Fri, May 21, 2021

Games with serious impacts: The next generation of serious games

GeoTech Cues by Richard J. Cordes

Technology & Innovation

When most people think of gaming, entertainment is the first thing to come to mind, but it may be possible to leverage that entertainment for research, improving artificial intelligence, and simulating everything from pandemic and food security scenarios to regulatory impacts. By 2012, some estimates suggested that humans had spent over 50 billion hours, or roughly 6 million years, playing World of Warcraft, a long-running massively multiplayer online (MMO) game. As high as that number is, it’s important to note that World of Warcraft is just one of many MMOs, and MMOs are just one of many categories of game. With most gamers averaging six hours a week of playtime, which likely does not reflect COVID-19 pandemic binge-gaming, and the growing blur of what constitutes a “gamer” due to the sustained popularity of mobile games like Candy Crush and Angry Birds, the amount of time spent gaming annually is now unfathomably large. Many have seen this as an unprecedented waste of time and effort, but as Buckminster Fuller once said: “[Waste] is nothing but resources we’re not harvesting—we allow them to disperse because we’ve been ignorant of their value.” In this case, the resources dispersed are human attention and potential.

Not all have been so dismissive of the potential value in time spent gaming. The potential for positive impacts in terms of mental health, problem solving, and social skills, as well as the cultivation of helping behaviors such as sharing and facilitating, has garnered attention in popular media, but others are going further, seeing it not as time wasted but as collective effort waiting to be tapped. For those just now considering what value could be extracted from that collective effort, such as continuing professional development, education, research, and simulation, a good first step is acknowledging that the idea isn’t new. Looking to the history of this concept and recent successes in implementing it will help leverage new opportunities and prevent the repetition of past mistakes.

The beginning of serious games

The use of games for purposes other than entertainment is arguably as old as chess and the intentional design of such games begins with Kriegsspiel, a wargame designed by the nineteenth-century Prussian Army. However, the recognition of the value of the aforementioned effort untapped begins in the early twentieth century with John von Neumann, a Hungarian mathematician and physicist famous for his work on the Manhattan Project and his contributions to (and his founding of) a wide variety of fields. His “Theory of Parlor Games,” presented in a paper published in German in 1928, looked at the nature of “statistical hazards” in social games and how the clear, formal rule sets of games allow for bounding an environment such that the utility of certain decisions and the probability of outcomes can be measured, thereby allowing better understanding, measurement, and observability of behavior. He began with roulette-style games and worked up to more complex ones, suggesting, subtly, that the hazards of the game and the behavioral responses available to players “reflect the essence” of the real world. Neumann later worked with Oskar Morgenstern, an economist, to produce the seminal text “Theory of Games and Economic Behavior.” The work essentially founded the field of “game theory” by building on both Neumann’s previous work and the work of other mathematicians, such as Emile Borel, who had also seen the value of the bounded environments within games for analyzing behavior and probability.

Since that time, the tools and insights of game theory have been thoroughly interwoven in mathematics and science but are often removed from the analysis of actual “games’ as they would be defined in common parlance. Game theory focuses primarily on abstract games, with simple ones typically represented as matrices and more complicated ones being structured as models. These representations help researchers better understand behavior, forecast outcomes, and design mechanisms, such as auction rules, for engineering outcomes. The study and use of actual games have instead been picked up by a relatively new field known as “serious games.” The focus of serious games is the use and study of game-like systems and actual games to engineer and understand outcomes.

A full accounting of the field’s beginnings would be difficult to describe in detail here, but its transition from obscurity began in the 1980s with an observation by management consultant Chuck Coonradt. Coonradt found that there is painstaking effort in keeping professionals working in refrigerated warehouses, but an increase in absenteeism in days with good skiing weather led him to note that under the right conditions people would not only expend more effort than they would at work but that they would pay for the privilege to do so.  

“People will pay for the privilege of working harder than they will work when they are paid”

The Game of Work, Charles Coonradt

Coonradt has sometimes been referred to as being the “grandfather of gamification,” a result of his book “The Game of Work.” Coonradt asserted that while games and sports require work and effort, they enjoy the status of being enjoyable pursuits for six key reasons:

  1. Clear goals: The objectives of the work are clear and well scoped, making navigation toward those goals manageable.
  2. Scorekeeping: The measurement of performance outcomes are clear, comparable, and unambiguous.
  3. Feedback: Given the clarity of objectives and performance outcomes, individuals participating in a game or gamified system have reasonable basis to consider the impact of certain behaviors on results.
  4. Choice: Games and game mechanisms provide players with choices, some clearer than others—the clearer the choices, the more valuable feedback becomes, and the more opportunities are provided for players to invest in understanding the impacts of their choices on outcomes and in innovating or adapting those choices.
  5. Field of play: The time and space in which the game is played are well scoped, so players have clear expectations entering this scope: they know what to expect, what is expected of them, and that the game will eventually end, and therefore that they will have time to rest if they exert themselves.
  6. Skin in the game: This concept from game theory was communicated to a much wider audience in the book of the same name by Nassim Taleb—that players need to acknowledge some value on the table, some potential cost or gain at stake that is tied to their performance in order to play effectively and fairly.

