do and what they say. This provides a rich and unprecedented record of activities and responses captured live for later analysis that neither degrade over time nor are reliant on retrospective accounts. Using this approach, a team of researchers at Tufts University gathered the server logs from the online game World of Warcraft in late 2005, when an outbreak of a virtual ‘bleeding’ virus ran amok through the community. The virus had initially been restricted to a secure area of the virtual world by the game developers, but was passed to the rest of the population when a pet dog – immune to the fatal effects – was summoned by its player-owner outside the area into the mainland. Within days, the fantasy realm of Azeroth was littered with the corpses of avatars unlucky enough to have been in the vicinity of an infected user. Blizzard, the game’s creator, quickly released a correction, returning the world to its former, disease-free state, but the outbreak proved a rich stream of inquiry.

In an unprecedented approach, the researchers studying Azeroth were able to analyse the unexpected and natural human effects of a disease, including host behaviour, which mathematical models are unable to take into account. Now, epidemiological models of infectious epidemics take into consideration these human factors.

Researchers need not wait for something as dramatic – or unanticipated – as a disease to capture real-time behaviour. Sociologists at the University of Surrey in the UK are working in virtual world Second Life to generate models about norm development in new community discussion groups by analysing transcripts of group discussions and physical group behaviour. When I was at the same institution I assessed where social network analysis added to real-world social psychological theories of influence over and above persuasion models by tracking the diffusion of attitudes and behvaviours via relationship connections using data provided by the development house. Legal scholars at New York Law School and Yale Law School have observed online interactions in EverQuest and other game-worlds, and have applied new theory to research outlining policy for new law and governance in developing countries. And recently, the National Science Foundation awarded the University of California at Irvine’s Department of Informatics $100,000 to study the different cultural social dynamics between Chinese and American players of the game.

Castronova has continued to explore real-life applications of the virtual economic principles that arise in online games. In his most recent project, he and a team at the University of Indiana at Bloomington created a Shakespeare-themed virtual world called Arden. The aim was to split it into two identical copies, welcome players to both and to manipulate the economic tenets of trade, barter and asset accumulation in one, keeping the other as control.

NATURAL BEHAVIOUR
Unlike most lab-based research into online behaviour, researchers who use virtual worlds are able to identify naturalistic patterns of emergent behaviour in context. These spaces attract a widely distributed demographic – both in terms of country of residence and other features like age, gender and socioeconomic status – which offers a far more representative population than the undergraduates traditionally relied upon in experimental social science research. However, because of the vast quantity of activity in these spaces, online community data can quickly multiply and researchers can be overcome with the amount of information captured. There are few computers in the world that can cope with such an abundance of riches, and establishing boundaries for data collection is essential when working in this space.

Microcosms of human behaviour on the computer may feel like a science fiction fantasy, but for the hundreds of millions who engage in virtual worlds, online games, social networks, chatrooms, forums and listservs, the virtual is just as real as reality. Like the technology itself, social science research using the web is in its infancy, but peel back the digital curtain and you will discover communities that reflect the real-world so closely – and with such promise – you will be amazed indeed.

This essay was originally written for Nature