Artificial Intelligence for Computational Sustainability: A Lab Companion/Preface
Long-term planet sustainability requires the intelligent use of intelligent computational systems, at least if we accept that planet-transforming technology and over-population are here to stay, taxing human abilities to plan intelligently and with the long-view. This requires socially-engaged computational thinkers who can build and evaluate intelligent systems technology, informed by sustainability applications, and able to work across disciplinary lines.
Jeannette Wing (2006) first forwarded the concept of computational thinking, and it is a concept that has evolved. AI, the area of computing that is concerned with managing the non-determinism of decision making, is at the heart of any computational thinking curriculum that is to have societal relevance (Fisher, 2008), most notably for societal and environmental sustainability.
To engage computing students on sustainability problems doesn’t simply require an AI education, leaving to hope that they will later see and act on its relevance to sustainability, but the connections are best made when they are made explicit, else the knowledge may be left inert in the student (Bransford, et al., 1990), inaccessible even when context would seem to demand it. Making connections clear may also induce students and practitioners with sustainability concerns into computing, who might not have otherwise chosen to pursue it.
This text is motivated by a desire for increasing and sustaining the numbers and diversity of the computing community engaged in environmental and societal sustainability, and in particular is focused on introducing sustainability content into undergraduate courses on AI. The desirable characteristics of the evolving textbook are that it be
- portable, a supplement to any primary textbook and other resources used in an undergraduate AI course;
- online and freely available, for use in courses worldwide, as well as for use in broader impact plans (NSF, 2007) by research teams/projects;
- compartmentalized into self-contained sections and exercises, enabling instructors to easily “snip out” portions of the textbook for use in their courses;
- interlinked with AI-related Wikipedia articles, text-books (e.g., Poole and Mackworth, 2010; Wikibooks Community, 2012), online courses and lectures (e.g., Ng, 2011), online research papers, and other resources (e.g., RadioLab, AAAI ); and
- community-developed, evolving as projects, assignments, and explanatory material at the intersection of AI and sustainability are written into the text.
The project aligns with larger efforts on computational sustainability and it is expected that this effort will add to existing educational efforts of that community, highlighting and encouraging the flow of knowledge from research to education and public outreach.
In addition to describing the AI and sustainability lab text generally, the chapter discusses the implications of the project for integration of research and education, and communicating science to the public; and spotlights the community nature of the project, and the hope that the project will help attract young people to computing, through AI, sustainability, and community.
Wikibooks, a sister project of Wikipedia, is intended to support collective creation of textbooks for both formal and informal education. Like Wikipedia, Wikibooks’ infrastructure supports knowledge evolution, such as edit histories that allow credit assignment and rollback; and mechanisms to support discussion scope and revision, both on small points of disagreement as well as large plans for article merging and splitting. Wikibooks fall under a Creative Commons Attribution ShareAlike license, allowing copying and redistribution of all or part of the labbook, with attribution, to another platform, perhaps even to primary AI textbooks, which would be an ideal eventuality. Wikibooks allows stable pdf versions along a book’s evolutionary path. There are also broader implications to using Wikibooks, stemming in large part from its relationship to Wikipedia.
Communicating Science to the Public
One of the most important broader impacts of using Wikibooks is the possibility of involving more academics in Wikipedia, and communicating science and technology to the public, generally. Wikipedia’s popularity as a source for students and many others, faculty included (though typically not in their areas of expertise), make accurate and complete communication of scientific knowledge in the medium all the more critical. The Association for Psychological Science Wikipedia initiative (APS, 2012), for example, calls upon its members to contribute to Wikipedia, directly and through their students, to better insure completeness and accuracy, while also exercising these scientists and their associates on the critical skills of communicating science to the public. In fact, there are many "WikiProjects" (http://en.wikipedia.org/wiki/Wikipedia:WikiProject) that are designed to insure the quality of corresponding content area articles.
An informal survey at AI-related articles on Wikipedia show the coverage to be wanting, not so much incorrect as grossly incomplete, most of all in depth. Articles generally covering unsupervised learning, to take but one example, most notably clustering, are dominated by material from classic data clustering of statistics, with very limited coverage of uniquely machine learning and AI perspectives. Even the article on computational sustainability itself is little more than a stub. Moreover, informal query of undergraduate and graduate CS students show a near 0% participation of students as Wikipedians. In addition, formal studies (Lam, et al., 2011) indicate that women are still very under-represented as Wikipedians.
