Using ecological models we can better understand and design:
For many years there has been a rich interplay between computing and biological models. Neurological models have been the inspiration for neural computing and also computing itself has been a key element in the development of modern cognitive science. Similarly evolutionary models have driven the development of genetic algorithms and programming, whilst computer simulations of genetics have been used to aid the understanding of biological evolution (for example Dawkin's 'insects').
More slowly, ecological thinking has found its way into thinking about computer systems and cognitive science. This is characterised by an emphasis on the interaction between people, computers and their environment.
The most obvious links are computer simulations of ecosystems such as James Lovelock's famous 'daisy world' models and the studies of interacting organisms in Artifical Life.
However, we can look more widely into the fields of HCI and CSCW and see the influence of ecological perspectives. Furthermore, we can exploit an understanding of natural ecology to drive the process of computation itself.
The workplace is an ecological setting where we interact with one another, with the aretfacts and with environment that surrounds us. This mode of thinking is strong in the CSCW literature beginning with work on distributed cognition (see works of Hutchins and Lave below) and situated action (Suchman) both of which focus, in different ways, on the importance of external and environmental influences in shaping work. Recent ethnographic studies have emphasised the importance of the ecology of the workplace; including whiteboards, calendars, individual papers and piles on desks (Herskind; Rouncefield; Sellen).
An ecological perspective of the workplace has been part of my own work, especially, my focus on interacting through the artefact in my CSCW framework and on the issues of long-term interaction. Ongoing theoretical work and case studies with Devina Ramduny, Julie Wilkinson and Janet Finlay have focused on the role of paper and other artefacts as the drivers, shapers and reminders for work processes. In particular, we have looked at the way environmental cues (such as the position of papers on a desk) act as the triggers that initiate action and in a sense are the place-holders of organisational processes. This has lead us to the 'socio-organisational Church-Turing hypothesis', that organisations have (amongst other things) an information processing function and therefore we should expcet to see analogies between the explicit and tacit aspects of organisations and computational artefacts.
Ecological imperatives can also be found in single-user interaction. Again distributed cognition and situated action have been important influences, recognising that people do not just think and plan, but act in interaction with their physical or virtual environment.
Ecological models have also been the inspiration for information foraging theory, developed at Xerox PARC by Peter Pirrolli and Stu Card. This uses existing mathematical models of the way animals forage for food to analyse the way users make decisions whilest information browsing. One of the key features of information foraging theory is the recognition that the process of searching for information is costly and that an optimal strategy involves a trade-off between the certain costs of searching against the potential, but uncertain, value of information yet to be found.
In retrospect I can see that my own work with Janet Finlay and Johnathan Hassell on the AMO (agents-medium-objects) framework had an ecological flavour. This concentrated on the importance of designing an environment for interaction (the medium) within which the user can manipulate passive objects (e.g. documents) and interact with intelligent agents or other users.
Finally, we may wonder whether a richer ecological viewpoint can help in the design of computer algorithms in the same way that neural networks and genetic algorithms have proved so powerful at tackling traditionally 'hard' problems. Indeed, there are signs in the genetic programming and artificial life literature where concepts of co-evolution and niches have been used.
On going work with Russell Beale is taking a radical route, making competition a central part of machine learning. Rules compete with one another in order to develop an ecosystem of rules which are not individually optimal over the whole data set, but between them give a better result than any indvidual could. We are calling this artificial ecosystems or eco-algorithms.
From this perspective we can look at neural network clustering algorithms, for example Kohonen networks and ART networks, and see that these too have some of the elements of an ecology - the clusters formed by the self-organising networks compete with one another to match data items.
My long-standing work (with others) on status-event analysis can also be seen as related to this perspective. Whereas event-based and also object-oriented paradigms focus on the direct interaction between agents, status-event analysis puts a strong emphasis on status phenomena which represent the elements of a shared environment within which agents interact.
Topics pages on status-event analysis and time.
A. Dix (2002).
Managing the Ecology of Interaction.
Proceedings of Tamodia 2002 - First International Workshop on Task Models
and User Interface Design, Bucharest, Romania, 18-19 July 2002
extended abstract
|| draft paper (PDF, 98K)
| http://www.hcibook.com/alan/topics/ecology/ | maintained by Alan Dix |