Charles Armstrong, CEO of software startup Trampoline Systems, came to visit last week and delivered a talk to the company. I heard him at ETech and just loved the combination of social insight and technology. I knew my Nature colleagues would find it fascinating too, and we weren't disappointed. Here are my notes from his talk.
Background
Charles studied social and political science at the University of Cambridge, and while there got interested in ethnography. Key elements include qualitative descriptions and holistic approaches. Once upon a time he worked for an internet marketing firm in the building next door to Nature's current London office. He was struck by dysfunction of corporate information systems and had a hunch that natural communities were much better at distributing and processing information. So he decided to spend 12 months living in the Isles of Scilly on the island of St Agnes in order to study this.
St. Agnes
His emphasis was on informal observation, questioning and learning. Towards the end of the project he added a formal interview process. Supported by two external academics, Ray Jobling and Michael Young.
Isles of Scilly consist of 5 inhabited islands with 2000 inhabitants. St. Agnes is 2 x 1.5km in size with 80 people. Small-scale tourism is the main economy. There is also flower farming but it no longer makes money. It is a crofting community and the same land is typically worked by several generations of the same family. There are half a dozen tractors and one school for children aged 5-11. The mail arrives on a boat from the main island, St Mary's. The daily arrival of the freight boat is one of the main social rituals -- about 20% of the population turns out.
Information from outside also arrives on the boat from St Mary's. This may be relevant to only 5 people on the island but would make its way to them in short order without lots of other people being troubled with it. (Charles notes that these results are impressionistic and based on informal polls, not rigorous statistics.)
Information dissemination
From these observations, Charles derived some rules for information dissemination in the community:
- Groups have implicit authorisation parameters that govern how information is relayed.
- Groups play a function in pooling intelligence about who needs to know certain pieces of information.
- Groups themselves can be identified as targets for relaying information.
- Each person and group is identified with certain semantic "triggers" that activate relaying behaviour.
- The "further away" two people or groups are in a social network, the higher the threshold for relaying information between them.
These five mechanisms seemed to explain the efficient distribution of information. This is in contrast to large organisations, which are intrinsically hierarchical and less efficient in terms of reaching the people who need to know (and only those people). For example, 34% of business email turns out to be completely unnecessary. But informally the same mechanisms as St Agnes are working even in large organisations.
Previous strategies to overcome this problem include:
- Collaborative scoring (e.g., digg and Slashdot). This doesn't work in a heterogeneous community.
- Semantic sorting of unstructured data (e.g., Autonomy). This isn't responsive to individual relevance and interest, only to topics.
Sociomimetics
Could a better information system be created by mimicking the natural social behaviour of small communities?
Biomimetics is the closest parallel -- this looks to nature for the solution to a human problem. For example, Velcro is based on the observation that seeds stick to dog fur. Also, Airbus has developed a low-friction wing surface based on the scale structure of shark skin. Sir Joseph Paxton based his building designs on the weight-supporting structure of water lillies.
Charles suggests the term 'sociomimetics' to describe borrowing ideas from natural societies in order to influence the design of new technologies, especially information technologies.
Trampoline
Trampoline, the system that Charles's company has created, is a sociomimetic approach to information management. It uses the following steps to route information appropriately and automatically:
- Index content
- Find content matches
- Determine authorisation
- Evaluate social network
- Examine recipient preferences
- De-duplicate messages
Their current focus is on email, but the process is general. It brings large organisations closer to the efficiency seen in natural communities.
Their first product was launched in 2004 and is being used by 6,000 people. They are now working on an enterprise email management product to be used alongside (e.g.) MS Exchange Server. It looks at the message flow and uses the rules listed above to direct information to people who might be interested but otherwise wouldn't have seen it -- a kind of automated village gossip. It's programmed in Java and uses lots of open-source libraries. The UK Foreign Office is among their current customers.
He then demos a tool that visualises the natural communication maps within organisations. This information is a by-product of the system as it works out which information to route to whom. -- it looks to me like a kind of org chart for the 21st century.
He then looks at the routing of a particular email sent to the 'Lab' group at Trampoline Systems. By setting the threshold appropriately, it can be automatically routed to people outside that group with a high degree of interest and the right authorisation. The system also takes into account how people respond to emails. (Responses to an email are a 'vote' for it in a similar way that links are vote for a web page in Google PageRank.) This allows for people who are prolific in sending emails, but not similarly influential.
Q&A (partial notes)
Q: Privacy issues: Maybe not everyone wants to be formally recognised as a key node in the network (i.e., a gossip).
A: Yes, we don't release this info to all users. People can only see their own networks.
Q:: Scalability of natural community approach.
A: The limit for natural communities is about 400-500. In addition, the imposition of a formal structure has a damaging effect on info management.
Q: What next?
A: Focussed on email for at least the next six months, then will start to weave in other info.