We can hope and expect that the sustainability text on Wikibooks will be a steppingstone for its contributors, largely self-selected to care substantially about educational and public outreach, to more broadly contribute to Wikipedia articles on AI. Some of these new contributors will be students.
Speaking to Authentic Audiences
There are also important broader impacts relating to formal pedagogy. Students turn in assignments for a grade, but they are more motivated to do quality work when the products of their labors are to be presented to what Light (2001) and others call an “authentic audience.” When working for authentic audiences, many students will pick up their game, as the work is perceived to be, and is in fact, more relevant (Bruff, 2011). Ideally, students who participate as Wikipedians will see their efforts as contributing to global pedagogy, both as it relates to sustainability and AI, and this will contribute to their self-image as people who can make a difference in society, in large part by working in community.
More broadly, universities are facing important questions of how to best use the World’s freely-available educational resources for a better onsite education – one that ideally fosters a commitment to place, a vital prerequisite to a commitment to sustainability. Just as importantly, and perhaps more so, are questions of how to contribute to the World’s pedagogical resources. Thus, an important motivation for the lab text is to give faculty an authentic audience to speak to as well as students; to give all a medium to contribute to the World’s freely-available educational resources, both for reasons of good World and scientific citizenship, as well as to maintain their vibrancy as educators in an environment that is increasingly global.
Overview of Lab Text Contents
At this stage the lab text is indexed primarily through AI topics, with secondary indexing and accessibility through sustainability topics.
- search (uninformed and heuristic),
- constraint-based reasoning and optimization,
- deterministic propositional and first-order inference,
- deterministic planning,
- reasoning and planning under uncertainty,
- machine learning (e.g., supervised, unsupervised, reinforcement),
- multi-agent systems, agent-based modeling,
- cognitive agent architectures,
and more. Across a large number of AI texts, it is not difficult to identify common units of content. Ideally, such indexing will serve to encourage incorporation of sustainability content into the primary AI textbooks themselves, as they are revised and new texts published. In addition, a primary organization based on AI topics will help to highlight sustainability-related problems that share similar problem structure, but which may be in very different sustainability domains, thus encouraging abstraction and generalization on the part of students and guarding against inert knowledge.
Types of Lab-Text Contributions
Sections of the lab text should include explanatory and illustrative material sufficient to orient readers on sustainability content and applications. Since the text’s primary organization is based on AI topics, this will likely lead to redundancies in the description of sustainability applications that cross AI subject lines. Such redundancies are likely to be acceptable, even desirable, for some time, but editors would likely merge such treatments in some cases, and perhaps split content that covers too many AI exercises in one section. Accompanying explanatory text on each sustainability topic, there will be various AI model assignments  of varying scope, in each case relevant to the sustainability topic:
- programming and/or paper projects from several weeks to a semester duration;
- programming and written assignments of a week or so; and
- in-class or other short-term exercises ranging from minutes to hours.
Throughout this material, there will be pointers to source material and other external references.
In addition to the dominant AI-based orientation, there is a final chapter of the lab text where sustainability problems can be laid out, with no or limited reference to AI concepts per se. The intent here is to enable students, perhaps as part of formal projects, to identify AI relevant approaches without biasing them as to what these approaches should be.
Examples of Lab-Text Contributions
There are many examples of AI for sustainability that can be placed in the lab companion. Numerous machine learning applications are amenable to translation to an educational setting, including biodiversity (Dietterich, 2009), climate tracking (Monteleoni, Schmidt & Saroha, 2010), electricity use (Gupta, et al., 2010) and distribution(Gross, et al., 2006) monitoring.
Like machine learning, optimization more generally is an area of AI with many pedagogical opportunities. Brunskill and Lesh (2010), for example, use optimization to find routes for developing-world health-care workers who must walk tens of miles a day to visit remote villages.
Another motivating application of optimization is corridor design. Conrad, et al. (2010) describe a new paradigm of optimization under budget constraints, one that is quite intuitive to CS students. The authors apply their methods to grizzly bear corridor design, finding paths between existing protected ecological reserves to allow for bear population mobility and increasing possibilities of genetic diversity. Much of the paper itself is in reach of upper-division computer science students, and was required reading in an upper-division special topics course for CS majors on Computing and the Environment. The problem, as laid out in the paper, invites experimentation with different approaches. One student’s end-of-semester project implemented a related corridor optimization strategy from Williams and Snyder (2005), replicating the paper’s experimental results.
