Ambient intelligence for scientific discovery
Extended abstracts of the 2004 conference on Human factors and computing systems - CHI '04
OBJECTIVES Human-Computer Interaction in scientific discovery has been gaining a momentum since the last decade due to its potential payoffs in applications, such as drug discovery, nano-materials or telecommunication. In HCI history, many advanced human interfaces were invented by members of the scientific community, such as the World Wide Web from CERN and the "Glass Cockpit" from NASA. With the growth of data streams and complexity of discovery tasks, there are demands for ambient
... e such as peripheral vision, soft agents, data mining, and semantic web. The goal of this one-day workshop is to gather interdisciplinary scientists and HCI professionals under one roof and explore ambient intelligence interfaces for the new needs in discovery. The expected results include a proceeding of selected papers, a position summary that states the current technologies and philosophy, and workshop for coming years. The workshop is focused on three themes: 1) information reduction, 2) novel representations, and 3) discovery networks. INFORMATION REDUCTION Modern science and data communication technologies produce more data than can be analyzed by all humans together on Earth. For example, with the growing number of satellites and improving spatial and spectral resolutions, the data down-link rate will be 1 gigabit per second in 2010 and up to 10 gigabit per second by 2020. Much of this data is redundant and irrelevant to specific purposes, and can be classified as noise. Scientists spend up to 70% of their time on preprocessing the imperfect, redundant and noisy data. In many instances, it is not yet known how to distinguish noise from real signals. This is also the case in Bioinformatics, where protein sequences need to be interpreted with respect to their structure, dynamics and function, but to distinguish an important building block in the protein sequence from simple filling material is an unsolved mystery in biology. Furthermore, medical doctors are faced with new drugs on a regular basis and the amount of biomedical knowledge that could be useful for diagnosis and treatment is increasing, but to link the important pieces of information together that would result in actual improvement in human health is not yet possible. Delays during the analysis and lack of sufficient insight into the nature of a problem hinder the discovery prediction of disastrous situations or significant physical features in all of these and other examples. In addition, the overflow of information distributed over the Internet might be the cause for its end if the problem is not efficiently addressed. We aim to reduce the information at the source side. Thus, we will share the experience from developers of ambient interfaces for data representation, semi-autonomous discovery agents, semantic data compression, feature indexing, and navigation tools for browsing large databases, knowledge or model based information understanding and information reduction. NOVEL REPRESENTATIONS Scientific discovery requires multidisciplinary insight and serendipity from routine work, i.e. the capability to make non-obvious connections between the complex interactions of the components of these systems. Such insightful solutions can often be found in an interactive and visual problem solving environment, as demonstrated for example by the fact that despite the modern numerical computing technologies, scientists today still use Gedanken experiments for concept development.  It is striking that simple intuitive simulation is still one of the most powerful approaches to creative problem solving. New ways of creating such interactive and visual problem solving environments are needed.