Coonradt argued that these principles could be applied when designing mechanisms for work-flows and processes, in order to achieve desired outcomes. Today many might refer to this idea as the “gamification” of work. Coonradt’s principles have since been tested and implemented in a wide variety of domains. An enormous amount of work has been done since the 1980s on both adapting work flows for gamification, and developing games to facilitate continuing professional development and education. This being said, many suspect that the value of gamifying work has been overhyped. Much of this critique is fair, sometimes targeting the nature of the concept but usually focusing on the nature of implementation and the unreliability of setting out to design engaging experiences. Designing games that are engaging is difficult enough when the center of gravity for design is player engagement, let alone when it is learning chemistry or debugging an operating system. As would be expected from the trajectory of hype in any particular concept, serious games has fallen from what the Gartner Hype Cycle defines as the “peak of inflated expectations” and is on the road toward more reasonable and practical applications.

A part of this progress has been acknowledging the “blood and bones in the path”— designing engaging experiences is difficult and unreliable— and that the driving questions need to change focus from designing games to engineer outcomes to adapting and adopting games to engineer outcomes. Moving from “design from scratch” to “adapt and adopt” has ushered in a new era of serious games, and this shift in prioritization began with an observation: “People are already playing these games for billions of hours, can we harness that work for something useful? Can we adapt the games for impact?”

MMOS and Eve Online

The speculation that games could be adapted and adopted rather than designed from scratch to engineer outcomes had been considered before, but when Attila Szantner, a co-founder of the pre-Facebook, Hungarian social media giant iWiW, considered the billions of hours being spent gaming in 2015, he decided to take action. Attila and his partner, Bernard Revazs, converted theory into practice by founding MMOS (Massively Multiplayer Online Science), which set out to apply the adapt-and-adopt approach for citizen science initiatives. In discussing citizen science and the perspective of MMOS on serious game design for this article, Attila demonstrated a grounded and realistic approach. He noted being inspired by previous work in designing serious games for research and education, particularly Zooniverse and other forms of “people-powered research.” Attila’s insights on the topic cautioned against the perception that the work of MMOS was entirely revolutionary—he claimed that there is no shortage of games designed for research and education and that their impacts should not be overlooked. The key difference between MMOS and these past endeavors however was that they avoided constructing any game from scratch, instead focusing on refining a recipe for turning players of existing games into “virtually limitless human computation engines for citizen science.” The recipe, or at least the outline for it, goes something like this:

  1. Find games that players already love;
  2. Find large-volume, modular research tasks (such as image and pattern classification);
  3. Map the tasks to potential adaptations in-game that facilitate the crowdsourcing of those research tasks;
  4. “Connect the dots with the in-game lore” to make the adaptations to the game “aesthetically fitting and thematically adoptable;” and
  5. Inform the players that by playing the game they already love, they’re contributing to science and making an impact.

Attila wasn’t shy in stating that there was another factor in making this recipe work: developing a relationship with CCP, an Icelandic game company that runs EVE Online. In fact, the recipe itself was developed in part by EVE Online’s game design director at the time, Petur Thorarinsson. No short description of EVE can do it justice, but, in short, it is a space-focused MMO game in which players pilot and manage ships in a vast set of star systems known as “New Eden,” competing and cooperating in their attempts to control territory and earn money in a myriad of ways. EVE is one of the longest running MMOs of all time and is innovative for a number of reasons, but of interest here was CCP’s willingness to commit to and experiment with academic and scientific collaborations, evidenced by the fact that they employ an in-house economist.

