Kaleidoscope: Programming ideas

The idea for Kaleidoscope is to build a learning interface where the joint environment of user, interface and digital dimension are considered as one intelligent place.

In programming this we are developing a disjointed architecture which allows the individual programs to be combined independently. This allows us to explore the programs in respect to the follow up projects.

Kaleidoscope is designed as a four part program:

Part 1:Tracker	Part 2:Behaviour Interpreter	Part 3:Dynamic Feedback Mapper	Part 4:Dynamic Interactive Environment
The interface can essentially be any interface giving a stream of input to the Behaviour Interpreter	1) Isolates and defines moment in the data as recognised 'actions'. 2) Defines a (finite) list of recognised behaviours which is dynamically updated over time by means of the data sent from the Dynamic Feedback Mapper.	1) Creates dynamic mappings to the output parameters of the Dynamic Interactive Environment. 2) Defines the fitness function and feeds it back to the Behaviour Interpreter. The fitness function is determined by the users repetition of action: the more an action is repeated the more asserted it becomes.	1) The digital environment is defined through a set of parameters for action. The parameters are open and combinable. 2) The Dynamic Interactive Environment renders the environment.
Passes on: raw tracking data	Passes on: defined behaviours	Passes on: Fitness function for the Behaviour Interpreter and mapping of behaviours to output parameters for Dynamic Interactive Environment	Passes on: visualisation of the environment
	Receives: Fitness function from Dynamic Feedback Mapper
In the first stage this will be the screen tracker from Spawn	In the first stage we are conceiving this as using a fuzzy pattern recognition.	The most important thing for the Dynamic Feedback Mapper is the notion of tunability. It is essential that the dynamic mappings can be tuned so as to give the user a more or less intuitive experience.	In Kaleidoscope the rendering will happen in passive stereoscopic projection.

Further requirements for the program is that the fuzzyness of the Behaviour Interpreter and the mapping of the Dynamic Feedback Mapper can be adjusted and tuned. It will be essential for the project that the looseness by which the environment learns about the user's actions can be adjusted, allowing for more or less Ôintuitive' pre-setups to define the learning curve of the user.

Ideas

Forgetting: that the system forgets actions which are not repeated

Exaggeration: the idea that as actions are repeated they become more and more exaggerated - allowing the user to sense of a firmness to the defined behaviour.

Randomness/noise: that the system remains fresh by holding a certain sense of randomness to the interaction

Knowledge transfer: that the system remembers actions from past users?

Soft interaction: the idea of fuzzyness

Ambient interfaces/Spatialised Media

Data muttering: just really like the word muttering - the idea that the data is happily conversing with its self

Theory

The idea that the actions are affirmed through the system - creating an environment for action defined through a mutual learning process sited between the user and the environment itself. This reciprocal process allows the environment to come into existence through the actions of the user.

The question is how is this framed: in Kaleidoscope it is like a toy - but also not very culturally engaging.
How can it engage the body: what are the other interfaces that we could build (do you move on the surface?)

Questions

What is the difference between our fuzzy pattern recognition combined with the Dynamic Feedback Mapper and reinforcement learning?

What is the difference between genetic programming and fuzzy logic: do we want it to be genetic?