methods for positive change at both the societal and
individual levels, exploring a variety of topics in
computational sustainability (Gomes 2009) and assistive technologies. As in the robotics course, students
developed a semester project in small teams, but with
a focus on having an impact to society through the
requirement to work with an external organization.
Although not strictly a course in integrated AI,
many of the topics and the semester team projects
involved the integration of multiple AI techniques.
The course covered the topics listed in table 2.
To ensure a balanced discussion between computational sustainability and assistive computing
throughout the semester, the course schedule interleaved topics from both categories. In addition, the
course reviewed the underlying AI methods and discussed project development (proposals, Heilmeier’s
Catechism, and other topics), presentation skills
(including talk and poster design, and elevator pitch-es), and how to work with external organizations. As
in the advanced robotics course, classes were a mix of
lectures by the instructor or guest speakers, student
presentations of research papers, seminar discussions
on the readings, and in-class workshops on the
The course project was more open than the service
robotics project, but placed a strong emphasis on
developing a project with a tangible impact to society.
Besides encouraging students to explore the social
dimensions of their work, this emphasis provided
strong motivation to students. Toward this goal, each
team was required to work with an external organization on their chosen project. These external organizations were not chosen ahead of time, which
allowed students to experience the full challenge of
launching and maintaining an external collaboration.
One team developed an ASL-to-Text chat program
that would recognize and transcribe a limited subset
of American Sign Language (ASL) visual signs into
text, integrating techniques from computer vision,
machine learning, and accessible interface design. To
ensure that the project would meet the needs of the
deaf community, the team worked with staff in the
ASL Program at the Penn Language Center at the Uni-
versity of Pennsylvania. Another team worked with
contacts at the Pennsylvania Game and Fish Com-
mission to develop a set of educational games on the
dangers of overfishing, combining mathematical
population models with maximum entropy models
learned from data. Since each team developed an
independent project from scratch, unlike in the
robotics course, this course provided no scaffolding
for the projects.
The sustainability and assistive computing projects
had significantly smaller scope than the robotics
projects, but the chance to address real societal and
environmental problems sparked the students’
enthusiasm. As with many collaborations, the teams
found it challenging to maintain their connections
to the external organization. They also experienced
the difficulty in working with raw data provided by
these organizations, with data cleaning and pre-pro-cessing becoming a major factor in obtaining good
results, as it is in many machine learning applications.
I find project-driven courses to be extremely reward-
ing, both for the students and as the instructor. In all
cases, the largest challenge is helping students to get
started quickly in the project, which would seem to
require scaffolding and more closed requirements.
However, in these types of courses, I believe that one
of the worst mistakes an instructor can make is to
restrict the project. Instead, leave it open ended and
encourage them to impress you — let their creativity
take over, give them the chance to push the project
as far as possible, and see how far they can go. I pre-
Computational Sustainability Assistive Computing
Species distribution modeling* Intelligent wheelchairs, smart prosthetics, and assistive
Electronic waste and green technology Smart home monitoring for patient and elderly care*
Sensor placement in water distribution
Human-computer interfaces for people with disabilities*
Ef;cient power and biofuel usage Assistive technologies and predictors of technology
Food and farm optimization*
Telemedicine and medical informatics*
Table 2. Course Topics.
Those topics marked with asterisks (*) included significant AI components.