The first collaboration with MMOS, championed by EVE Online’s former executive producer, Andie Nordgren,  produced “Project Discovery,” a minigame within the game that connected players with tasks assisting researchers at the KTH Royal Institute of Technology in Sweden with the Human Protein Atlas project, helping to identify subcellular localization of proteins in cells. The task chosen was an image classification activity traditionally delegated to researchers with specialized training. The reigning consensus at the time was that only large-scale machine learning algorithms could handle parsing the millions of images that needed to be classified. Given expectations calibrated by past projects, when Emma Lundberg, the KTH researcher leading the Human Protein Atlas Project, first engaged in the collaboration, among her team’s higher estimates was forty thousand classifications a day. Very early into the launch, they saw as high as nine hundred thousand classifications per day, far exceeding expectations. When critics suggested that machine learning solutions could outperform EVE’s citizen scientists, they weren’t necessarily incorrect—Loc-CAT, a state-of-the-art tool for protein classification tasks did outperform players on common classes of proteins. However, the aggregate data from EVE players outperformed Loc-CAT in identifying rare cases and novel patterns. The aggregate data was integrated into the machine learning model using “transfer learning,” and this injection of human insight boosted the performance of Loc-CAT “significantly.”

The project generated tens of millions of image classifications. These were true milestones for the medium that would soon be rivaled by future collaborations between MMOS and EVE Online. In the next iteration of Project Discovery in 2017, EVE players were tasked with the analysis of astronomical data from the CoRoT telescope in collaboration with Reykjavik University and the University of Geneva. The project itself was championed by none other than Professor Michel Mayor, the famed co-discoverer of the first exoplanet, 51 Pegasi b. Similar to Lundberg, at the start of the project Mayor seemed to temper his expectations, suggesting that the opportunity for outreach and participation was in and of itself a reasonable goal. MMOS did so as well, noting that they did not expect to achieve the same level of success as the previous iteration. However, by week two they had achieved a peak classification rate of 1,539 classifications per minute and a total of 13.2 million classifications, breaking all previous daily records. By the project’s end, 279 million classifications had been performed by 422,000 players. Mayor, speaking about Project Discovery after attending EVE Online’s yearly “Fanfest” conference, stated, “I discovered a new world twenty years ago with a telescope and another one this year when I learned about EVE Online.”

Project Discovery is currently in its third iteration, this time tackling COVID-19 in collaboration with the University of Modena and Reggio Emilia, McGill University, and the University of British Columbia. A statement from Dr. Ryan Brinkman, a professor in medical genetics at the University of British Columbia reads:

“This project is crashing through all my expectations, with players continuing to show great engagement and interest in the work we are doing, as well as providing huge amounts of high-quality data for our research. Their efforts will not only contribute to the understanding of COVID-19, but the data they are generating will also be freely and widely shared with the entire scientific community. There is very high interest in re-using their results for the generation of machine learning algorithms. There is simply no other resource out there for this anywhere close to what is now being generated.”

In response to a request for an update on the undertaking, the Project Discovery team indicated that over 82.2 work-years have been spent by 327,000 active players thus far, with a running total of 119 million classifications.

MMOS is now bringing the lessons of its success with EVE to other franchises, such as Borderlands 3, a first-person shooter developed by the company Gearbox. In a collaboration among Gearbox, McGill University, and the UC San Diego Microsetta Initiative’s “American Gut Project,” the game’s millions of players now have the opportunity to help map the human gut microbiome, which is instrumental in developing a wide variety of medical treatments. That study, too, has vastly exceeded expectations. Even the humble Attila was proud to note that they had performed in a single day five times the analysis that some traditional web-based citizen science projects had done in ten years. The success in a mainstream first-person shooter was a key test for the approach, indicating a reliability in implementation that allows risk-averse organizations to consider adoption and ambitious ones to experiment further.

Social complexity in virtual worlds

Looking further into EVE Online begs a question: what else can the game be adapted to produce? One only needs to take a single, real look at EVE’s player base and complex social organizations to consider its potential for continuing professional development, large scale social simulation, and wargaming in complex environments. EVE draws a very interesting global crowd: it is infested with policy wonks, scientists, military and intelligence professionals, members of the foreign service, politicians and lobbyists, traders and fund managers, and other working professionals, some of whom have publicly admitted to using the game to hone skills they apply in the real world. For example, a significant portion of players are information professionals, and EVE’s provision of a robust API for developers to build on gave them an opportunity to practice their craft by providing a wide variety of complex tools for other players. It is a game often played by people with serious jobs, and their impacts as professionals are seen in the presence of organizational wikis, thousands-strong organizations with formalized roles and workflows, functioning intelligence networks, mercenary organizations, logistics companies, months-long military campaigns, and complex financial instruments. When asked about EVE’s dark, dystopian world of political and financial intrigue, and its tendency to attract hedge fund managers, Russian tycoons, and working professionals, the company’s CEO, Hilmar Veigar Pétursson, simply asked “Have you looked outside?”—a response that closely parallels von Neumann’s aforementioned observation that games often reflect the essence of the real world.

