Swarm Intelligence and Human Systems

I'm gonna introduce Nancy Eckhart she is talking about swarm intelligence and human systems so Nancy has been a hacker hailing from right here or right here in this local area Seattle Washington she's been playing games on the internet since the early 90s and was a purveyor of text-based muds she is now a champion strategist and community organizer in the world of competitive augmented reality gaming Nancy leads hundreds of players and capture the flag events around the world can we all give a warm welcoming round of applause to Nancy hello Portland thank you so much for coming here all right I'm Nancy Eckert some of you may actually know me by my gaming handle Pongal in or just Pongo like some of you

I imagine I found my way here today because of games so my first exposure to the internet was through games on a spring morning in 1993 or so some friends of mine invited me to a lab in a basement at the local college where some of our older siblings were taking classes to play a text-based adventure game called a mud and in a few short hours I learned that I could use something called telnet to talk to people in Sweden now I didn't know what telnet was and 1993 was a year before you could really Google things to find answers to things like that but I wasn't gonna let that stop me so within a few

weeks I taught the modem at my school library to connect to the University using this thing called telnet so I could play games at school and talk to my friends across the world the following year I scheduled a free class period in the library as a monitor and I became the best student library worker my school had ever seen and I was never caught I even helped fix the computers so games are incredible motivators and we'll talk about that a little bit more later on what I love about gaming though is that gaming it offers some really unique opportunities to simulate systems that may or may not actually exist in the real world in order to solve problems

that do you can make a game simulate just about anything furthermore games remove obstacles for communication they give us opportunities to take calculated risks in environments that are generally lower stakes than the real world humans have used games for thousands of years to exchange information and facilitate cooperation to teach important survival skills if you can think of some of the games that we might have played of his children house hide-and-seek war play is critical in the animal kingdom for important teaching important survival skills and perhaps one of the most important survival skills in human systems is how to function as part of a group so there's a specific game that I'd like to talk about today very

briefly it's called ingress this is an augmented reality video game anybody play ingress yeah enlightened resistance would you tell me if I asked I'm enlightened anyway so this is an augmented reality game the predecessor to pokemon go and Harry Potter unlike these two games ingress has always been player versus player it's a team-based player versus player game there are two opposing player teams it's fixed sum so two teams are competing against each other directly for the same resources in a very capture-the-flag style format the game is also played in physical space in real time there is no off button and generally speaking the game is played by going to a specific location and performing some action on your phone

hopefully you have internet access so this team-based competition and physical space brings two very important and powerful primal forces into play tribalism and territoriality the competitive game involves going to places that are in some way challenging to get to which can sometimes require special equipment skills or intelligence big points are scored in the game by means of secret operations involving precise cooperation between dozens to even hundreds of people across the world international operations exist operations exist where people do not share the same common language the emergent social game is ruled pretty much by spreadsheet spreadsheets and intrigue and in many of these operations if one person fails to do their job correctly the entire operation is a wash

generally it's hundreds of hours of planning just down the drain now I like to study and occasionally interact with systems particularly intelligent ones so I played this game pretty competitively for about five years it became a community organizer in an area of operations roughly covering the Cascadia watershed I've won several tournaments across the world and I also helped found a guild that eventually became a religious order ordained its own priesthood so I guess you could say I'm a cult leader so I did learn a lot in this game about how groups interact in physical and digital spaces I learned a lot about social engineering that is persuading people to act in certain ways usually to go to a place to do a thing

so you can beat the other team for the win I learned a lot about why people cooperate and why they often don't so some basic challenges to cooperation are illustrated by two classic problems in game theory which I'll briefly briefly described here the first one is called the stag hunt so in French philosopher jean-jacques Rousseau described the stag hunt with a kind of story two hunters can either cooperate to hunt a stag which is harder than a hare or they can individually Hunta hair which only feeds the person that caught it cooperation is a choice that requires trust that the other party will not defect but this trust has a cost associated with it to

give you maybe a real-life example you have a security consultant and a client this is a cooperation built on trust you must cooperate and toward to achieve your mutual goals of better security however cooperation comes with a cost the security consultant doesn't want their hardwork to be wasted finding solutions that will not be implemented the client doesn't want their money to be wasted on solutions that cannot be implemented so this is an example of a stag hunt perhaps a better known example is the prisoner's dilemma this also describes the problem I'll be it in a slightly different way so suppose the two players here are represented by the colors of green and blue similar to the

