Tag Archives: complexity

Dragon Trainer begins

Good news: a research project I helped to write has been approved for funding by the European Commission’s Future and Emerging Technologies program. The project is led by one of the scientists I admire the most, David Lane, and rests firmly in the complexity science tradition associated to the Santa Fe Institute. We intend to attack a big, fundamental problem: innovation is out of control. Humans invent to solve problems, but they end up creating new and scary ones. Which they tackle by innovating more, and the cycle repeats itself. Cars improve mobility, but they come with global warming and the urban sprawl. Hi tech agriculture mitigates food scarcity, but it also gives rise to the obesity epidemics. To quote one of our working documents:

While newly invented artifacts are designed, innovation as a process is emergent. It happens in the context of ongoing interaction between agents that attribute new meanings to existing things and highlight new needs to be satisfied by new things. This process displays a positive feedback […] and is clearly not controlled by any one agent or restricted set of agents. As a consequence, the history of innovation is ripe with stories of completely unexpected turns. Some of these turns are toxic for humanity: phenomena like global warming or the obesity epidemics can be directly traced back to innovative activities. We try to address these phenomena by innovation, but we can’t control for more unintended consequences, perhaps even more lethal, stemming from this new innovation.

We want (1) build a solid theory that concatenates design end emergence in innovation and (2) use it to forge tools that the civil society can use to prevent the nefarious consequences of technical change. It does not get any bigger! And in fact we got a stellar evaluation: 4.5 out of 5 for technical and scientific excellence and 5 out of 5 for social impact.

The project commits to building Dragon Trainer, an online community management augmentation software. The idea is to make a science of the art of “training” online communities to do useful things (like policy evaluation), just as you would train an animal too large and strong to push around. I am responsible for producing Dragon Trainer, and it is quite a responsibility.

I am superhappy, but worried too. Taxpayers foot most of the bill, and this makes it even more imperative to produce the absolutely best result we can. I will need to work very, very hard. I am seriously thinking of devoting myself to full time research for a couple of years starting in 2012. Does this make sense? What do yo think?

Dragon training: computer-aided online community management

In my book Wikicrazia I claim that the public sector, society’s system to pursue the common good, can be made smarter by mobilizing the citizenry’s collective intelligence. Accessing collective intelligence entails enabling a large number of individuals to coordinate on some common goal. Normally, this is done by means of online commmunities, that use the Internet as their technological infrastructure and where interaction is mediated by some kind of social bargain, with somebody to resolve conflicts and keep the group focused on the goal.

There’s a problem here. On the one hand, online communities cannot be run by top-down command and control: it is exactly the free action of their different participants that make online communities so incredibly effective in processing large amounts of information. On the other hand, public policies have by definition a goal which is set exogenously with respect to the community itself: whereas Facebook users are on Facebook to hang out, and it does not really matter what they do with it, the users of Peer to Patent are there to process patent application; those of Kublai to write up creative business plans; those of Wikipedia (not a public policy, but similar in this respect) to write an encyclopedia. Community managers, myself included, are trapped in this dilemma: practically the only way we have to figure out the social dynamics in our communities is to spend an unreasonable amount of time participating in them, and we try to steer them by rhetoric and persuasion. We end up navigating pretty much by gut feelings. And as communities scale – even to just a few thousand participants – it gets really hard to understand what is really going on.

I thought our work would improve a lot if we could augment our ability to read social dynamics of online communities by using software. In essence, a policy community is a social network, and as such it can be represented by a graph with nodes and links, and studied mathematically. The community’s social dynamics should be encoded into the mathematical characteristics of the graph that represents it: for example, the creation of a cohesive group of senior users in Kublai in 2009 was picked up by the crystallization of a structure called k-core. If we managed to build some sort of dictionary that maps social dynamics onto mathematical characteristics of the graph, we could use network analysis to detect community dynamics that are invisible to the eye, because they happen at a scale too large for human participants: and this would work even for very large communities, at least in principle.