Experience indicates that corridor design is engaging to undergraduates and that data sets and environments can be created for compelling projects and assignments. The Reserve Design Game (Rochester and Possingham, 2008) and the Chesapeake Bay Game are examples of other pedagogical material that could be folded into the lab text in this area.
Importantly, these examples of AI for sustainability are but a few—we should not be surprised if the projects, assignments, and exercises that could be placed in this text are coextensive with the 'cross-product' of AI subject areas and Sustainability problem areas!
Deep Infusion of Sustainability
There are courses, some with varying material on the Web, that are focused particularly on AI and sustainability, or computing and sustainability (e.g., "Sustainability and Assistive Computing" at http://cs.brynmawr.edu/Courses/cs380/fall2010/; "Artificial Intelligence for Health and Sustainability" at http://www.cs.cmu.edu/~ebrun/spring2012_ai.pdf; "Computing and the Environment" at http://www.vuse.vanderbilt.edu/~dfisher/socially-engaged-computational-thinkers/ComputingEnvironment.pdf).
In contrast to the very positive trend towards sustainability-themed courses, this lab companion is an example of deep infusion of sustainability material into NON-sustainability-themed courses. Indeed, the intent of this effort is not to displace one bit of AI content from AI courses, but rather to facilitate the easy adoption by instructors of sustainability problems that can be used to motivate AI methods and concepts, while passing on sustainability knowledge in the process. There is no reason that this strategy of deep infusion could not be used in other computing courses, ranging from computer architecture to database, to name but two.
Integration of Research and Education
Generally, inspiration for educational material can come from research papers in AI that address sustainability issues; these papers often spend a fair amount of time explaining the AI relevance to a larger sustainability audience. Conrad, et al. (2010) is but one example of research of interest in AI and optimization circles per se, but the presentation is at multiple levels of intuition and formality for the benefit of non-AI audiences too.
AI in sustainability research papers, because they are framed for larger audiences, often with a particular application in mind, are ideal jumping off points for translation of research into educational material. The Institute for Computational Sustainability (ICS, 2012) has promising examples, some of which are highlighted herein.
Ideally, researchers and educators who follow such a path will continually contribute to the Wikibooks lab text. In addition, a hope is that the evolving lab text will be a resource for broader impact and education plans of research projects, such as NSF proposals and awards, in large part because this opportunity will be promoted, perhaps going so far as to provide templates for such activities on the lab book site, as well as virtual-meeting tutorial sessions on Wikibooks editing. This will be but one semi-formal way of encouraging participation in the sustainability lab text project.
The ideal design is of a community-written and edited lab text that complements AI undergraduate courses on the one hand, as well as existing efforts on computational sustainability on the other. The text is intended as a mechanism for infusing sustainability into AI curricula and texts, the start of a larger goal of infusing computational sustainability into the computing curriculum generally. There are important broader impact motivations too, notably to foster integration of research and education, better communication of science to the public, and presenting opportunities to students and faculty for contributing, in community, on a global stage. This lab text can be edited by anyone, and there are no areas so in need of community participation as societal and environmental sustainability.
- Wing, J. (2006) Communications of the ACM, March 2006, Vol. 49, No. 3, pp. 33-35, retrieved from http://www.cs.cmu.edu/afs/cs/usr/wing/www/publications/Wing06.pdf February 17, 2012
- Fisher, D. (2008) AI and Developing Socially-engaged Computational Thinkers. In AAAI Spring Symposium on “Using AI to Motivate Greater Participation in Computer Science” Retrieved from http://www.vuse.vanderbilt.edu/~dfisher/socially-engaged-computational-thinkers/
- Bransford, J. D., Sherwood, R. D., Hasselbring, T. S., Kinzer, C. K., & Williams, S. M (1990). Anchored instruction: Why we need it and how technology can help. In C. Nix & R. Spiro (Eds.), Cognition, Education and Multimedia: Exploring Ideas in High Technology (pp. 115-141). Hillsdale, NJ: Lawrence Erlbaum Associates.