game and that each player can either choose to cooperate or defect if both players cooperate they both receive a payoff for cooperating if they both defect they both receive a punishment if blue defects while green cooperates then blue receives a larger payoff well Green receives sort of a sucker's payoff a punishment similarly if blue cooperates while Green defects blue receives a punishment while green receives the payoff so both of these games illustrate a basic problem of cooperation misplaced trust has a significant penalty for the nice party that doesn't defect first so this model the prisoner's dilemma assumes that the parties can't actually communicate with one another however there's a very interesting variant on the prisoner's

dilemma called the iterated prisoner's dilemma where you can involve a strategy based on what's happened in the past so there's a professor professor Robert Axelrod created a series of contests in order to find a winning strategy to this iterated prisoner's dilemma he invited experts from many different industries to compete and the strategy submitted ranged from very nice that is always cooperate no matter what happens to always defect there mean strategies and Axelrod discovered some pretty interesting things about humans playing this game the first one was that cooperation required in almost every case for there to be a chance of a repeat interaction later on that is to say on the last round of the game

everybody knew that it was the last round everybody defected to maximise their games Axelrod's contest also found out that the long-term gains over time were achieved by a strategy or strategies that used a very simple tactic so simple it took about four lines of code to express basically it was this lead with cooperation and then do whatever your opponent did last it's called tit for tat and it won everything Axelrod ran another contest and it won again even against he'd let everybody know that tit for tat was going to be in the second contest so people were intentionally trying to beat it and we're not able to do so so he analyzed the short and the long term games

against nice strategies and mean strategies and specific vulnerabilities against them and eventually he arrived at a slightly revised version of this strategy it was called tit for two tats basically lead for lead with cooperation forgive the first affection and then do whatever your opponent did last so world peace right we figured out human behavior no humans are messy human systems have all sorts of biases and competing goals and new problems that are introduced because of these competing goals and they always become more pronounced at scale one criticism of game theory to predict human behaviors is that the models quickly become too complex to even model and understand still we might have touched on something here with this idea of

leading with cooperation the problem of cooperation with mmo's and augmented reality games can also be mapped out to a kind of stag hunt or prisoner's dilemma and many times the incentive is built right into the game rules words usually loot require more cooperation and in many cases the more cooperation you get the better reward you get and a lot of times again if one person defects everybody goes home empty-handed or pays a penalty of some kind so maybe the most interesting takeaway from Axelrod's experiment was that he found out that even small isolated groups using the tip for tat strategy could eventually out score long term against large populations of greedy strategies or mean strategies the secret

to this was something called clustering which is the tendency of people using nice strategies to interact favorably with each other the key was that these small pockets of cooperation connecting to each other was what enabled them to basically payout these long term gains so this is a really important thing to keep in mind when you're seeking to coordinate large groups of people the relationships between the people in the group particularly their likelihood to of cooperation or defection among each other are fully as important as the people themselves and so the organization of free agents in a system is a basic problem of cooperation and this is one more example of how games can simulate real world systems with

slightly lower stakes and this is where guilds come from guilds are systems for organizations and games in order to facilitate cooperation and one would hope without resorting to force or bribery when we're building an organization like a gaming guild this is often the first system that we try to build it's a pyramid this is a very intuitive structure you have your founders at the top guild leaders calling the shots maybe they have a cabal of trusted advisors near the top that are come up with many strategic solutions and then at the bottom of the pyramid you have your so-called rank and file in gaming a lot of times they're newer players maybe they're casual players but if the people at the top are

understood to be the strategist doing the planning the people on the bottom are the ones with executing those plans hopefully in service to some goal common to the entire team information flows from the top down and if you're at the bottom chances are you have an individual goal of working your way up to the top so this approach has some strengths again chain of command is very well understood it answers that question of who's in charge very decisively which can be invaluable in many ad hoc situations particularly emergency situations however it also lacks flexibility it lacks the ability to adapt to changing needs and systems analysts refer to this as a brittle system it's fairly

vulnerable to attack in a few key places mainly near the top of the pyramid or where the information comes from where the decision-making comes from if you compromised the guild leader or the CEO you own the entire system but what if we could do something like this at first glance this may just look like an unintelligent mob but in actuality there's some crazy teamwork going on here each node in this network is a group of starlings flocking is interdependent to maximize its effectiveness now this isn't a new concept it's used widely in programming and this is not accidental just as artificial neural networks are patterned on our understanding of how the brain works artificial swarm intelligence is in

