Tag Archives: Kublai

The rise and fall of Kublai: the difficult interface between bureaucracies and networks

2 Replies

I have resigned as director of Kublai, the first web 2.0 project of the Italian central administration. My directorship had been an interim one for the past year: the Department for development policies of the Ministry of economic development, Kublai’s home base administration, has initiated a procedure for launching its final year and recruiting a new director, but the procedure met several difficulties and is still pending, after more than a year. The officer in charge of Kublai, Tito Bianchi, had asked me to stay on and bridge this gap, so I ended up with an unpaid role of interim director. Meanwhile the Department itself had to face a crisis, not directly connected with Kublai; as a consequence, starting in February, Tito’s contract was not renewed. In this new situation I did not feel like I could stay on, even as an interim director. I still hang out in Kublai as a member of the community, and still learn a lot from Kublaians. I am extremely proud of the liveliness the community displays even in difficult times: in these days we are discussing whether to spin it off from government project to the third sector. If it works, it will be very unique: a demand-driven policy spinoff! And even if it fails, it will make for an extraordinary experience trying it. I am very grateful to the Department for the opportunity to develop such an advanced project.

Over and above the role I played in it, Kublai’s story seems to be representative of a tough problem in public policy: administrations find it hard to manage the interface with the online communities they collaborate with – even if they originate them. Online public policies can be incredibly efficient, because they enable citizens to contribute by taking actions — I should know, since I have designed and delivered more than one. In March 2009, ten months after kickoff, an independent study showed that two thirds of the comments feeding into business plans writing came from the community itself, and only one third fron the project’s paid staff. The key to high efficiency is the community’s activism, and community ties can break down if they are not lovingly maintained. A decrease in the credibility of the founding group, or even simply a slump in alertness, might take away the motivation to participate at any time.

In the last few years I spent a lot of time thinking very hard about policy communities, especially those that live online, and I learned much. It turns out that running them effectively requires some conditions to be met:

  • an adequate planning horizon; if you need three years to build a resilient community, you have to have three years of guaranteed operations. Stopping before achieving your goals means wasting all of the resources invested.
  • radical transparency: access to what goes on behind the scenes has to be allowed to those interested (and not imposed to those who are not). Active participation requires participants to feel ownership of the project, and we don’t feel that if we think we are not being told the whole story.
  • low costs: a project based on voluntary work needs to show volunteers that its staff brings to the table a value worth the resources paid to it in salaries. This is a sensitive point, because the collaboration between staff and community is inherently unequal — we collaborate, but I get paid and you don’t — but, if the staff works well and hard, it is still acceptable. Justifying high overheads, on the other hand, is really hard. Every penny that goes down that route instead of the project’s own activities weakens its credibility.
  • fast delivery, with little or no timeline uncertainty, and therefore streamlined procedures. Social dynamics in online communities are fast and inherently emergent, which means you cannot control them top-down. When stuff happens, it happens, and you have to act. If something needs doing now, you need to be able to go out and do it now. If this is not possible, you need to commit to a date for it to get done. “We don’t know, we are waiting for a green light From On High” destroys credibility, hence participation, hence efficiency,
  • lowering the barriers between paid staff and the community. People participating to an online community on a voluntary basis make great candidates to become its staffers. They know it well and love it, they share its meeting places and times. Conversely, many public sector employees are harnessed by rigid office hours and badges that keep track of the time they spend in the office; many more are sequestered behind corporate firewalls that inhibit the use of social networks (major tools of this particular trade). Requiring a person in this situation to work on a project like Kublai means asking for trouble.
  • All of the above requires a lot of management leeway; this makes it inevitable that the assessment of these projects is carried out by weighing its results rather than checking its conformity to procedure.

