Network of Innovators: Expert Networking for Open Data

Open data is a growing but still nascent field where innovators need all the help they can get to find technical, policy, and practical expertise for opening data and translating that data into impact.

At the IODC, we are testing the Network of Innovators ADD URL, an expert network to help open data practitioners find the know how they need.

Expert networks are a relatively new phenomenon in the government context – with notable exceptions like the United States Air Force Research Lab’s Aristotle and Australia’s business-development-focused Expert Network. They have, however, existed in various fields for some time. VIVO connects scientists within and across disciplines to enable knowledge sharing and collaboration. Zapnito is even offering businesses the ability to create their own internal expert networks. And, of course, millions use LinkedIn build their professional networks.

At its core, an expert network is: (a) a database of people’s profiles with some rich information about them; (b) the ability to navigate the database along various dimensions;  and (c) the ability to take some concrete actions on the results.

This is actually not that different from a dating site when you think about 🙂

 

Designing the database

The first task at hand is the design of a people-centric schema to describe members of the network. Beyond the traditional and objective attributes – e.g. name, location, picture, education –  the schema must describe a set of skills and experiences related to the application domain, open data for us. Taxonomies and ontologies will be critical to make sure people use the same ‘language’ when describing their expertise.

Populating the database

Now that we have an empty database with a rich structure, it is time to populate it.

The most obvious approach is to have network members self-report information about themselves, with the well-known pitfalls – e.g., people bragging about expertise they don’t have or people omitting expertise they do. Another option is to have network members report about other members. This is what LinkedIn tries to achieve using endorsements. Yet another option is to rely on external sources to populate members’ profiles.

These various approaches can be used either separately or combined. Note the use of external sources has been working very well in academic fields where publications and patents are a rich source of expertise about people – Harvard Catalyst Profiles is a prime example.

The Network of Innovators adopts a blended approach. We combine open sources of data, such as basic conference registration information about who is attending the IODC, with self-reported data gleaned – not from open ended questions – but by asking people to identify the questions they could answer if asked. For example, how might I gain organizational approval for opening data?

 

Navigating the database

Just like for any information retrieval task, we have the choice between two complementary approaches: search and browse.

In the search mode, people looking for an expert will express their need using a query that will leverage the rich schema we described above. A query for an expert might include requirements for location, spoken language and, of course, expertise. The search interface might incorporate some extra knowledge such as synonyms, support for natural language and richer information about geography.

In the browse mode, people start from an entry point and start navigating along various dimensions, e.g., “people from the same country,” “people from the same organization,” “people speaking the same language,” etc.

Of course, a typical user might combine both approaches iteratively to find the best matches.

For each result set, the tool may also provide some nice visualization to help compare candidates and feature-rich profile pages.

 

Connecting with the experts

Once a set of experts has been identified, one can take some concrete actions, e.g. adding them to one’s address book or contacting them directly (via phone, email, SMS). More advanced social features can also be added, such as the ability to ‘follow’ people.

The modes of interaction and the actions that can be taken are largely network-specific and should respect people’s privacy and preferred way of interaction.

Two elements not to be ignored when building such an expert network are the engagement layer and the experiment layer. The former enriches the tool with dashboard and game mechanics to encourage people to continue use or to discover new features. The latter allows for A/B experiments and a better understanding of how the network is actually being used.

 

This is just the beginning

The technical components are a necessary but not sufficient condition for an expert network to be successful. Just like building any community (offline or online, see [2]), bootstrapping such a network is hard since early adopters will find little value in a largely uninhabited network. Other challenges include:

  • devising incentives to convince people to populate their profile and keep them fresh and accurate;
  • articulating a clear value proposition for people to actually use the network, especially on the demand side; and
  • motivating people to integrate the platform into existing workflow and communication processes.

In order to overcome the incentives challenge, we are testing the tool here at IODC where people are eager to learn from one another. But we will learn what works and what doesn’t.

The government setting poses another big challenge: asking for help and sometimes admitting ignorance is not necessarily part of the culture.

At GovLab, we are exploring some of these issues as part of our Network of Innovators project. Our first experiment will take place during the International Open Data Conference. We encourage you to try the app and visit our booth.

 

Resources