computing is pattern on observe observed behaviors in the animal world fish scrolling for defense reducing threats to the individual through collective action army ants solve complex problems through cooperation adaptive specialization and communication they use pheromone scent trails to communicate with each other and they make rapid decisions through a kind of consensus building it's very fascinating so what exactly are they doing to use a programmers analogy they're following a set of very simple instructions generally things like don't run into other animals and they're processing information from their nearest neighbors how close is my nearest neighbor what direction are they going that that sort of thing so when it first started speaking on this subject I used swarm intelligence

to refer to a set of competitive strategies in ingress that my teammates and I developed an augmented reality gaming through a highly iterative process basically to meet the challenges of tournament play rules would change frequently as would our playing field and the composition of our teams so over time our strategies had to adapt and become more flexible swarm intelligence started as a metaphor for how to organize movements of large groups hundreds sometimes over a thousand basically larger than Dunbar's number in public spaces this is these are actually two slides from a tournament in downtown Portland in 2017 so using this metaphor for pre-programmed instructions and nearest neighbor communications we came up with a strategic model that was

consistently successful in tournament play we found the traditional command and control structure while useful was terrible at adaptation it was also very slow when you have information that's being relayed through multiple channels all the way from the top down to the bottom you lose precious seconds and that could cost you the game it also leaned heavily on on single points of failure prone to information bottlenecks and siloing it scaled poorly guilds ran by a single figurehead in that sort of pyramid model tended to max out at about a hundred and fifty there's Dunbar's number again players or so we started looking at other games we ran across Evon lines appropriately named goon swarm infamous for rolling over their

competition with armies of new players but things really started to get interesting when we discovered that these tactics in use outside the game and one of the first examples we found was the US military so General Stanley McChrystal perhaps best known for his command of the Joint Special Operations Command fighting al-qaeda in the mids describes their decentralized organization basically a swarm as something the US military forces under his and actually had to emulate this is something that al-qaeda was doing that they had to emulate in order to fight them effectively and what they found was that timely solutions for unknown problems tended to emerge from the bottom up rather than from the top down so he refers to this as the team of

teams approach one very different implementation of swarm intelligence was the Swedish Pirate Party founded in 2006 the organisation elected members to Parliament and expanded to 33 countries through an approach that is literally described by its founder as a swarm he wrote a very excellent book on this I do actually have a reading list at the end of this slide presentation if you're you're interested but eh it's success he or he he owes relates its success to a conscious effort to empower the end nodes in his organization many of them were young people

so to reiterate the qualities of a swarm and why we might want to organize something in this manner so this offers solutions for many of the challenges we just talked about in the command-and-control pyramid approach when done well swarming isn't it incredibly fast and efficient way of coordinating large groups hundreds or more it's flexible one of the most desirable qualities of the swarm intelligence framework is adaptability these strategies can be applied to several different scenarios and they can integrate right really well with more traditional hierarchies it's also quite resilient when you attack one part of the system in contrast to our command and control structure we talked about earlier attacking any single part of the

system doesn't really accomplish a whole lot since the swarm doesn't need information from a leader of the entire swarm removing any one node here merely establishes a new nearest neighbor so it has this quality of resilience that resembles a self-healing organism organism and is efficient it leverages the skillset of the entire group as opposed to just the people at the top again this is in particular is what the swedish pirate party attributes its massive success to now these are all really excellent qualities and advantages but there's also a hidden possibility here and that's collective intelligence that is problem-solving at scale this is really the brass ring that we're shooting for here so when individual is acting autonomously they

have a strong purpose mastery in what they're doing and communication with their nearest neighbors collective intelligent behaviors will start to emerge swarm intelligence strategies prioritize empowerment again of the entire network focusing on what everyone can do swarms benefit from and leverage diverse perspective and skill sets and because they're adaptable because they're resilient because they're efficient they can solve big problems problems that are too complex for any one individual to understand sometimes without even consciously realizing that doing it any DNG fans out here who plays a bard tell me what you think about charisma based skills they break the game don't they anybody ever DM with a bard in the party yeah it's it's fun social engineering is