Unfortunately, these conditions do not match well with the work style commended to Weberian bureaucracies — and most public administrations are Weberian bureaucracies. In the case of Kublai:

  • it has not been possible to guarantee the planning horizon. As we prepared to launch, the Department decided to unbundle it into three steps, one per year. The reason: a three-year project would have required a European tendering process, that was estimated would require a year and a half. The results: at the second year the project was funded, but in a way that undermined its effectiveness (see below) and with a gap of three months for formally reinitiating the project and seven months to actually go out and do stuff; for the third year we designed an architecture that was on paper more effective and used two delivery channels, but one of them was never initiated. My personal history is a sign of the stop-and-go nature of administrative processes: I directed Kublai for three years, and I had a regular contract for a total of sixteen months. The remaining twenty months were interim.
  • full transparency is technically difficult. Even when public decision makers have the best intentions, they live in a system so complicated as to be incomprehensible for the uninitiated. For example, Kublai’s first year was delivered through an in-house company of the State, Studiare Sviluppo. In its second year, the Department’s legal office changed its mind and claimed that would not be possible: though Studiare Sviluppo and the Department itself have a common origin in the Ministry of the economy, the creation of the Ministry of economic development in 2006 (which contains the Department, but not Studiare Sviluppo) had created a rift between the two organizations. Delivery would have to go through Invitalia, another in-house of the State. This distinction is difficult to understand, so much that it remains controversial within the Department: in 2009 some of its programmes continued to flow through Studiare Sviluppo, but not Kublai. How do you tell a story like this in the “About” page?
  • costs tend to balloon. Kublai 2009 had a higher budget than its 2008 version, but its effectivess seemed to decrease. Why? Partly because the overhead had gone up. Studiare Sviluppo charges 15% of the project’s budget to take care of all the administrative details; Invitalia’s overhead is more difficult to estimate and it varies from one project to another; with exceptions, it seems to oscillate between 30 and 40% (source: talks with several employees and former employees of the Department. I asked Invitalia for a precise figure, and am waiting for an answer).
  • decision making always takes a long time, and it is always subject to uncertainty. To give you an idea, Kublai 3 should have started in January 2010. The idea was to activate a project, led by Invitalia, and use that as co-financing of a larger project, funded by the European Regional Development Fund. This, however, would have required a change in Invitalia’s accounting system (in ERDF projects human labor must be accounted for as a direct cost, excluding overhead). In May we realized that Invitalia could not or would not make this change fast enough (I’m told they are still in the process). Tito had to initiate a second procedure which tapped into a different financial source; this new procedure was launched in June, but moved very slowly, with a lot of backtracking and all kind of pitfalls springing up, without us apparently being able to see them coming and avoid them. We thought it would be over in September, then in November, then in February, and we turned out to be wrong every time. In February Tito and I agreed to come out. The truth is that no one can tell what will happen and when.
  • in Kublai’s first year we used Studiare Sviluppo’s flexibility to recruit our staff directly from the community. We used very small contracts that recognized and encouraged voluntary work without forcing creatives to turn themselves into consultants to the public sector to work at Kublai. The vision behind this was to lower the barriers between “professionals” and “committed citizens”, and to build an ethics of fairness, where it could not happen that two people would be doing more or less the same work with one of them getting paid and the other one working for free. Starting in 2009 this process stopped: Invitalia’s internal rules dictate that only a small part of a project’s budget can be outsourced (15% of the budget in the case of Kublai); what’s more, it’s normal for recruitment procedures to take months, and even the smallest contract has to be signed by the CEO. Result: the administrative load of many small contracts was simply unsustainable. We had to retreat into a polarized version of the staff: on one side professional business plan advisors (public or quasi-public employees, who typically appreciate the stability a steady job gives them, and have little or no firsthand experience of business creation, let alone in the creative sectors), on the other side the creatives from the community, that we could not even cover travel expenses of if they held a presentation about Kublai at a creative industry event.
  • as for evaluation, Tito and I built a structured evaluation activity into Kublai 3, but that’s stuck with the rest of the project: the project is at risk not to be assessed at all.

