Sorensen

This paper argues for the rich perspectives in applying an interactivist design heuristics on issues of development and organization of Ambient Intelligence (AmbI). It sketches design principles for AmbI on the basis of a general outline of the interactivist model, focusing on adaptive dynamics. Lastly the discusses implementational issues for future interactivist AmbI research.

New dynamic IT: AmbI

Designing IT is facing more challenges than ever with increasingly pervasive and complex IT systems which have to get les fragile and more self-maintaining. Besides an increasing interweaving of IT into most everyday practices calls for more adaptive assistance. Different poles of the IT research spectrum calls for a change of design strategy for future IT. Some related to maintaining vast pervasive IT systems (cf. IBM’s Autonomic Computing project[1] ) and others to the need for IT that supports increasingly dynamic user practices (Ducatel, K. et al., Harris and Henderson 2000, Kaasgaard forthcoming, Rheingold 2002).

To meet both ‘internal’ (infrastructural and organizational) and ‘external’ (functional) improvements, IT has to be designed with adaptive capabilities – sometimes referred to as smart or intelligent technology. In accordance with the growing embeddedness of IT into more parts of daily life I will refer to this new technology as AmbI.[2]

To develop AmbI the only venue is to study phenomena showing intelligent characteristics, i.e. natural adaptive systems. The paper argues for a bio-mimetic design heuristic for AmbI, with a specific focus on organizational and developmental issues provided by an interactivist approach. Interactivism provides powerful tools for AmbI design combining optimized infrastructural organization and improved functionality. Primarily by explaining how more adaptive assistance in AmbI rises in concert with infrastructural and functional self-organizational capacities.

Why Bio-mimetic design heuristics?

There is several reasons why a general bio-inspired tendency has grown within IT-research and it is worth taking a look at some of these.

A looming issue is development and maintenance of vast complex IT systems. Nature is the only domain for self-maintaining and recursively adaptive dynamics we know of, and since we already have got sciences – traditionally biology but also transdisciplinary fields such as dynamic and complex systems theory - concerned with the organization of complex adaptive systems it is instructive to look to some of the models from fields inspiration for design.

Theories such as interactivism use the same tools to describe individual constituents and global complex systems. Such theories harbors great scaling and nesting capacity which is especially important in order to avoid disintegrated views. Not only does the design of technology depend on an understanding of interacting systems on different levels, from devise-device to whole networking societies. The systems themselves has to be designed to facilitate the rapidly changing practices and mobile, long-distance and trans-media interaction which characterizes use of IT.

Besides humans have evolutionary constrained cognitive capabilities and cannot fully overview very complex design task. Acknowledging this in our design ideology we should ally ourselves with some of nature’s benevolent pattern creating principles. An adaptive dynamic technology would provide ‘design for free’ because artifacts will evolve and develop in response to interactive use. AmbI will thus create genuine functionalities bottom up.

Self-organization in complex systems: An interactivist approach[3]

The self-maintaining robustness and developmental capabilities inherent in adaptive systems [4], stems from the structural and functional self-organization of the systems. Self-organization emerges as a consequence of the dynamic interaction of the system constituent following onboard value-guided navigation. Values rises in natural adaptive systems because they are thermodynamically non-equilibrium and hence depend on a controlled input of energy and information to maintain their functional organization. This openness creates a bias; some environmental features facilitate self-maintaining interaction others do not. Outcomes of interactions – success or failure – give rise to construction of internal indicative constraints for subsequent actions.

Adaptive complex systems has different means to adapt, facilitated by both short and long term dynamics. In short terms systems adapt by learning. Learning capabilities are necessary in adaptive systems since the niches they inhabit are not stable nor are their innate responses flawless. Learning is a process towards improving the anticipatory capabilities in the system which is contextual and implicit in the beginning, but becomes increasingly generic as the system constructs better anticipative models for managing interaction.

Adaptive systems also (as species) improve adaptability by procreative means. In reproduction, combinatorial (sexual) or/and reconfigurable (mutational) possibilities for adaptation arises which can travel farther in the fitness landscape. This is an advantage if the niche has changed drastically or just to explore better spaces in the fitness-landscape.

Both short and long term adaptation rest on the same ordering principle namely variation and selection cycles. This principle is mostly known in phylogenic processes, but it is also the driving force in ongoing interactive trial and error cycles in ontogenetic processes (Bickhard and Campbell in press).

AmbI: Interactivist architectural principles

According to interactivist design principles AmbI will be organized as a kind of biosphere constituted by heterogeneous artifacts striving for self-maintenance by interacting with their environment and occupying functional niches. A tight functional coupling between users and artifacts will be obtained through recursive variation and selection cycles providing optimized assistance. All AmbI devices carries onboard value systems which are organized in levels regarding e.g. access to eternal computing support, information, power supply and perhaps some more abstract ‘value units’. The overall normative constraint for AmbI artifacts is getting positive feedback from their functional niche, including users and other artifacts. AmbI will gradually exploit mutual supportive or even symbiotic organizational dynamics in pursue of self-maintaining feedback.

Variation will be provided by evolutionary computing methods with an interactivist ingredient. Since disposable or reconfigurable hardware has not been developed enough to render lethal mutations costless (other than at a strict software-level), variation must facilitate developmental dynamics by mutations that is within an acceptable frame of deviance. Dysfunctional mutations are mostly unacceptable, (especially in critical situations), however mutation is an important part of the adaptive dynamics. The interactivist model provides tools for designing constrained but open-ended variation for ontogenetic variation by way of functional ‘themes’ (Bickhard and Richie 1983). Themes are interactive aspects (not components) which together form relevant functionalities in given contexts. Trials happen within frames of creative but relevant outcomes dynamically determined by functional themes created by previous interactions.