- NSF (2007). Merit Review Broader Impacts Criterion: Representative Activities, http://www.nsf.gov/pubs/gpg/broaderimpacts.pdf
- Poole, D. and Mackworth, A. Artificial Intelligence: Foundations of Computational Agents, Cambridge University Press is freely available on the Web (http://artint.info/index.html)
- Wikibooks, Artificial Intelligence, retrieved February 17, 2012 from http://en.wikibooks.org/wiki/Artificial_Intelligence
- Ng, A. 2011. Machine Learning video lectures. Retrieved from http://www.ml-class.org/course/video/preview_list
- RadioLab podcast, for example http://www.radiolab.org/2011/may/31/
- AAAI Resources for Educators, retrieved from http://aaai.org/AITopics/EducatorResources
- Gomes, Carla. 2009. "Computational Sustainability: Computational Methods for a Sustainable Environment, Economy, and Society." The Bridge 39(4):5-13.
- APS (2012). Association for Psychological Sciences Wikipedia Initiative: retrieved February 2012 from http://www.psychologicalscience.org/index.php/members/aps-wikipedia-initiative
- Lam, S.K.; Uduwage, A.; Dong, Z.; Sen, S.; Musicant, D.R.; Terveen, L.; Riedl, J., WP:Clubhouse? An Exploration of Wikipedia’s Gender Imbalance. WikiSym 2011, Mountain View, CA, October, 2011
- Light, R. (2001). Making the most of college: students speak their minds, Cambridge, MA: Harvard University Press
- Bruff, D. (2011). A Social Network can be a Learning Network http://chronicle.com/article/A-Social-Network-Can-Be-a/129609
- Russell, S. & Norvig, P. (2010). Artificial Intelligence: A Modern Approach (Third Edition). Prentice Hall, NJ
- Luger, G. (2009) Artificial Intelligence: Structures and Strategies for Complex Problem Solving, 6th Edition. Addison-Wesley
- EAAI Model Assignments (2012), retrieved from http://modelai.gettysburg.edu/
- Dietterich,T. (2009). Machine Learning in Ecosystem Informatics and Sustainability, International Joint Conference on Artificial Intelligence, also http://videolectures.net/ijcai09_dietterich_mleis/
- Monteleoni, C., Schmidt, G. A., Saroha, S. (2010). Tracking Climate Models. CIDU, pp. 1-15
- Gupta, S., Reynolds, M.S., and Patel, S.N. "ElectriSense: Single-Point Sensing Using EMI for Electrical Event Detection and Classification in the Home," Proc. Conf. Ubiquitous Computing (UbiComp), ACM Press, 2010, pp. 139–148.
- Gross, P., et al., "Predicting Electricity Distribution Feeder Failures Using Machine Learning Susceptibility Analysis," Proc. 18th Conf. Innovative Applications of Artificial Intelligence (IAAI06), AAAI Press, 2006, pp. 1705–1711.
- Brunskill, E. and Lesh, N. (2010). Routing for Rural Health: Optimizing Community Health Worker Visit Schedules. In AAAI Spring Symposium on AI and Development, March 2010, Palo Alto, CA
- Conrad, J., Gomes, C., van Hoeve, W., Sabharwal, A., and SuterConrad, J. Incorporating Economic and Ecological Information into the Optimal Design of Wildlife Corridors, Computing and Information Science Technical Reports,URI:http://hdl.handle.net/1813/17053. Cornell University, Ithaca, NY, 2010
- Williams, J. C., and Snyder, S. A. 2005. Restoring habitat corridors in fragmented landscapes using optimization and percolation models. Environmental Modeling and Assessment. 10(3):239–250
- Rochester, W. and Possingham, H. (2008). The Reserve Design Game. http://www.uq.edu.au/marxan/resgame/index.html#intro
- Learmonth, G., Smith, D. E., Sherman, W. H., White, M. A., and Plank, J. (2011). “A Practical approach to the complex problem of environmental sustainability: The UVA Bay Game,” The Innovation Journal: The Public Sector Innovation Journal 16 (1) http://www.innovation.cc/scholarly-style/learmonth_sustain_inviroment_v16i1a4.pdf
- ICS (2012). Institute for Computational Sustainability, retrieved from http://www.cis.cornell.edu/ics/projects/overview.php, February 2012