With all of these problems, it is a wonder that the project was not born dead! If we we got anything at all done (2500 members, 400 projects being discussed and tens of businesses launched are an important result after all) is because some people poured an unreasonable amount of work and passion into Kublai. My staff was very generous, turning from paid staffers to volunteers when contracts expired and were not renewed (I absolutely need to quote Cristina Di Luca, Walter Giacovelli and Marco Colarossi); the officials in charge for Kublai at the Department (Paola Casavola, at the time the director general who made the initial decision, Tito Bianchi, Marco Magrassi and Giampiero Marchesi above all) have been visionary leaders from the policy maker’s side (Tito in particular has really put himself on the line for Kublai); Alfredo Scalzo at Studiare Sviluppo and Nicola Salvi, Federico Venceslai, Angelita Levato and Danila Sansone at Invitalia used all the autonomy their respective organizations allowed them to try and meet this project’s special needs. We have been a kind of pirate ship, part Luke Skywalker and part Dilbert: I remember Nicola posting from his own laptop and through the neighbor’s open wi-fi, because the corporate firewall would not allow him access to Second Life. Despite all of this commitment, we were a competent and motivated crew sailing a leaking ship. When the work input of the non-Invitalia staff, — the least embedded in bureaucratic procedures and way of thinking — was discontinued, Kublai clearly lost momentum.

A conclusion I drew from this extraordinary experience is that the interface between bureaucracy and network is not easy to navigate. Kublai can be viewed as an attempt to coproduce and codeliver a public service — assistance towards business creation — traditionally delivered (not too well) by local offices. It is a very hot theme: it is, for example, one of the key themes of NESTA’s Public Services Lab, that I have been in conversation with lately. From my vantage point at the Council of Europe, all European governments are moving in this direction, to defend the welfare state in the face of a structural fiscal crisis. I predict it will be difficult to get lasting results if the conditions I listed won’t be guaranteed.

I see only a way that they can be guaranteed: a new deal between government and the women and men who work for it. Such a new deal would work like this: administrations have to give trust and breathing space to their servants, be they employees or consultants like myself; and then assess their results, rewarding people who get results and punishing people who don’t. If there are abuses of that trust, they will be dealt with on a case-by-case basis: designing an entire system to prevent abuse is at a high risk to making it too rigid, disabling people to offer their best ideas. Designers of social media have a mantra: if you design a space to deal with trolls, your users will be trolls. Similarly, if you design a system to harness lazy or corrupt employees, your best staff members, frustrated and humiliated, will leave, and you will be left with the lazy and corrupt ones. It’s time to make choices; who will future public policy be designed for?

A quiet Wikicratic Tuesday

Leave a reply

On Tuesday, February 15th, here in Milan, I will take part in two events I find interesting from the point of view of wikicracy, that is collaborative redesign of the world we live in.

The day (starting at 9.30) is to be dedicated to a conference on service design organized by the Association of Industrial Designers and Milan Polytechnic (info). I am even a speaker: Kublai is, in this context, treated as a well-designed service, and my team and I as its designers. We are even metadesigners, because another of the services showcased is to be CriticalCity Upload, that was incubated from within Kublai.

The evening is dedicated to an intriguing experiment. Pietro Speroni, mathematician and supergeek, has invented a way to find shared answers to open questions, based on what he calls “human genetic algorhythms”, and is going to test it with the help of whoever shows up. The idea is this: we start with a question. Anyone can offer an answer. Then everyone votes each answer (only “yes” and “no” votes allowed). A software looks at the votes and extracts its Pareto front, i.e. the set of non-dominated answers: it means that an answer voted by Alice and Bob will be discarded if there is at least one other answer voted by Alice, Bob and Chuck. Mathematically, this second answers dominates the first one, because it has all of its voters plus at least one. Then, everyone is asked to write new answers, feeling free to use bits and pieces from the ones in the first-round Pareto front. This gives rise to a second generation of answers, which use the “genetic material” already selected by the first iteration; then a new vote is called, the software computes the Pareto front and on it goes until a unique dominant answer evolves. The software, naturally, is called Vilfredo.

Pietro and his group have been experimenting, and so far the human genetic algorhythm always converges. In a way it converges too fast: Vilfredo gets to an unique dominant answer in ten generations or less, whereas non-human genetic algorhythms need more like ten thousands generations to converge. Makes sense: evolution has to blindly try every conceivable mutant and let genetic cul-de-sacs go extinct, whereas human discussion can discard a priori a lot of unpromising alternatives. If the question is “what shall we do tonight?”, answers like “let’s declare war to Guatemala or “let’s all go bathe in molten lead” don’t need to be discussed and discarded: they just never come up. This should teach us that emergence in the natural world (where agents are dumb, like neurons or ants) is very different from emergence in the social world (where agents are smart, like people or companies). At 8 p.m. at Via Boifava 29/A (Facebook event). You can say anything, but not that we don’t live in interesting times.