At the ontogenetic level themes will be active in learning. To speed up evolution and to maintain a lower tolerance than in natural variation for the sake of reliable functionality, learning functions could be applied in selections at the phylogenetic level as well, excluding mutations that showed lethal or dysfunctional beforehand.

Since artifacts are not thermodynamic non-equilibrium systems[5], value constraints will not emerge autonomously but will have to be imposed as fitness-functions. Within evolutionary robotics, fitness functions has been applied successfully enabling arbitrary ’genotypes’ to spontaneously develop light-seeking (recharging) behavior over generations (Floreano & Mondada 1998). Interestingly few pre-coded ‘genes’ (lines of code) with many combinatorial possibilities seem to provide the best basis for adaptive evolution. The task of designing ‘genotypes’ for various AmbI devices can probably be handed over to the evolutionary dynamics.

Selection frequencies will differ for different kinds of devices, such that micro- or swarmbased services (e.g. communication ‘scouts’ handling access and optimal bandwidth) will have a shorter lifespan than macro-services (e.g. OS’s). Roughly lifespan will be proportional to length of ‘genom’ and infra-structural complexity.

Services will be nested such that e.g. an OS acts as niche for simpler micro-services, struggling to enter a symbiotic cooperation with the OS. Users will act as niches for certain AmbI services and they will therefore stick to this (type of) user perfecting a tight adaptivity while others will inhabit more generic niches. This division of labor provides an extremely dynamic technology filling out every functional niche on the fly.

References:

Bickhard, M. H.: Error Dynamics: The Dynamic Emergence of Error Avoidance and Error Vicariants. Journal of Experimental and Theoretical Artificial Intelligence, 13 , 199-209. 2001a

Bickhard, M. H.: Function, Anticipation, Representation. In D. M. Dubois (Ed.) Computing Anticipatory Systems. CASYS 2000 - Fourth International Conference. (459-469). Melville, NY: American Institute of Physics. 2001b

Bickhard, M. H., Richie, D. M.: On the Nature of Representation: A Case Study of James Gibson's Theory of Perception. New York: Praeger Publishers. (1983)

Bickhard, M. H., Terveen, L.: Foundational Issues in Artificial Intelligence and Cognitive Science - Impasse and Solution . Amsterdam: Elsevier Scientific. 1995

Bickhard, M. H., Campbell, D. T.: Topologies of Learning and Development. New Ideas in Psychology, 14 (2), 111-156. 1996

Bickhard, M. H., Campbell, D. T.: Variations in Variation and Selection: The Ubiquity of the Variation-and-Selective-Retention Ratchet in Emergent Organizational Complexity. Foundations of Science. In press.

Christensen, W. D., Bickhard, M. H.: The Process Dynamics of Normative Function. Monist, 85 (1), 3-28. 2002

Christensen, W. D., Hooker, C. A.: An Interactivist-Constructivist Approach to Intelligence: Self-Directed Anticipative Learning. Philosophical Psychology, 13(1), 5-45. 2000a

Christensen, W. D., Hooker, C. A.: Autonomy and the emergence of intelligence: Organised interactive construction. Communication and Cognition - Artificial Intelligence vol. 17 no. 3-4, pp. 133-157, 2000b

Christensen, W. D., Hooker, C. A.: Self-directed agents. In MacIntosh, J. (Ed.) Contemporary Naturalist Theories of Evolution and Intentionality ,Canadian Journal of Philosophy, 31, Special Supplementary Volume. (in press)

Ducatel, K. et al.: Scenarios for Ambient Intelligence in 2010. Report for the European Commission Community Research. IPTS Seville (2001)

Emmeche, C.: The garden in the machine. The emerging science of artificial life. Princeton University Press, Princeton. 1994

Floreano, D., Mondada, F.: Evolutionary Neurocontrollers for Autonomous Mobile Robots In Neural Networks, 11(7-8. (1998)

Harris, J., Henderson, A.: Evolution in Action: HCI in a World of Pliant Systems. http://www.pliant.org/Papers-Area.html

Kaasgaard, K.: Beyond Formalisms. The Art and Science of Designing Pliant Systems. Interview with Jed Harris and Austin Henderson. In Kaasgaard, K.: Genres of Usability. Forthcoming http://www.pliant.org/Papers-Area.html

Nolfi, S., Floreano, D.: Evolutionary Robotics: The Biology, Intelligence, and Technology of Self-Organizing Machines. MIT Press/Bradford Books. Cambridge 2001

Rheingold, H.: Smart mops – the next social revolution. Perseus Publishing, Cambridge 2002

1 http://www.research.ibm.com/autonomic/

2 AmbI denotes the interesting aspects by future IT: 1) In opposition to Pervasive or Ubiquitous Computing it centers on functionality – intelligent assistance – and not some means to obtain it, while still denoting the highly distributed nature of IT. 2) ‘Ambient’ indicates the right sort of ‘non-intruding but present at hand’ assistance aimed for, captured under the slogan ‘If there’s to be computers everywhere they’d better get out of the way’. 3) ‘Intelligence’ not only denotes the behavior but the intrinsic characteristics of recursively adaptive systems, i.e. their structural and functional development and organization.

3 The interactivist approach I will use is in resonance with e.g. Bickhard 2001a, 2001b, Bickhard and Campbell in press, Bickhard and Richie 1983, Bickhard and Terveen 1995, Christensen and Bickhard 2002, Christensen and Hooker 2000a, 2000b, in press.

4 There are simpler forms of complex systems than recursively adaptive ones which are the focus here.

5 All though they most likely will depend on electricity, they will not disintegrate by lack of energy. Besides the process is reversible so that ‘dead’ cell phones can be recharged.