This set of EVE players does not necessarily bring the seriousness of their jobs to the game, but that doesn’t mean they don’t make serious impacts. For example, EVE’s “Space Pope,” a player who once influenced many thousands of players to engage in a religious crusade, is actually a thirty-year veteran of NASA’s Jet Propulsion Laboratory. EVE’s Council of Stellar Management, an elected board that represents players to the developers of the game, has amongst its members a Republican politician and maritime law lobbyist. While running for election he noted, “If you replace ‘government’ with CCP, ‘union members’ with the player base, and ‘country’ with the game world, I’m already basically a [member of] CSM. It’s literally my day job.” A fact well known to the EVE community, but perhaps not to those outside of it, is that Sean Smith, a Foreign Service Information Management Officer and Airforce Veteran among the casualties of the 2012 attack on the US Consulate in Benghazi, was a key figure within EVE. Known within EVE as “Vile Rat”, Sean is memorialized as having shaped the geopolitical environment there to such an extent that even Hillary Clinton noted his impact in the game. Before shirking off what it means for these individuals to have had an impact on the geopolitical environment of a virtual world, it should be noted that whole books have been written on the histories of conflicts within EVE Online, and certain assets in-game can be worth tens of thousands of dollars and take months to build.

In response to a request for more information, EVE’s Senior Strategist Tryggvi Hjaltason, a former intelligence analyst who was originally hired to focus on monitoring financial activity in the game, discussed how EVE generated such a complex economic and political environment. Tryggvi, in reference to EVE’s reliability in generating social complexity, lovingly referred to EVE as the “friendship machine” and laid out the framework for the mechanisms that drive it:

  1. Agency: The game does not restrict identity in the way other MMOs do—EVE simply provides for trade-offs in skill-choices and has some formal roles for participation within in-game corporations. The rest is up to the individuals. It is up to the player to define their identity. This self-definition means a wider variety of choices and strong consideration of the impact of their choices in game.
  2. Loss is Real: This concept maps perfectly to the aforementioned “skin in the game” principle. Unlike other games famous for high-priced game assets such as Second-Life, high value items in EVE can be destroyed in conflict, and players do not always have a choice in whether or not conflict happens. A player could spend many months and thousands of dollars of in-game currency building something valuable just to have it destroyed or taken from them hours later. The cost of mega-projects and the inescapable danger of loss creates a center of gravity for cooperation, contractual relationships, and the development of trust. These scale from small groups of individuals protecting each other from emergent mutual threats all the way to thousands of individuals participating in large-scale conflicts to defend trading territories. Tryggvi noted that war bonds have even been issued to fund some of these conflicts—such instruments in-game rely entirely on institutional trust.
  3. Active Facilitation. An insightful factor rarely found in analyses of social complexity in games, but have found in analyses in successful communities of practice such as Complexity Weekend, is the presence of Facilitators, whom Tryggvi referred to as “helpers,” or participants in the system of interest who are incentivized by having a positive impact on others. Tryggvi believes that EVE attracts these helpers because of the aforementioned agency and potential for loss. The ability to construct a unique identity means that one can manifest an identity as a helper, and the impact of their actions in the construction of that identity is proportionate to the potential for catastrophe that EVE allows for.

It is remarkable that these factors mapped so well to descriptive, but difficult to test, models from organizational research and cybernetics. When asked whether EVE Online would make use of this “friendship machine” for research beyond classification tasks, for contributing to organizational and personality research for example, Tryggvi noted that it was of interest and that there were parties who might be interested in collaborating on some of these fronts, but he also noted caution would be needed as CCP strictly abides by EU data privacy regulations. Given the success of the numerous iterations on Project Discovery, it’s likely that they could actually translate those interests and ambitions into results.

The future of serious games

Given the success of the adapt-and-adopt model in applying serious game approaches to citizen science and the social and political complexity EVE has shown is achievable in virtual worlds, some ideas for serious games that had previously been cast aside for being too ambitious should be reconsidered. Perhaps the greater challenge was for a group of researchers studying social and political complexity to build a game as engaging as EVE, not whether or not that level of complexity could be achieved in a game. There are a series of potential applications of serious games that can now be resurrected:

