An exciting week in Singapore with a series of events including IRAHSS 2013 all held in the Raffles City Convention Centre. That also means an upgraded corner room in the Fairmont Hotel which also has my second favourite swimming pool of any hotel in the world. Tuesday and Wednesday I have the job of summarising each day which means I need to stay alert and focused. Thursday and Friday is a mixture of a foresight event and a complexity conference and I then return home on Saturday. Today was an expert review of RAHS 4.0, the next generation of Risk Assessment and Horizon Scanning system for Singapore and as one of the main designers of 1.0 (which itself spawned SenseMaker®) I had the job of reflecting on history at the start of the event, then being part of a three man review panel with Peter Ho and Jeffrey Cooper at the end. The opening reflections were made with Jeff Jonas who like me was acquired by IBM but unlike me is still there! As it happened we were both dressed in black, something captured by the graphic recording.
Now Jeff and I provide different perspectives on the whole issue of how you find patterns, my emphasis on human metadata compliments his ability to mash the numbers. A fascinating conversation with him over lunch on the predictive work he is doing on asteroid collisions really brought the home to me not only what is possible here, but also the interesting way he things about the problem. Either way we are going to spend some time together in August exploring things further and I am looking forward to that.
For the moment I had the task of providing a quick overview of what I thought was important in the next generation of RAHS, some of which was in the original vision. The three I came up stand as three general principles of any decision support system and I share them here:
The need to move from Induction to Abduction as a basic assumption
Induction assumes a case based approach, we can use the past to make statements about the future. Abduction, or the logic of hunches is all about finding the most plausible route between apparently unconnected things. As such it accepts the problem of samples of one or less which is the reality of managing for uncertainty and creating resilience. It is a very different design philosophy and it also means a necessary emphasis on advocacy, and the need for research to persuade decision makeers to act on the basis of coherence rather than proof.
Human sensor networks are key
You can't train people to overcome cognitive bias, you need to work with it. That means any system must be designed to enable the rapid deployment not just of data analytics, but also of mass human sensors who see things from different perspectives and who in aggregate will scan more. Models are more limited than such networks in their real time, fast feedback capability. Their use is also what I am starting to call a post-big-data approach; proactively creating data rather than just analysing what is there. Both are important, but there is too great a tendency to want to black box analytics.
Enable ordinary purpose for extraordinary need
Any system (which includes human sensor networks) has to have capability which is routinely used, which can then be activated quickly to create an ability to handle the unexpected. This is all about resilience and includes key concepts such as switching from trying to anticipate the future to triggering anticipatory awareness. It is the way that a Government or company can respond asymmetrically to asymmetric threat rather than the resource hungry symmetric responses that dominate too often.
Later in the day, as part of table 1 which joyously contained all the complexity experts and leading curmudgeons and cynics (not sure what the collective noun is here) we added some more:
- The need to create an object based architecture from which applications emerge through the interaction of people with software objects, it is not about application design.
- People need to be modelled in the design as identities (the human equivalent of objects). This is a big issue and I will blog more on this in the future.
- Metadata is more important than data itself, and is key to knowledge sharing between silos.
In effect you are designing an ecology, with genetic regulation (long term in the DNA, short term in the chemical activation deactivation) rather then drawing up blue prints for a bridge.