Un tranquillo martedì wikicratico

Leave a reply

Martedì 15 febbraio, qui a Milano, parteciperò a due eventi interessanti dal punto di vista della wikicrazia, cioè della riprogettazione partecipata del mondo in cui viviamo.

La giornata (a partire dalle 9.30) è dedicata al convegno sul design dei servizi organizzato dall’Associazione per il Design Industriale e dal Politecnico di Milano (programma e invito qui). Sono anche relatore: gli organizzatori considerano Kublai un servizio ben progettato, e in questo senso io e il mio team ne saremmo i designer. Siamo anche metadesigner, perché un’altra relazione racconterà CriticalCity Upload, incubato proprio dentro a Kublai.

La serata è dedicata a un esperimento molto intrigante. Pietro Speroni, matematico e supergeek, ha inventato un modo per trovare risposte condivise a domande aperte, basato su quelli che lui chiama “algoritmi genetici umani”, e ne farà un collaudo insieme agli intervenuti. L’idea è questa: c’è una domanda. Chiunque può dare una risposta. Tutti votiamo le risposte (“Sì” o “No”). Un software esamina la votazione e ne trova il cosiddetto fronte di Pareto, cioè l’insieme delle risposte non dominate: significa che una risposta che ha avuto il voto di Anna e Beatrice verrà scartata se ce n’è almeno un’altra che è stata votata da Anna, Beatrice e Carlo. Si dice che la seconda risposta “domina” la prima, perché ha tutti i votanti della prima più almeno uno. Fatto questo, si chiede ai partecipanti di riscrivere, usando – se lo vogliono – i pezzi delle risposte selezionate nel primo round. Questo dà luogo a una seconda generazione di risposte, che usano il “materiale genetico” già selezionato, su cui di nuovo tutti votano e il software cerca il fronte di Pareto, e si itera fino a che non evolve una sola risposta dominante. Il software, naturalmente, si chiama Vilfredo.

Gli esperimenti di Pietro e del suo gruppo finora hanno mostrato che l’algoritmo genetico umano converge. In un certo senso converge troppo in fretta: Vilfredo arriva a un’unica risposta dominante in dieci generazioni o meno, mentre gli algoritmi genetici in senso stretto convergono in diecimila generazioni. E ti credo: l’evoluzione deve provare tutte le mutazioni alla cieca, e lasciare che i vicoli ciechi genetici portino all’estinzione, mentre la discussione tra umani scarta a priori un gran numero di alternative. Se la domanda è “cosa facciamo stasera?”, per esempio, le risposte tipo “dichiariamo guerra al Guatemala” o “andiamo tutti a fare il bagno nel piombo fuso all’Italisider” non hanno bisogno di essere discusse e poi scartate, ma proprio non compaiono mai. Questo dovrebbe insegnarci che i modelli di emergenza nel mondo fisico (i cui agenti sono stupidi, come i neuroni o le formiche) sono molto diversi da quelli del mondo sociale (i cui agenti sono persone, e quindi intelligenti). Alle 20.00 a Musicopoli, Via Boifava 29/A (evento Facebook qui). Potete dire tutto, ma non che non viviamo in tempi interessanti.Pietro and his group have been experimenting, and so far the human genetic algorhythm always converges. In a way it converges too fast: Vilfredo gets to an unique dominant answer in ten generations or less, whereas non-human genetic algorhythms need more like ten thousands generations to converge. Makes sense: evolution has to blindly try every conceivable mutant and let genetic cul-de-sacs go extinct, whereas human discussion can discard a priori a lot of unpromising alternatives. If the question is “what shall we do tonight?”, answers like “let’s declare war to Guatemala or “let’s all go bathe in molten lead” don’t need to be discussed and discarded: they just never come up. This should teach us that emergence in the natural world (where agents are dumb, like neurons or ants) is very different from emergence in the social world (where agents are smart, like people or companies). At 8 p.m. at Via Boifava 29/A (Facebook event). You can say anything, but not that we don’t live in interesting times.