concerned with exploiting psychological vulnerabilities to influence people and trust an information flow impact relationships between agents in a systems so it makes these intuitive routes to soft power so we talked about leading with cooperation for max gains over time so how do we do that when we're building trust in a large-scale social social system whether it's gaming or not newbies are the most vulnerable and important agents in the system and why is that bluntly it's because they're the least informed in a tribal situation whoever gets to the newbies first will often have great influence over them later in MMOs and a RGS helping the newbies is a great way to build a guild and it's it is also a path to leadership

a lot of people will establish trust with new players via power leveling they use mentorship and sponsorship to build trust and to convey information about our about your values and ultimately to define new win conditions this is a double-edged sword of course like any viral infection information offensives work by targeting the most vulnerable members of a population in this case the least informed in some cases it can be recognized as an actual bid for power so we've seen in in games there have been incidents of say bad actors reaching out to new players in public channels establishing themselves as gate keepers promoting competing viewpoints and in a couple very unfortunate cases isolating people from the community so

let's step back to motivation for a second and social engineering so any hacker knows that the tools to disrupt the system are the same as the ones to use to empower the system so how can we use these tools for good anybody recognize this badge yes so when one of the guys of the local hackerspace heard that I was going to Def Con last year he gave me a small plastic badge with a bunch of dice printed on one side he told me the badge contained a secret message about an exclusive party and he challenged me to break the code within about 48 hours I knew more about cryptography than an online course from MIT had taught me in a month I also had

learned one or two things about cross-site scripting but that's a story for another time gamification is in fact the most reliable motivator on the planet games have been shown in many cases to be more reliable than just about any other motivator to get people to act for or against their own self-interests now the reason for this is in large part due to two very important chemicals in our brain one of them is oxytocin which is fundamental to trust and bonding it is present in all social games and really all social systems the other is dopamine which controls our reward centers and defines what we think of as worthwhile experiences add them together and you have a powerful cocktail for social

engineering at scale so in 2008 a group of researchers at the University of Washington created an online puzzle game called Foldit which gamified of all things protein folding fold folded leveraged the pattern recognition abilities of thousands of human players and in some cases outmatched the problem-solving abilities of computer algorithms the researchers refer to fold it as a serious game and they would go on to produce several more so these serious games have crowdsource many advancements in science and medicine basically by breaking these complex problems down into teachable bits presenting those bits in a game of fide context and then just catalyzing large-scale cooperation and communication among its players meanwhile at a conference like this one

we have many games to increase engagement but also to teach the skills relevant to security so I hope all of us here at besides are involved in a game like this but what I'd like to do today is break down the structure a bit because I'd like you eventually to all create your own the first thing you need to do is establish your win condition this is the reason your game exists or organization it's sometimes called a vision or a purpose it can be something like collect flags fold proteins work through this mountain of data to find the patterns capture the flag security capture the flag is a very obvious example the second thing you need to do is break it

down to those teachable bits those first principles as far as possible you need to break these skills down to digestible bits that can be mastered and this is important autonomously it is not enough for you to give people the information they need and help them along you need to empower them to not only master these skills but you need them to be able to teach these skills to other people without your oversight so ultimately agents in the swarm need to be able to train others to do the same this unconscious competence mastery is one of the necessary conditions to unlock this collective intelligence we were talking about earlier so the more we can do autonomously the more mental bandwidth

we all have for focusing on higher level tasks so one of my favorite examples for a game that does this is Lock sport one thing that I love about it is it's got a great learning curve somebody can pick their first lock in an hour of practice and they will come back with a completely transform perspective on physical security this makes it a wonderful game for people new to the security industry post it's cool it's fun I like picking locks the third thing you need to do is establish and facilitate communications with everybody's nearest neighbors within your swarm again it is not enough to make sure you give them all of the information they need you as the game

designer need to get people talking to one another one other great example would be the many of the badge games that are created for InfoSec conferences I particularly enjoyed the the DEFCON badge game from DEFCON 26 where that enabled chance encounters I think they all do this but enable chance encounters between strangers for a win condition it's got people talking to one another but also talking to one another about the things that they cared about about these skills that they had mastered so here's where we start to see we can how we can use games as a way not just to teach skills mastery but to perform but to perform social engineering at scale from an organization as small as a

gaming guild to grassroots movements with thousands of members there's there's a real power here and it's a power that we all possess you all possess you all have this power it can be used to empower and disrupt and it is my hope that you all now can recognize when it's being used and in whose interests anyway you all now have the guidelines make good choices