Why Artificial Intelligence and Sustainable Design?

Long-term environmental and societal sustainability requires that artifacts, materials, systems and processes be designed to minimize energy and waste and to maximize reuse and utility; we should hope that the days of designing things that are ultimately thrown away are rapidly coming to an end. The ‘design for X’ paradigm considers downstream objectives, such as reusability, early in the design process. Designers are being challenged to consider factors that had been previously given little attention, like life cycle costs along many dimensions; including energy requirements during manufacture, use and end-of-use phases, and material loss and environmental damage at the end of a product’s life. A vision for sustainable design is cradle-to-cradle design (McDonough and Braungart, 2002), in which products are designed and built in ways that enable full reuse at low costs (e.g., energy), with nothing thrown out and nothing degraded. Our motivation to organize the symposium stemmed from our presumption that the increased complexity of design necessitated by a desire for very long-term planet sustainability requires application of and advances in artificial intelligence.

This blog is a continuing discussion that started at the AAAI Spring Symposium on Artificial Intelligence and Sustainable Design at Stanford University from March 21-23, 2011. The symposium focussed on domains, problems and challenges of sustainable design and the role that AI can play in sustainable design.

The symposium brought together researchers from three primary areas:
• AI and Design
• Computational Sustainability
• Design for Sustainability
AI and design is an established field already, with conferences and journals; AI has provided computational approaches to design processes and the representation of design knowledge. Design for sustainability is not so much a field as it is a set of principles, which are implemented on an ad hoc basis. Computational sustainability is a nascent field, but already influential in moving AI and other computational fields into addressing sustainability questions.

About 25-30 people physically attended the symposium. There was also a virtual participation option, which was made available to co-authors, colleagues, and students of the authors of the papers as a way of broadening participation without requiring additional travel — and as a result about 10 avatars attended in Second Life at various times. A total of 18 papers were presented over the course of the two-and-a-half days. Presentation topics included smart buildings, environmentally smart occupants, decision support systems to support energy analyses, optimization to inform design decision making, and theoretical models of cradle-to-cradle design.

The symposium included two invited talks. Alice Agogino (University of California, Berkeley, Department of Mechanical Engineering) described the Smart Lighting project – a wireless sensor network for customizable commercial lighting control. This application requires decision making in the face of uncertainty, with needs for system self-configuration and learning. Kirstin Gail (Autodesk) spoke on Bloom, a recyclable laptop developed by graduate students of a Stanford University course on design innovation; Bloom represents a new class of electronic products that can be easily disassembled for recycling by the consumer at end of product life. In both of these talks, the speakers noted that sustainability was only achievable by considering human behavior as a significant factor. In fact, this recognition of the importance of social influences and implications of human behavior was an overarching theme of the symposium generally.

There were several major sustainability themes in the papers and discussions, including energy conservation, material recycling, lifecycle and environmental impact analysis. The AI and computational themes included approaches to modeling sustainable design knowledge and processes, sustainability through computational support for biologically inspired design, characterizing and managing uncertainty, machine learning and case-based reasoning for sustainable design, and computational tools for estimating sustainability costs in design alternatives.

Participants also engaged in three separate breakout groups to brainstorm on forward-thinking research directions at the intersection of AI and sustainable design. Many themes were discussed, including (1) the role of AI in modeling the effect of human behavior on sustainable design of artifacts and systems; (2) the role of AI in biologically inspired sustainable design, and (3) the role of customization in fitting designs to individuals and groups in a manner that improves sustainability indices.

The Symposium was successful in achieving a common interest among the participants, which we hope will be maintained through a continued effort to communicate and share research on this topic. Our expectation is that AI will provide formal models, languages and methods of sustainable design, thus helping to push sustainable design from the status of interest area into the status of a discipline.

Douglas Fisher and Mary Lou Maher were co-chairs of this symposium. The papers of the symposium were published as AAAI Press Technical Report SS-11-01.

Douglas Fisher is an associate professor at Department of Electrical Engineering and Computer Science, Vanderbilt University.

Mary Lou Maher is a senior research scientist at the Human-Computer Interaction Lab, University of Maryland.

Reference: McDonough, W. and Braungart, M. (2002) Cradle To Cradle: Remaking The Way We Make Things, North Point Press, 2002.