I intend to develop this software as my Ph.D. thesis. Colleagues at University of Alicante and the European Center of Living Technology will help. I call it Dragon Trainer, because doing policy through an online community is like training a dragon, an animal too large and dangerous to order around. If you are interested in learning how we plan to do this, you can watch the video above (12 mins).

Dragon training: gestione di comunità online assistita dal computer

Nel mio libro Wikicrazia sostengo che il settore pubblico, il pezzo della società deputato al perseguimento dell’interesse comune, possa essere reso più intelligente mobilitando l’intelligenza collettiva dei cittadini. Ricorrere all’intelligenza collettiva vuole dire abilitare un gran numero di individui a lavorare in modo coordinato su obiettivi comuni. Questo in genere si traduce in comunità online, che usano Internet come infrastruttura tecnologica e in cui si interagisce sulla base di un patto sociale e di qualcuno che media i conflitti e fa in modo che non si perda di vista l’obiettivo.

Qui però si pone un problema. Da una parte, le comunità online non si possono gestire con il comando top-down: è proprio l’azione libera delle tante persone che le compongono a produrre la loro straordinaria efficienza nell’elaborare grandi quantità di informazione. Dall’altra, le politiche pubbliche hanno per definizione una missione da compiere che viene dall’esterno della comunità che le attua: mentre gli utenti di Facebook sono su Facebook per stare insieme, e non importa quello che poi faranno usando quella piattaforma, quelli di Peer to Patent sono lì per valutare domande di brevetto; quelli di Kublai per elaborare progetti di impresa creativa; quelli di Wikipedia (non è una politica pubblica, ma ne ha alcune caratteristiche) per scrivere un enciclopedia. I community managers, me compreso, si dibattono in questo dilemma come possono: quasi l’unico modo che hanno per interpretare le dinamiche sociali delle loro comunità è passare una quantità spropositata di ore online, e cercano di influenzarle con la persuasione, l’esempio, la retorica. Ma si lavora molto a istinto, questo è chiaro. E quando le comunità diventano relativamente grandi — anche solo qualche migliaio di persone – è davvero difficile capire cosa sta succedendo.

Ho pensato che il nostro lavoro migliorerebbe molto se avessimo un software che accresce le nostre capacità di lettura delle dinamiche sociali online. In essenza, una comunità di policy è una rete sociale, e quindi può essere rappresentata con un grafo di nodi e link e studiata matematicamente. Le dinamiche sociali della comunità si dovrebbero riflettere sulle caratteristiche matematiche del grafo che la rappresenta: per esempio, la creazione di un gruppo coeso di utenti senior in Kublai nel 2009 veniva segnalata dalla formazione di una struttura che si chiama k-core. Se riusciamo a costruire una specie di vocabolario che traduca le dinamiche sociali in cambiamenti nelle caratteristiche matematiche del grafo, possiamo usare l’analisi di rete per individuare le dinamiche di comunità che a occhio nudo non si vedono, perché sono “troppo macro”: e questo funziona anche per comunità molto grandi, almeno in linea di principio.

Sviluppare questo software sarà il lavoro della mia tesi di Ph.D. Mi aiuteranno i colleghi dell’Università di Alicante e dell’European Center for Living Technology. Per ora si chiama Dragon Trainer, perché gestire una comunità online che deve svolgere un compito esogeno è come ammaestrare un drago: che è troppo grosso e pericoloso per essere costretto a fare quello che vuoi, e quindi va sedotto o convinto. Se ti interessa capire come sarà fatto, guarda il video qui sopra (12 minuti).I intend to develop this software as my Ph.D. thesis. Colleagues at University of Alicante and the European Center of Living Technology will help. I call it Dragon Trainer, because doing policy through an online community is like training a dragon, an animal too large and dangerous to order around. If you are interested in learning how we plan to do this, you can watch the video above (12 mins).