  1. Wargaming in the gray zone: Using virtual worlds as a basis for conducting wargames isn’t new, but using them for gray zone warfare is a challenging proposition. Simulating society, geopolitical environments, or creating a synthetic internet are monumentally difficult endeavors and potentially not achievable depending on what expectations are present. Perhaps games like EVE could be adapted or adopted to allow for exploration, experimentation, and education through wargaming. Wargames building on games that already have complex markets and information systems might be uniquely suited to prepare participants for the kind of campaigns seen in the modern operating environment.
  2. Pandemic research: EVE and MMOS are already making an impact here, but there’s always more that could be done. Research on contact tracing, policy impacts on trade, and network impacts of narrative influence on behavior could all be potential research initiatives of interest. In particular, epidemiological models could be greatly enhanced with data from these kinds of games, evidenced by the fact that World of Warcraft once served as an accidental testbed for pandemic research after an in-game illness escaped its intended environment.
  3. Next-generation simulation: There is an ongoing discussion about the impact of focusing on “toy problems” and failing to communicate the limitations of models in a variety of fields, especially economics and artificial intelligence. Further, the simulation mechanisms intended to generate data “closer to the real thing” come with their own limitations. When trying to study human behavior, nothing is better than actual human behavior. Virtual worlds could provide a sandbox for research of a variety of complex social and organizational phenomena without the ethical dilemmas posed by research projects like Facebook’s psychological experiments on its users. Further, given the proof-of-concept impact EVE player-data had on its AI-counterparts, there is an opportunity for researchers using agent-based models to study phenomena like foraging and information dynamics to both see their ideas tested and to collect data to better inform future models.
  4. Research beyond classification tasks: There is a paradox in researching organizational behavior: the constrained environments of the lab and the timescales used means better data, but on behavior that might not reflect the real world. On the other hand, the unconstrained environments of the real world mean limited or illusory data, and the generated insights can take many years to reduce to mechanisms that are (or are not) reproducible in the lab. Virtual worlds allow for both large-timescale experimentation and more iterations of short-timescale experimentation in spaces that can provide participants with an enormous amount of flexibility in their choices while still ensuring data collection can be robust and accurate.
  5. Non-scientific annotation and knowledge-management problems: There are plenty of places outside the context of traditional scientific research where annotation and classification data is needed: the creation of nuanced training data-sets for specific machine learning use-cases and filling in the gaps on citation data, for example. While these needs have been approached in the past using other crowdsourcing methods, applying a serious games approach would greatly lower costs and increase the rate of completing these tasks.
  6. Regulatory sandboxing: Regulation and legislation are especially difficult to plan and implement as societies become increasingly complex, jurisdictions become increasingly interconnected, and technology rapidly changes. A recent solution to this has been “Regulatory sandboxing,” where companies are placed into incubators or special, closed regulatory environments for new but potentially risky business models not yet addressed by current regulations. This environment allows for those business models to be tested in a supervised environment while regulators consider how their operations would be affected under new regulations. Virtual worlds with sufficiently complex economies and political environments may offer opportunities to actually test many iterations of regulations and policies to get some sense of loopholes and impact.

In response to concern about how some of the approaches might affect gameplay in live MMO environments, a few players of EVE Online provided their take on the potential impacts. The use of regulatory sandboxing and general research didn’t seem to faze any of them, even where it involved data collection. However, when they expressed initial confusion about the potential impacts of having professionals conducting gray zone wargames to impact in-game geopolitics, asking for a strict definition of “gray zone.” When given one, they laughed and clarified that it is simply what the EVE community refers to as the “metagame.” Spycraft, narrative campaigns, social engineering, and sabotage were already so commonplace as to be considered comically benign. When asked what it would mean to have professionals actively attempting to “ruin” the game for others by attempting to impact market prices or disrupt trade in order to better understand food security risks, interviewed players pointed to a recent post from one of CCP’s own developers responding to a player on a public forum. The developer noted that many had come to EVE to ruin the game for other players but have done little more than generate “oodles” of content for other players, and they proudly stated that “we are all still here, trying to ruin it for each other. And it is great.”  Maybe few game platforms are appropriate or resilient enough for this particular kind of productive and participatory serious game. However the fact that one exists bodes well for the future of serious game implementations.

Wed, May 19, 2021

Games with serious consequences: Extremist movements and kayfabe

Extremist movements and emergent collectives have found a home in online communities and platforms. In this piece, Nonresident Fellow R.J. Cordes outlines how such groups blur the lines between games and reality and presents a strategy for handling online collectives effectively.

GeoTech Cues by Richard J. Cordes

Digital Policy Disinformation
Logo of the Commission on the Geopolitical Impacts of New Technologies and Data. Includes an 8-point compass rose in the middle with the words "Be Bold. Be Brave. Be Benevolent" at the bottom.

GeoTech Center

Championing positive paths forward that societies can pursue to ensure new technologies and data empower people, prosperity, and peace.